analysis_sbermarket_landing.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "moving-hunter",
   "metadata": {},
   "source": [
    "# Анализ эксперимента лендинга Сбермаркета\n",
    "**Задание аналитика в [Карьерный цех](https://analytics.careerfactory.ru/)**"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "tracked-summary",
   "metadata": {},
   "source": [
    "## Описание\n",
    "\n",
    "На текущем варианте лендинга пользователь сначала выбирает ритейлера, переходит в каталог, ищет нужные ему товары и только потом может узнать, что по его адресу выбранного магазина может не оказаться. Проведён эксперимент (A/B тестирование) с двумя вариантами лендинга.\n",
    "\n",
    "**Исходные данные:**\n",
    "\n",
    "Данные состоят из файлов, полученных из разных источников:\n",
    "\n",
    "- `AB Test Hit.csv` — пользователи и распределение по группам эксперимента;\n",
    "- `Landing Viewed.csv` — время захода пользователя на лэндинг;\n",
    "- `Main Page Viewed.csv` — время захода на страницу ретейлера;\n",
    "- `Address Change Initiated.csv` — время открытия модалки ввода адреса;\n",
    "- `Address Not In Delivery Zone.csv` — время получения сообщения \"Адрес вне зоны доставки\";\n",
    "- `Shop Celection Started.csv` — время открытия модалки выбора магазина;\n",
    "- `Shop Selected.csv` — время успешного выбора магазина;\n",
    "- `Product Added.csv` — время фактического добавления товара в корзину;\n",
    "- `Add To Cart Clicked.csv` — время попытки (клика) добавления товара в корзину;\n",
    "- `Order Completed.csv` — время оформления заказа.\n",
    "\n",
    "**Цели:**\n",
    "- проанализировать дизайн эксперимента и предложить идеи для улучшения. Необходимо проверить гипотезу, что выбор адреса до этапа выбора магазина поможет избежать “кривых” сценариев без повышения отказов с самого лендинга;\n",
    "- по метрикам bounce rate лендинга (в тестовых группах он не должен стат значимо вырасти) и конверсии добавления товара в корзину принять решение какой вариант лендинга лучше.\n",
    "\n",
    "_bounce rate - это термин веб-аналитики, обозначающий процентное соотношение количества посетителей покинувших сайт или лендинг прямо со страницы входа или просмотревших не более одной страницы сайта._"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "round-steam",
   "metadata": {},
   "source": [
    "### Структура проекта\n",
    "* [1. Загрузка и изучение общей информации датасета](#start)\n",
    "* [2. Предобработка данных](#preprocessing)\n",
    "* [3. Исследовательский анализ данных](#eda)\n",
    "* [4. Расчёт метрик (bounce rate лендинга и конверсию)](#metric)\n",
    "* [5. Диагностика дизайна эксперимента](#experiment)\n",
    "* [6. Выводы](#conclusion)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "blond-butterfly",
   "metadata": {},
   "source": [
    "<a id=\"start\"></a>\n",
    "## 1. Загрузка и изучение общей информации датасета"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "joint-chart",
   "metadata": {},
   "source": [
    "### Импортируем необходимые библиотеки"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "yellow-cannon",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Pandas version: 1.2.4\n",
      "Seaborn version: 0.11.1\n"
     ]
    }
   ],
   "source": [
    "import re\n",
    "import math\n",
    "from IPython.display import display\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "\n",
    "print(f\"Pandas version: {pd.__version__}\")\n",
    "print(f\"Seaborn version: {sns.__version__}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "least-nightlife",
   "metadata": {},
   "source": [
    "### Загрузим данные"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "equipped-granny",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "./dataset/\n"
     ]
    }
   ],
   "source": [
    "# input path of datasets\n",
    "PATH = input()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "monthly-attribute",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Dataset has successfully read\n"
     ]
    }
   ],
   "source": [
    "list_df = []\n",
    "dataset_ab_test  = 'AB Test Hit.csv'\n",
    "dataset_landing = 'Landing Viewed.csv'\n",
    "dataset_mainpage = 'Main Page Viewed.csv'\n",
    "dataset_address_change = 'Address Change Initiated.csv'\n",
    "dataset_delivery_zone = 'Address Not In Delivery Zone.csv'\n",
    "dataset_shop_started = 'Shop Celection Started.csv'\n",
    "dataset_shop_selected = 'Shop Selected.csv'\n",
    "dataset_product = 'Product Added.csv'\n",
    "dataset_cart = 'Add To Cart Clicked.csv'\n",
    "dataset_order = 'Order Completed.csv'\n",
    "\n",
    "try:\n",
    "    df_ab_test = pd.read_csv(f'{PATH}{dataset_ab_test}', sep=',')\n",
    "    df_landing = pd.read_csv(f'{PATH}{dataset_landing}', sep=',')\n",
    "    df_mainpage = pd.read_csv(f'{PATH}{dataset_mainpage}', sep=',')\n",
    "    df_address_change = pd.read_csv(f'{PATH}{dataset_address_change}', sep=',')\n",
    "    df_delivery_zone = pd.read_csv(f'{PATH}{dataset_delivery_zone}', sep=',')\n",
    "    df_shop_started = pd.read_csv(f'{PATH}{dataset_shop_started}', sep=',')\n",
    "    df_shop_selected = pd.read_csv(f'{PATH}{dataset_shop_selected}', sep=',')\n",
    "    df_product = pd.read_csv(f'{PATH}{dataset_product}', sep=',')\n",
    "    df_cart = pd.read_csv(f'{PATH}{dataset_cart}', sep=',')\n",
    "    df_order = pd.read_csv(f'{PATH}{dataset_order}', sep=',')\n",
    "    print('Dataset has successfully read')\n",
    "    list_df.extend([df_ab_test, df_landing, df_mainpage, df_address_change, df_delivery_zone,\n",
    "                    df_shop_started, df_shop_selected, df_product, df_cart, df_order])\n",
    "except Exception as err:\n",
    "    print(err)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "coupled-steam",
   "metadata": {},
   "outputs": [],
   "source": [
    "df_descriptions = {\n",
    "    'df_ab_test': 'AB Test Hit',\n",
    "    'df_landing': 'Landing Viewed',\n",
    "    'df_mainpage':'Main Page Viewed',\n",
    "    'df_address_change': 'Address Change Initiated',\n",
    "    'df_delivery_zone': 'Address Not In Delivery Zone',\n",
    "    'df_shop_started': 'Shop Celection Started',\n",
    "    'df_shop_selected': 'Shop Selected',\n",
    "    'df_product': 'Product Added',\n",
    "    'df_cart': 'Add To Cart Clicked',\n",
    "    'df_order': 'Order Completed'\n",
    "}\n",
    "\n",
    "# Get name of dataframe\n",
    "def get_df_name(df):\n",
    "    name =[x for x in globals() if globals()[x] is df][0]\n",
    "    return name"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "assured-lodge",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "df_ab_test\n",
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 502784 entries, 0 to 502783\n",
      "Data columns (total 6 columns):\n",
      " #   Column        Non-Null Count   Dtype \n",
      "---  ------        --------------   ----- \n",
      " 0   hit_at        502784 non-null  object\n",
      " 1   anonymous_id  502784 non-null  object\n",
      " 2   group         502784 non-null  object\n",
      " 3   device_type   502784 non-null  object\n",
      " 4   browser       502784 non-null  object\n",
      " 5   os            502783 non-null  object\n",
      "dtypes: object(6)\n",
      "memory usage: 23.0+ MB\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "None"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "df_landing\n",
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 341481 entries, 0 to 341480\n",
      "Data columns (total 2 columns):\n",
      " #   Column        Non-Null Count   Dtype \n",
      "---  ------        --------------   ----- \n",
      " 0   timestamp     341481 non-null  object\n",
      " 1   anonymous_id  341481 non-null  object\n",
      "dtypes: object(2)\n",
      "memory usage: 5.2+ MB\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "None"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "df_mainpage\n",
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 613509 entries, 0 to 613508\n",
      "Data columns (total 3 columns):\n",
      " #   Column        Non-Null Count   Dtype \n",
      "---  ------        --------------   ----- \n",
      " 0   timestamp     613509 non-null  object\n",
      " 1   anonymous_id  613509 non-null  object\n",
      " 2   retailer_id   613509 non-null  int64 \n",
      "dtypes: int64(1), object(2)\n",
      "memory usage: 14.0+ MB\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "None"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "df_address_change\n",
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 103434 entries, 0 to 103433\n",
      "Data columns (total 3 columns):\n",
      " #   Column        Non-Null Count   Dtype \n",
      "---  ------        --------------   ----- \n",
      " 0   timestamp     103434 non-null  object\n",
      " 1   anonymous_id  103434 non-null  object\n",
      " 2   source        103434 non-null  object\n",
      "dtypes: object(3)\n",
      "memory usage: 2.4+ MB\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "None"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "df_delivery_zone\n",
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 6167 entries, 0 to 6166\n",
      "Data columns (total 2 columns):\n",
      " #   Column        Non-Null Count  Dtype \n",
      "---  ------        --------------  ----- \n",
      " 0   timestamp     6167 non-null   object\n",
      " 1   anonymous_id  6167 non-null   object\n",
      "dtypes: object(2)\n",
      "memory usage: 96.5+ KB\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "None"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "df_shop_started\n",
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 57407 entries, 0 to 57406\n",
      "Data columns (total 3 columns):\n",
      " #   Column        Non-Null Count  Dtype \n",
      "---  ------        --------------  ----- \n",
      " 0   timestamp     57407 non-null  object\n",
      " 1   anonymous_id  57407 non-null  object\n",
      " 2   source        57407 non-null  object\n",
      "dtypes: object(3)\n",
      "memory usage: 1.3+ MB\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "None"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "df_shop_selected\n",
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 211728 entries, 0 to 211727\n",
      "Data columns (total 3 columns):\n",
      " #   Column        Non-Null Count   Dtype \n",
      "---  ------        --------------   ----- \n",
      " 0   timestamp     211728 non-null  object\n",
      " 1   anonymous_id  211728 non-null  object\n",
      " 2   source        211728 non-null  object\n",
      "dtypes: object(3)\n",
      "memory usage: 4.8+ MB\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "None"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "df_product\n",
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 1150060 entries, 0 to 1150059\n",
      "Data columns (total 2 columns):\n",
      " #   Column        Non-Null Count    Dtype \n",
      "---  ------        --------------    ----- \n",
      " 0   timestamp     1150060 non-null  object\n",
      " 1   anonymous_id  1150060 non-null  object\n",
      "dtypes: object(2)\n",
      "memory usage: 17.5+ MB\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "None"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "df_cart\n",
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 1223417 entries, 0 to 1223416\n",
      "Data columns (total 2 columns):\n",
      " #   Column        Non-Null Count    Dtype \n",
      "---  ------        --------------    ----- \n",
      " 0   timestamp     1223417 non-null  object\n",
      " 1   anonymous_id  1223417 non-null  object\n",
      "dtypes: object(2)\n",
      "memory usage: 18.7+ MB\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "None"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "df_order\n",
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 16650 entries, 0 to 16649\n",
      "Data columns (total 3 columns):\n",
      " #   Column        Non-Null Count  Dtype  \n",
      "---  ------        --------------  -----  \n",
      " 0   timestamp     16650 non-null  object \n",
      " 1   anonymous_id  16650 non-null  object \n",
      " 2   order_sum     16650 non-null  float64\n",
      "dtypes: float64(1), object(2)\n",
      "memory usage: 390.4+ KB\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "None"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "for _df in list_df:\n",
    "    print(get_df_name(_df))\n",
    "    display(_df.info())"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "empirical-february",
   "metadata": {},
   "source": [
    "Выявили 1 пропуск в типе ОС пользователя"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "turkish-smoke",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>hit_at</th>\n",
       "      <th>anonymous_id</th>\n",
       "      <th>group</th>\n",
       "      <th>device_type</th>\n",
       "      <th>browser</th>\n",
       "      <th>os</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>353476</th>\n",
       "      <td>2020-11-26 08:04:14.164 UTC</td>\n",
       "      <td>43c2bbe6-ba74-4b76-9597-421d3175ad30</td>\n",
       "      <td>default</td>\n",
       "      <td>desktop</td>\n",
       "      <td>Firefox</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                             hit_at                          anonymous_id  \\\n",
       "353476  2020-11-26 08:04:14.164 UTC  43c2bbe6-ba74-4b76-9597-421d3175ad30   \n",
       "\n",
       "          group device_type  browser   os  \n",
       "353476  default     desktop  Firefox  NaN  "
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_ab_test.loc[df_ab_test.isna().any(axis=1),:]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "figured-apparatus",
   "metadata": {},
   "source": [
    "<a id=\"preprocessing\"></a>\n",
    "## 2. Предобработка данных"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "lesbian-commercial",
   "metadata": {},
   "source": [
    "### Приведём дату к типу `datetime` во всех\n",
    "И обрежем время до секунды"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "prescription-screw",
   "metadata": {},
   "outputs": [],
   "source": [
    "df_ab_test['hit_at'] = df_ab_test['hit_at'].apply(lambda x: re.sub(r\"\\.[0-9]{1,3}\", \"\", x))\n",
    "df_ab_test['hit_at'] = pd.to_datetime(df_ab_test['hit_at'], format='%Y-%m-%d %H:%M:%S %Z')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "danish-queens",
   "metadata": {},
   "outputs": [],
   "source": [
    "for _df in list_df:\n",
    "    if _df.equals(df_ab_test):\n",
    "        continue\n",
    "    _df['timestamp'] = _df['timestamp'].apply(lambda x: re.sub(r\"\\.[0-9]{1,3}\", \"\", x))\n",
    "    _df['timestamp'] = pd.to_datetime(_df['timestamp'], format='%Y-%m-%d %H:%M:%S %Z')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "celtic-daisy",
   "metadata": {},
   "source": [
    "### Удалим все полные дубликаты во всех таблицах\n",
    "\n",
    "Неопытный аналитик мог совершить ошибку в сборе данных и задублировать их"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "bronze-finnish",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "124 rows duplicated has removed in df_ab_test table\n",
      "2324 rows duplicated has removed in df_landing table\n",
      "7932 rows duplicated has removed in df_mainpage table\n",
      "920 rows duplicated has removed in df_address_change table\n",
      "3 rows duplicated has removed in df_delivery_zone table\n",
      "562 rows duplicated has removed in df_shop_started table\n",
      "22348 rows duplicated has removed in df_shop_selected table\n",
      "4400 rows duplicated has removed in df_product table\n",
      "4400 rows duplicated has removed in df_cart table\n",
      "0 rows duplicated has removed in df_order table\n"
     ]
    }
   ],
   "source": [
    "for i, _df in enumerate(list_df):\n",
    "    df_name = get_df_name(_df)\n",
    "    n_duplicates = _df[_df.duplicated(keep='first')].shape[0]\n",
    "    list_df[i] = list_df[i].drop_duplicates()\n",
    "    print(f\"{n_duplicates} rows duplicated has removed in {df_name} table\")\n",
    "\n",
    "# Prevent copy of dataframes\n",
    "df_ab_test = list_df[0]\n",
    "df_landing = list_df[1]\n",
    "df_mainpage = list_df[2]\n",
    "df_adress_change = list_df[3]\n",
    "df_delivery_zone = list_df[4]\n",
    "df_shop_started = list_df[5]\n",
    "df_shop_selected = list_df[6]\n",
    "df_product = list_df[7]\n",
    "df_cart = list_df[8]\n",
    "df_order = list_df[9]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "alien-pearl",
   "metadata": {},
   "source": [
    "<a id=\"eda\"></a>\n",
    "## 3. Исследовательский анализ данных"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "close-survival",
   "metadata": {},
   "source": [
    "### Оставим только тех пользователей которые были на странице лендинга"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "floral-artist",
   "metadata": {},
   "source": [
    "Оставим только первый заход на лендинг для каждого пользователя"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "secret-farmer",
   "metadata": {},
   "outputs": [],
   "source": [
    "df_landing = df_landing.sort_values('timestamp').drop_duplicates('anonymous_id')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "express-wheel",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "282744(56.25%) rows of clients not viewed landing has removed\n"
     ]
    }
   ],
   "source": [
    "df = df_ab_test.copy().merge(df_landing, how='left', on='anonymous_id')\n",
    "#df = df.groupby(['anonymous_id'])['timestamp'].min().reset_index() # keep only first visit of landing\n",
    "\n",
    "df = df.rename(columns={\"timestamp\": \"timestamp_landing\"})\n",
    "df_nan = df['timestamp_landing'].isna()\n",
    "\n",
    "print(f\"{df_nan.sum()}({(df_nan.sum()/df.shape[0])*100:.2f}%) rows of clients not viewed landing has removed\")\n",
    "\n",
    "df = df[~df_nan]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "naval-pierre",
   "metadata": {},
   "source": [
    "Удалили **56%** записей по пользователями которые не бывали на лендинге. Это очень большая доля от всех пользователей."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "tough-glenn",
   "metadata": {},
   "source": [
    "### Если пользователь несколько раз попал в группу A/B тестирования то оставим только одну запись"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "current-relay",
   "metadata": {},
   "source": [
    "#### Оставим только первую по времени запись"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "entire-barrel",
   "metadata": {},
   "outputs": [],
   "source": [
    "first_unique_id = df[['hit_at', 'anonymous_id']].groupby('anonymous_id').min()\n",
    "df = first_unique_id.reset_index().merge(df, how='left', on=['anonymous_id', 'hit_at'])"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "different-gregory",
   "metadata": {},
   "source": [
    "#### Не будем учитывать пользователей которые попали в обе группы"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "middle-fraction",
   "metadata": {},
   "source": [
    "Проверим есть ли дублирование пользователей по группам"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "desperate-instrumentation",
   "metadata": {},
   "outputs": [],
   "source": [
    "group_doubles = df.duplicated(subset=['anonymous_id', 'hit_at', 'device_type', 'browser', 'os'], keep=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "skilled-backup",
   "metadata": {},
   "source": [
    "Удалим дубликаты"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "professional-dragon",
   "metadata": {},
   "outputs": [],
   "source": [
    "df = df[~group_doubles]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "respective-newark",
   "metadata": {},
   "source": [
    "Проверим, что у нас осталось только по 1 записи на уникального пользователя"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "related-administration",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df[df.duplicated(subset='anonymous_id', keep=False)].shape[0]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "technological-revision",
   "metadata": {},
   "source": [
    "### Временные характеристики эксперимента"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "affected-corner",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Начало эксперимента: 2020-11-26 00:00:10+00:00\n",
      "Конец эксперимента: 2020-12-10 23:59:57+00:00\n",
      "Продолжительность эксперимента: 14 days 23:59:47\n"
     ]
    }
   ],
   "source": [
    "print(f\"Начало эксперимента: {df_ab_test['hit_at'].min()}\")\n",
    "print(f\"Конец эксперимента: {df_ab_test['hit_at'].max()}\")\n",
    "print(f\"Продолжительность эксперимента: {(df_ab_test['hit_at'].max() - df_ab_test['hit_at'].min())}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "fifth-claim",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import warnings\n",
    "warnings.filterwarnings('ignore')\n",
    "\n",
    "fig = plt.figure(figsize=(10,6))\n",
    "graph = df_ab_test.groupby('group')['hit_at'].apply(lambda x: sns.distplot(x, hist=False, rug=False, label='time'))\n",
    "graph.iloc[0].set_xticklabels(labels=df_ab_test['hit_at'].dt.strftime('%Y-%m-%d'), rotation=45)\n",
    "graph.iloc[0].set(xlabel='Time')\n",
    "\n",
    "fig.legend(labels=['default','address_first'])\n",
    "plt.title('Density of clients')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "catholic-cooperative",
   "metadata": {},
   "source": [
    "Можем сделать вывод, что число пользователей каждый день было примерно одинаково независимо от группы относительно общего числа пользователей группы. Скачки в числе показов лендинга с адресом требуют дополнительного исследования."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "exterior-transaction",
   "metadata": {},
   "source": [
    "### Cоотношение распределение пользователей в A/B тестировании"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "arabic-history",
   "metadata": {},
   "outputs": [],
   "source": [
    "def percentage_string(x):\n",
    "    '''Return percent as string'''\n",
    "    if x > 0.01 or math.isclose(x, 0):\n",
    "        return f\"{round(x,2)*100:.0f}%\"\n",
    "    else:\n",
    "        return \"<1%\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "smart-abuse",
   "metadata": {},
   "outputs": [],
   "source": [
    "COLORS = ['Red','Orange', 'Blue', 'Purple', 'Green', 'Magenta']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "pleasant-thirty",
   "metadata": {},
   "outputs": [],
   "source": [
    "def plot_bar(data, column='count', title=None):\n",
    "    plt.figure(figsize=(len(data.index)*1.5,5))\n",
    "    graph = plt.bar(data.index, data[column], color=COLORS)\n",
    "\n",
    "    plt.title(title)\n",
    "\n",
    "    for p, i in zip(graph, range(len(graph))):\n",
    "        width = p.get_width()\n",
    "        height = p.get_height()\n",
    "        x, y = p.get_xy()\n",
    "        plt.text(x+width/2,\n",
    "                 y+height*.5,\n",
    "                 data.percent[i],\n",
    "                 ha='center')\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "sealed-dryer",
   "metadata": {},
   "source": [
    "#### Посмотрим соотношение распределение пользователей по группам A/B тестирования"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "boring-argentina",
   "metadata": {},
   "outputs": [],
   "source": [
    "groups_count = df[['anonymous_id', 'group']].groupby('group').agg(['size'])\n",
    "groups_count.columns = ['count']\n",
    "groups_count['percent'] = groups_count.transform(lambda x: x/x.sum())\n",
    "groups_count['percent'] = groups_count['percent'].transform(percentage_string)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "nutritional-vaccine",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 216x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_bar(groups_count, title='Distribution of user groups')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "increased-composition",
   "metadata": {},
   "source": [
    "Наблюдаем, что распределение пользователей по группам которым показывался разный вариант лендинга сильно несбалансировано."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "occasional-extraction",
   "metadata": {},
   "source": [
    "#### Посмотрим с каким платформ заходили на лендинг"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "found-second",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th>anonymous_id</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>group</th>\n",
       "      <th>device_type</th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th rowspan=\"5\" valign=\"top\">address_first</th>\n",
       "      <th>console</th>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>desktop</th>\n",
       "      <td>11246</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mobile</th>\n",
       "      <td>11993</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>smarttv</th>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>tablet</th>\n",
       "      <td>358</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"5\" valign=\"top\">default</th>\n",
       "      <th>console</th>\n",
       "      <td>8</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>desktop</th>\n",
       "      <td>90428</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mobile</th>\n",
       "      <td>94111</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>smarttv</th>\n",
       "      <td>7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>tablet</th>\n",
       "      <td>2705</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                           anonymous_id\n",
       "group         device_type              \n",
       "address_first console                 1\n",
       "              desktop             11246\n",
       "              mobile              11993\n",
       "              smarttv                 1\n",
       "              tablet                358\n",
       "default       console                 8\n",
       "              desktop             90428\n",
       "              mobile              94111\n",
       "              smarttv                 7\n",
       "              tablet               2705"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.pivot_table(df, values='anonymous_id', index=['group', 'device_type'], aggfunc='count')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "active-alert",
   "metadata": {},
   "source": [
    "Кол-во пользователей с консолей и смартТВ незначительное. Отметим, что число пользователей с планшета невелико."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "dressed-compromise",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 540x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "device_count = df[['anonymous_id', 'device_type']].groupby('device_type').agg(['size'])\n",
    "device_count.columns = ['count']\n",
    "device_count['percent'] = device_count.transform(lambda x: x/x.sum())\n",
    "device_count['percent'] = device_count['percent'].transform(percentage_string)\n",
    "device_count = device_count.sort_values('count')\n",
    "\n",
    "plot_bar(device_count, title='Distribution of user by devices')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "proof-consideration",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1512x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "os_count = df[['anonymous_id', 'os']].groupby('os').agg(['size'])\n",
    "os_count.columns = ['count']\n",
    "os_count['percent'] = os_count.transform(lambda x: x/x.sum())\n",
    "os_count['percent'] = os_count['percent'].transform(percentage_string)\n",
    "os_count = os_count.sort_values('count')\n",
    "\n",
    "plot_bar(os_count, title='Distribution of users by OS')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "living-cookie",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>count</th>\n",
       "      <th>percent</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>browser</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Chrome</th>\n",
       "      <td>250166</td>\n",
       "      <td>50%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Mobile Safari</th>\n",
       "      <td>83765</td>\n",
       "      <td>17%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Yandex</th>\n",
       "      <td>53589</td>\n",
       "      <td>11%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Chrome WebView</th>\n",
       "      <td>18897</td>\n",
       "      <td>4%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>WebKit</th>\n",
       "      <td>18045</td>\n",
       "      <td>4%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Samsung Browser</th>\n",
       "      <td>16020</td>\n",
       "      <td>3%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Firefox</th>\n",
       "      <td>15088</td>\n",
       "      <td>3%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Opera</th>\n",
       "      <td>12907</td>\n",
       "      <td>3%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>MIUI Browser</th>\n",
       "      <td>12526</td>\n",
       "      <td>2%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Safari</th>\n",
       "      <td>8853</td>\n",
       "      <td>2%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Edge</th>\n",
       "      <td>6078</td>\n",
       "      <td>1%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>GSA</th>\n",
       "      <td>3594</td>\n",
       "      <td>&lt;1%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Facebook</th>\n",
       "      <td>1691</td>\n",
       "      <td>&lt;1%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Android Browser</th>\n",
       "      <td>556</td>\n",
       "      <td>&lt;1%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>UCBrowser</th>\n",
       "      <td>429</td>\n",
       "      <td>&lt;1%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Opera Touch</th>\n",
       "      <td>124</td>\n",
       "      <td>&lt;1%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>IE</th>\n",
       "      <td>95</td>\n",
       "      <td>&lt;1%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Maxthon</th>\n",
       "      <td>60</td>\n",
       "      <td>&lt;1%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Puffin</th>\n",
       "      <td>39</td>\n",
       "      <td>&lt;1%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Chromium</th>\n",
       "      <td>39</td>\n",
       "      <td>&lt;1%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Vivaldi</th>\n",
       "      <td>17</td>\n",
       "      <td>&lt;1%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Waterfox</th>\n",
       "      <td>15</td>\n",
       "      <td>&lt;1%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>PaleMoon</th>\n",
       "      <td>14</td>\n",
       "      <td>&lt;1%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Mozilla</th>\n",
       "      <td>13</td>\n",
       "      <td>&lt;1%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>QQBrowser</th>\n",
       "      <td>9</td>\n",
       "      <td>&lt;1%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>jasmine</th>\n",
       "      <td>6</td>\n",
       "      <td>&lt;1%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>WeChat</th>\n",
       "      <td>4</td>\n",
       "      <td>&lt;1%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>IEMobile</th>\n",
       "      <td>3</td>\n",
       "      <td>&lt;1%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Slim</th>\n",
       "      <td>3</td>\n",
       "      <td>&lt;1%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Avast Secure Browser</th>\n",
       "      <td>3</td>\n",
       "      <td>&lt;1%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Brave</th>\n",
       "      <td>2</td>\n",
       "      <td>&lt;1%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>MetaSr</th>\n",
       "      <td>2</td>\n",
       "      <td>&lt;1%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Opera Coast</th>\n",
       "      <td>2</td>\n",
       "      <td>&lt;1%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>SeaMonkey</th>\n",
       "      <td>2</td>\n",
       "      <td>&lt;1%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Silk</th>\n",
       "      <td>2</td>\n",
       "      <td>&lt;1%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>IceDragon</th>\n",
       "      <td>1</td>\n",
       "      <td>&lt;1%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Iceweasel</th>\n",
       "      <td>1</td>\n",
       "      <td>&lt;1%</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                       count percent\n",
       "browser                             \n",
       "Chrome                250166     50%\n",
       "Mobile Safari          83765     17%\n",
       "Yandex                 53589     11%\n",
       "Chrome WebView         18897      4%\n",
       "WebKit                 18045      4%\n",
       "Samsung Browser        16020      3%\n",
       "Firefox                15088      3%\n",
       "Opera                  12907      3%\n",
       "MIUI Browser           12526      2%\n",
       "Safari                  8853      2%\n",
       "Edge                    6078      1%\n",
       "GSA                     3594     <1%\n",
       "Facebook                1691     <1%\n",
       "Android Browser          556     <1%\n",
       "UCBrowser                429     <1%\n",
       "Opera Touch              124     <1%\n",
       "IE                        95     <1%\n",
       "Maxthon                   60     <1%\n",
       "Puffin                    39     <1%\n",
       "Chromium                  39     <1%\n",
       "Vivaldi                   17     <1%\n",
       "Waterfox                  15     <1%\n",
       "PaleMoon                  14     <1%\n",
       "Mozilla                   13     <1%\n",
       "QQBrowser                  9     <1%\n",
       "jasmine                    6     <1%\n",
       "WeChat                     4     <1%\n",
       "IEMobile                   3     <1%\n",
       "Slim                       3     <1%\n",
       "Avast Secure Browser       3     <1%\n",
       "Brave                      2     <1%\n",
       "MetaSr                     2     <1%\n",
       "Opera Coast                2     <1%\n",
       "SeaMonkey                  2     <1%\n",
       "Silk                       2     <1%\n",
       "IceDragon                  1     <1%\n",
       "Iceweasel                  1     <1%"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "browser_count = df_ab_test[['anonymous_id', 'browser']].groupby('browser').agg(['size'])\n",
    "browser_count.columns = ['count']\n",
    "browser_count['percent'] = browser_count.transform(lambda x: x/x.sum())\n",
    "browser_count['percent'] = browser_count['percent'].transform(percentage_string)\n",
    "browser_count = browser_count.sort_values('count')\n",
    "\n",
    "browser_count.sort_values('count', ascending=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "stone-stone",
   "metadata": {},
   "source": [
    "Половина всех пользователей среди большого количества браузеров используют Google Chrome"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "agricultural-advisory",
   "metadata": {},
   "source": [
    "<a id=\"metric\"></a>\n",
    "## 4. Расчёт метрик (bounce rate лендинга и конверсию)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "discrete-barrier",
   "metadata": {},
   "source": [
    "### Рассчитаем конверсию добавления товара в корзину"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "about-contrast",
   "metadata": {},
   "outputs": [],
   "source": [
    "df_product = df_product.sort_values('timestamp').drop_duplicates('anonymous_id')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "damaged-heating",
   "metadata": {},
   "outputs": [],
   "source": [
    "df = df.merge(df_product, how='left', on='anonymous_id')\n",
    "df = df.rename(columns={\"timestamp\": \"timestamp_added_cart\"})"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "expected-ferry",
   "metadata": {},
   "source": [
    "#### Рассчитаем сколько времени потребовалось между просмотром лендинга и добавлением товара в корзину"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "polar-stephen",
   "metadata": {},
   "outputs": [],
   "source": [
    "df['time_to_cart_add'] = df['timestamp_added_cart'] - df['timestamp_landing']"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "turned-receptor",
   "metadata": {},
   "source": [
    "Удалим записи в которых добавление в корзину позже просмотра лендинга"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "delayed-kernel",
   "metadata": {},
   "outputs": [],
   "source": [
    "df = df[~(df['time_to_cart_add'] < pd.Timedelta(0))]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "frozen-addiction",
   "metadata": {},
   "source": [
    "#### Проверим конверсию по группам A/B тестирования"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "charged-badge",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>count_cart</th>\n",
       "      <th>count_landing</th>\n",
       "      <th>conversion</th>\n",
       "      <th>conversion_%</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>group</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>address_first</th>\n",
       "      <td>3516</td>\n",
       "      <td>23234</td>\n",
       "      <td>0.151330</td>\n",
       "      <td>15%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>default</th>\n",
       "      <td>23737</td>\n",
       "      <td>184510</td>\n",
       "      <td>0.128649</td>\n",
       "      <td>13%</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "               count_cart  count_landing  conversion conversion_%\n",
       "group                                                            \n",
       "address_first        3516          23234    0.151330          15%\n",
       "default             23737         184510    0.128649          13%"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "t1 = pd.pivot_table(df, values='timestamp_added_cart', index=['group'], aggfunc='count')\n",
    "t1.columns=['count_cart']\n",
    "t2 = pd.pivot_table(df, values='anonymous_id', index=['group'], aggfunc='count')\n",
    "t2.columns=['count_landing']\n",
    "df_conversion = pd.concat([t1, t2], axis=1)\n",
    "df_conversion['conversion'] = df_conversion['count_cart'].values / df_conversion['count_landing'].values\n",
    "df_conversion['conversion_%'] = df_conversion['conversion'].transform(percentage_string)\n",
    "df_conversion"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "attractive-eugene",
   "metadata": {},
   "source": [
    "**По всем пользователям конверсия с новым лендингом выше на 2% (15% и 13%)**\n",
    "\n",
    "При значении статистической значимости в 0.05 с высокой вероятностью конверсия нового лендинга выше, чем старого (p < 0.001)\n",
    "\n",
    "<img src=\"pics/conversion_significance.png\">"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "shared-seating",
   "metadata": {},
   "source": [
    "#### Проверим конверсию по типам устройств используемых пользователями"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "gorgeous-benefit",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th>count_cart</th>\n",
       "      <th>count_landing</th>\n",
       "      <th>conversion</th>\n",
       "      <th>conversion_%</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>group</th>\n",
       "      <th>device_type</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th rowspan=\"5\" valign=\"top\">address_first</th>\n",
       "      <th>console</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>desktop</th>\n",
       "      <td>2345</td>\n",
       "      <td>10936</td>\n",
       "      <td>0.214429</td>\n",
       "      <td>21%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mobile</th>\n",
       "      <td>1135</td>\n",
       "      <td>11946</td>\n",
       "      <td>0.095011</td>\n",
       "      <td>10%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>smarttv</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>tablet</th>\n",
       "      <td>36</td>\n",
       "      <td>350</td>\n",
       "      <td>0.102857</td>\n",
       "      <td>10%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"5\" valign=\"top\">default</th>\n",
       "      <th>console</th>\n",
       "      <td>0</td>\n",
       "      <td>8</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>desktop</th>\n",
       "      <td>17593</td>\n",
       "      <td>88133</td>\n",
       "      <td>0.199619</td>\n",
       "      <td>20%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mobile</th>\n",
       "      <td>5903</td>\n",
       "      <td>93694</td>\n",
       "      <td>0.063003</td>\n",
       "      <td>6%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>smarttv</th>\n",
       "      <td>2</td>\n",
       "      <td>7</td>\n",
       "      <td>0.285714</td>\n",
       "      <td>29%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>tablet</th>\n",
       "      <td>239</td>\n",
       "      <td>2668</td>\n",
       "      <td>0.089580</td>\n",
       "      <td>9%</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                           count_cart  count_landing  conversion conversion_%\n",
       "group         device_type                                                    \n",
       "address_first console               0              1    0.000000           0%\n",
       "              desktop            2345          10936    0.214429          21%\n",
       "              mobile             1135          11946    0.095011          10%\n",
       "              smarttv               0              1    0.000000           0%\n",
       "              tablet               36            350    0.102857          10%\n",
       "default       console               0              8    0.000000           0%\n",
       "              desktop           17593          88133    0.199619          20%\n",
       "              mobile             5903          93694    0.063003           6%\n",
       "              smarttv               2              7    0.285714          29%\n",
       "              tablet              239           2668    0.089580           9%"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "t1 = pd.pivot_table(df, values='timestamp_added_cart', index=['group', 'device_type'], aggfunc='count')\n",
    "t1.columns=['count_cart']\n",
    "t2 = pd.pivot_table(df, values='anonymous_id', index=['group', 'device_type'], aggfunc='count')\n",
    "t2.columns=['count_landing']\n",
    "df_conversion = pd.concat([t1, t2], axis=1)\n",
    "df_conversion['conversion'] = df_conversion['count_cart'].values / df_conversion['count_landing'].values\n",
    "df_conversion['conversion_%'] = df_conversion['conversion'].transform(percentage_string)\n",
    "df_conversion"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "negative-advantage",
   "metadata": {},
   "source": [
    "**Конверсия между группами пользователей среди пользователей с мобильных устройств отличается на 4% или в ~1.5 раза (10% и 6%)**\n",
    "\n",
    "\n",
    "<img src=\"pics/conversion_mobile_significance.png\">"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "worthy-batman",
   "metadata": {},
   "source": [
    "**Конверсия между группами пользователей среди пользователей с ПК 21.4% и 20%**\n",
    "\n",
    "\n",
    "<img src=\"pics/conversion_desktop_significance.png\">"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "independent-carnival",
   "metadata": {},
   "source": [
    "По каким-то причинам ни один пользователь с консоли не добавил товар в корзину. И в группе пользователей `address_first` ни один пользователь не добавил товар в корзину, поэтому всех пользователей которые зашли на лендинг с консоли и смартТВ.\n",
    "\n",
    "Необходимо провести дополнительное расследование нет ли ошибки в лендинге которая не позволяет дойти до добавления товара в корзину при использовании смартТВ или консоли. Нужно учитывать незначительное число пользователей с этих типов устройств."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "molecular-group",
   "metadata": {},
   "source": [
    "#### Проверим конверсию по типам ОС"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "accepted-eligibility",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th>count_cart</th>\n",
       "      <th>count_landing</th>\n",
       "      <th>conversion</th>\n",
       "      <th>conversion_%</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>group</th>\n",
       "      <th>os</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th rowspan=\"12\" valign=\"top\">address_first</th>\n",
       "      <th>Android</th>\n",
       "      <td>779</td>\n",
       "      <td>8050</td>\n",
       "      <td>0.096770</td>\n",
       "      <td>10%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Chromium OS</th>\n",
       "      <td>1</td>\n",
       "      <td>4</td>\n",
       "      <td>0.250000</td>\n",
       "      <td>25%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Fedora</th>\n",
       "      <td>0</td>\n",
       "      <td>4</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Linux</th>\n",
       "      <td>20</td>\n",
       "      <td>3637</td>\n",
       "      <td>0.005499</td>\n",
       "      <td>&lt;1%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Mac OS</th>\n",
       "      <td>229</td>\n",
       "      <td>711</td>\n",
       "      <td>0.322082</td>\n",
       "      <td>32%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>PlayStation</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Tizen</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Ubuntu</th>\n",
       "      <td>5</td>\n",
       "      <td>32</td>\n",
       "      <td>0.156250</td>\n",
       "      <td>16%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Windows</th>\n",
       "      <td>2081</td>\n",
       "      <td>6497</td>\n",
       "      <td>0.320302</td>\n",
       "      <td>32%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Windows Phone</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Windows Phone OS</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>iOS</th>\n",
       "      <td>401</td>\n",
       "      <td>4295</td>\n",
       "      <td>0.093364</td>\n",
       "      <td>9%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"13\" valign=\"top\">default</th>\n",
       "      <th>Android</th>\n",
       "      <td>4053</td>\n",
       "      <td>63555</td>\n",
       "      <td>0.063772</td>\n",
       "      <td>6%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Chromium OS</th>\n",
       "      <td>2</td>\n",
       "      <td>20</td>\n",
       "      <td>0.100000</td>\n",
       "      <td>10%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Fedora</th>\n",
       "      <td>3</td>\n",
       "      <td>8</td>\n",
       "      <td>0.375000</td>\n",
       "      <td>38%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>FreeBSD</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Linux</th>\n",
       "      <td>217</td>\n",
       "      <td>29502</td>\n",
       "      <td>0.007355</td>\n",
       "      <td>&lt;1%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Mac OS</th>\n",
       "      <td>1970</td>\n",
       "      <td>5806</td>\n",
       "      <td>0.339304</td>\n",
       "      <td>34%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>NetBSD</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>PlayStation</th>\n",
       "      <td>0</td>\n",
       "      <td>8</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Tizen</th>\n",
       "      <td>2</td>\n",
       "      <td>6</td>\n",
       "      <td>0.333333</td>\n",
       "      <td>33%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Ubuntu</th>\n",
       "      <td>48</td>\n",
       "      <td>216</td>\n",
       "      <td>0.222222</td>\n",
       "      <td>22%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Windows</th>\n",
       "      <td>15311</td>\n",
       "      <td>52168</td>\n",
       "      <td>0.293494</td>\n",
       "      <td>29%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Windows Phone</th>\n",
       "      <td>0</td>\n",
       "      <td>12</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>iOS</th>\n",
       "      <td>2131</td>\n",
       "      <td>33206</td>\n",
       "      <td>0.064175</td>\n",
       "      <td>6%</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                count_cart  count_landing  conversion  \\\n",
       "group         os                                                        \n",
       "address_first Android                  779           8050    0.096770   \n",
       "              Chromium OS                1              4    0.250000   \n",
       "              Fedora                     0              4    0.000000   \n",
       "              Linux                     20           3637    0.005499   \n",
       "              Mac OS                   229            711    0.322082   \n",
       "              PlayStation                0              1    0.000000   \n",
       "              Tizen                      0              1    0.000000   \n",
       "              Ubuntu                     5             32    0.156250   \n",
       "              Windows                 2081           6497    0.320302   \n",
       "              Windows Phone              0              1    0.000000   \n",
       "              Windows Phone OS           0              1    0.000000   \n",
       "              iOS                      401           4295    0.093364   \n",
       "default       Android                 4053          63555    0.063772   \n",
       "              Chromium OS                2             20    0.100000   \n",
       "              Fedora                     3              8    0.375000   \n",
       "              FreeBSD                    0              1    0.000000   \n",
       "              Linux                    217          29502    0.007355   \n",
       "              Mac OS                  1970           5806    0.339304   \n",
       "              NetBSD                     0              1    0.000000   \n",
       "              PlayStation                0              8    0.000000   \n",
       "              Tizen                      2              6    0.333333   \n",
       "              Ubuntu                    48            216    0.222222   \n",
       "              Windows                15311          52168    0.293494   \n",
       "              Windows Phone              0             12    0.000000   \n",
       "              iOS                     2131          33206    0.064175   \n",
       "\n",
       "                               conversion_%  \n",
       "group         os                             \n",
       "address_first Android                   10%  \n",
       "              Chromium OS               25%  \n",
       "              Fedora                     0%  \n",
       "              Linux                     <1%  \n",
       "              Mac OS                    32%  \n",
       "              PlayStation                0%  \n",
       "              Tizen                      0%  \n",
       "              Ubuntu                    16%  \n",
       "              Windows                   32%  \n",
       "              Windows Phone              0%  \n",
       "              Windows Phone OS           0%  \n",
       "              iOS                        9%  \n",
       "default       Android                    6%  \n",
       "              Chromium OS               10%  \n",
       "              Fedora                    38%  \n",
       "              FreeBSD                    0%  \n",
       "              Linux                     <1%  \n",
       "              Mac OS                    34%  \n",
       "              NetBSD                     0%  \n",
       "              PlayStation                0%  \n",
       "              Tizen                     33%  \n",
       "              Ubuntu                    22%  \n",
       "              Windows                   29%  \n",
       "              Windows Phone              0%  \n",
       "              iOS                        6%  "
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "t1 = pd.pivot_table(df, values='timestamp_added_cart', index=['group', 'os'], aggfunc='count')\n",
    "t1.columns=['count_cart']\n",
    "t2 = pd.pivot_table(df, values='anonymous_id', index=['group', 'os'], aggfunc='count')\n",
    "t2.columns=['count_landing']\n",
    "df_conversion = pd.concat([t1, t2], axis=1)\n",
    "df_conversion['conversion'] = df_conversion['count_cart'].values / df_conversion['count_landing'].values\n",
    "df_conversion['conversion_%'] = df_conversion['conversion'].transform(percentage_string)\n",
    "df_conversion"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "hollywood-childhood",
   "metadata": {},
   "source": [
    "### Рассчитаем bounce rate лендинга"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "european-hebrew",
   "metadata": {},
   "source": [
    "Для расчёта bounce rate лендинга необходимо рассчитать перешёл ли пользователь с лендинга на следующую страницу (для каждой группы это разная страница)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "complimentary-salad",
   "metadata": {},
   "outputs": [],
   "source": [
    "df_mainpage = df_mainpage.sort_values('timestamp').drop_duplicates('anonymous_id')\n",
    "df_adress_change = df_adress_change.sort_values('timestamp').drop_duplicates('anonymous_id')\n",
    "df_delivery_zone = df_delivery_zone.sort_values('timestamp').drop_duplicates('anonymous_id')\n",
    "df_shop_started = df_shop_started.sort_values('timestamp').drop_duplicates('anonymous_id')\n",
    "df_shop_selected = df_shop_selected.sort_values('timestamp').drop_duplicates('anonymous_id')\n",
    "df_cart = df_cart.sort_values('timestamp').drop_duplicates('anonymous_id')\n",
    "df_order = df_order.sort_values('timestamp').drop_duplicates('anonymous_id')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "atlantic-march",
   "metadata": {},
   "outputs": [],
   "source": [
    "df = df.merge(df_mainpage, how='left', on='anonymous_id')\n",
    "df = df.rename(columns={\"timestamp\": \"timestamp_mainpage\"})\n",
    "\n",
    "df = df.merge(df_adress_change, how='left', on='anonymous_id')\n",
    "df = df.rename(columns={\"timestamp\": \"timestamp_address\"})\n",
    "\n",
    "df = df.merge(df_delivery_zone, how='left', on='anonymous_id')\n",
    "df = df.rename(columns={\"timestamp\": \"timestamp_delivery\"})\n",
    "\n",
    "df = df.merge(df_shop_started, how='left', on='anonymous_id')\n",
    "df = df.rename(columns={\"timestamp\": \"timestamp_shop_started\"})\n",
    "\n",
    "df = df.merge(df_shop_selected, how='left', on='anonymous_id')\n",
    "df = df.rename(columns={\"timestamp\": \"timestamp_shop_selected\"})\n",
    "\n",
    "df = df.merge(df_cart, how='left', on='anonymous_id')\n",
    "df = df.rename(columns={\"timestamp\": \"timestamp_cart\"})\n",
    "\n",
    "df = df.merge(df_order, how='left', on='anonymous_id')\n",
    "df = df.rename(columns={\"timestamp\": \"timestamp_order\"})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "id": "french-labor",
   "metadata": {},
   "outputs": [],
   "source": [
    "df_default = df[df['group']=='default']\n",
    "df_address = df[df['group']=='address_first']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "id": "floating-poultry",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>count_maipage</th>\n",
       "      <th>count_landing</th>\n",
       "      <th>rate</th>\n",
       "      <th>rate_%</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>group</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>default</th>\n",
       "      <td>84127</td>\n",
       "      <td>184510</td>\n",
       "      <td>0.455948</td>\n",
       "      <td>46%</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "         count_maipage  count_landing      rate rate_%\n",
       "group                                                 \n",
       "default          84127         184510  0.455948    46%"
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "t1 = pd.pivot_table(df_default, values='timestamp_mainpage', index=['group'], aggfunc='count')\n",
    "t1.columns=['count_maipage']\n",
    "t2 = pd.pivot_table(df_default, values='anonymous_id', index=['group'], aggfunc='count')\n",
    "t2.columns=['count_landing']\n",
    "df_bounce = pd.concat([t1, t2], axis=1)\n",
    "df_bounce['rate'] = df_bounce['count_maipage'].values / df_bounce['count_landing'].values\n",
    "df_bounce['rate_%'] = df_bounce['rate'].transform(percentage_string)\n",
    "df_bounce"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "id": "plastic-branch",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>count_address</th>\n",
       "      <th>count_landing</th>\n",
       "      <th>rate</th>\n",
       "      <th>rate_%</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>group</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>address_first</th>\n",
       "      <td>9165</td>\n",
       "      <td>23234</td>\n",
       "      <td>0.394465</td>\n",
       "      <td>39%</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "               count_address  count_landing      rate rate_%\n",
       "group                                                       \n",
       "address_first           9165          23234  0.394465    39%"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "t1 = pd.pivot_table(df_address, values='timestamp_address', index=['group'], aggfunc='count')\n",
    "t1.columns=['count_address']\n",
    "t2 = pd.pivot_table(df_address, values='anonymous_id', index=['group'], aggfunc='count')\n",
    "t2.columns=['count_landing']\n",
    "df_bounce = pd.concat([t1, t2], axis=1)\n",
    "df_bounce['rate'] = df_bounce['count_address'].values / df_bounce['count_landing'].values\n",
    "df_bounce['rate_%'] = df_bounce['rate'].transform(percentage_string)\n",
    "df_bounce"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "cheap-dating",
   "metadata": {},
   "source": [
    "**Bounce rate дефолтного лендинга 46%, а лендинга с вводом адреса в начале 39%**"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "increasing-costume",
   "metadata": {},
   "source": [
    "При значении статистической значимости в 0.05 с высокой вероятностью bounce rate нового лендинга ниже, чем старый (p < 0.001)\n",
    "<img src=\"pics/bounce_rate_significance.png\">"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ordered-pizza",
   "metadata": {},
   "source": [
    "<a id=\"experiment\"></a>\n",
    "## 5. Диагностика дизайна эксперимента"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "boxed-northeast",
   "metadata": {},
   "source": [
    "В дизайне эксперимента не учтено, что пользователь зарегистрирован, уже ввёл адрес и не заходил на страницу лендинга.\n",
    "Пользователей с уже введённым адресом необходимо не учитывать в эксперименте."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "intellectual-architect",
   "metadata": {},
   "source": [
    "<a id=\"conclusion\"></a>\n",
    "## 6. Выводы"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "known-tobacco",
   "metadata": {},
   "source": [
    "- выявлена аномалия в различии в числе заходов пользователей на лендинг в течении дня\n",
    "- **56%** всех пользователей не бывали на странице лендинга\n",
    "- группы A/B тестирования несбалансированы (89% `default` и 11% `address_first`)\n",
    "- более **98%** пользователей использовали ПК или смартфон\n",
    "- с высокой вероятностью конверсия нового лендинга выше, чем старого (**15.13%** и **12.86%** соответственно)\n",
    "- с высокой вероятностью bounce rate нового лендинга ниже, чем старого (**39.44%** и **45.59%** соответственно)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}