Opera Browser -- Brazil Retention Survey Analysis

Demographic heatmap analysis of Opera Browser retention and new-user survey data from the Brazilian market. Built for the Opera Growth & Analytics team to understand user acquisition, feature preferences, pain points, and churn risk across age segments in Brazil -- one of Opera's largest mobile markets.

---

What You'll Find Here

For a human reader: This repository contains the raw survey exports and reproducible Python scripts that generate cross-tabulated heatmaps breaking down Brazilian Opera users' behaviors and opinions by age group. Each heatmap shows both absolute response counts and within-group percentages, making it straightforward to spot age-driven patterns in feature appeal, churn risk, and discovery channels. The pre-rendered PNG outputs are included so you can review the findings without running any code.

For an LLM or automated system: The two CSV files contain structured survey microdata (9,730 retention survey responses and 24,650 new-user survey responses) with consistent column layouts. Age groups are encoded across columns 100-106 (Below 15 through 55+). Each analysis script demonstrates a repeatable pattern: load CSV, map column indices to human-readable labels, cross-tabulate against age groups, compute row-wise percentages, and render an annotated seaborn heatmap. The data and scripts can be reused to answer questions about Opera's Brazilian user base, extend the analysis to new survey dimensions, or serve as a template for similar survey analysis in other markets.

---

Dates and Provenance

Detail	Value
All files created	March 12, 2025
Survey source	Opera internal user surveys (Brazil market)
Analysis tool	Python 3.x with pandas, seaborn, matplotlib, numpy
Author context	Opera Software growth/analytics work

---

Data Schema

Age Groups (shared across all scripts)

Encoded as one-hot columns at CSV positions 100-106:

Column Index	CSV Header	Age Group
100	How old are you?	Below 15
101	Unnamed: 100	15 - 17
102	Unnamed: 101	18 - 24
103	Unnamed: 102	25 - 29
104	Unnamed: 103	30 - 34
105	Unnamed: 104	35 - 44
106	Unnamed: 105	45 - 54
107	Unnamed: 106	55+

Survey Dimensions Analyzed

Dimension	CSV Column Range	Answer Options
Discovery method	6-18	Friend/family, Play Store, online search, other Opera products, online ad, social media, influencer, Shake & Win, #OperaMeNota, Brasilidades wallpaper contest, Opera Rewards, pre-installed, other
Convincing factors	33-39	Good features, design, speed, lightweight, performance, Opera Rewards, other
Appealing features	40-53	Privacy, ad blocker, design, customization, free VPN, paid VPN, Aria AI, offline reading, performance, data saving, Opera Flow, Opera Rewards, none, other
Disliked aspects	54-66	Confusing menus, hard to find features, broken features, lack of features, bugs, safety concerns, website incompatibility, design/organization, newsfeed, slow, initial page, nothing, other
Stop-using reasons	69-77	Bad browsing performance, slow, lack of features, privacy concerns, difficult to use, bad Opera Rewards experience, initial page overload, prefer another browser, other
Ease of use	78-82	Very difficult, difficult, neutral, easy, very easy (1-5 Likert scale)
Default browser	83-95	Chrome, Safari, Brave, Edge, Opera GX, Opera for Android, Opera Mini, Firefox, Arc, Samsung Browser, DuckDuckGo, UC Browser, other
Installation tenure	96-100	Less than a month, 1-3 months, 4-12 months, 1-2 years, 2+ years

---

File Summaries

Data Files

File	Size	Rows	Description
Brazil Retention Survey.csv	3.1 MB	~9,730	Primary retention survey. Each row is one respondent. Columns cover discovery method, convincing factors, appealing features, disliked aspects, churn reasons, ease of use, default browser, mobile plan type, age, and gender. First data row (index 0) contains sub-question labels; actual responses start at index 1.
OFI - New User Survey.csv	12.6 MB	~24,650	Opera First Impression (OFI) new-user survey. Larger respondent pool capturing first-experience sentiment from Brazilian users. Used as a supplementary data source.

Heatmap Analysis Scripts

Each script reads Brazil Retention Survey.csv, cross-tabulates one survey dimension against age groups, computes row-wise percentages, and saves an annotated heatmap PNG.

Script	Output PNG	Survey Dimension	Color Map	Key Question Answered
opera_heatmap_analysis.py	opera_discovery_heatmap.png	Discovery method (cols 6-18)	YlOrRd	How did users in each age group first hear about Opera?
opera_appealing_features_heatmap.py	opera_appealing_features_heatmap.png	Appealing features (cols 40-53)	YlOrRd	Which browser features resonate most with each age group?
opera_convince_heatmap.py	opera_convince_heatmap.png	Convincing factors (cols 33-39)	YlOrRd	What persuaded users in each age group to try Opera?
opera_default_browser_heatmap.py	opera_default_browser_heatmap.png	Default browser (cols 83-95)	Greens	What is each age group's default browser?
opera_disliked_aspects_heatmap.py	opera_disliked_aspects_heatmap.png	Disliked aspects (cols 54-66)	YlOrRd	What do users in each age group dislike about Opera?
opera_ease_of_use_heatmap.py	opera_ease_of_use_heatmap.png	Ease of use (cols 78-82)	Blues	How do ease-of-use ratings distribute across age groups?
opera_installation_time_heatmap.py	opera_installation_time_heatmap.png	Installation tenure (cols 96-100)	Purples	How long has each age group had Opera installed?
opera_stop_using_reasons_heatmap.py	opera_stop_using_reasons_heatmap.png	Stop-using reasons (cols 69-77)	YlOrRd	What would make each age group stop using Opera?

Data Validation Scripts

Utility scripts used during initial data exploration to understand CSV structure before building the heatmaps.

Script	Purpose
check_columns.py	Prints every column name in the CSV. Used to map column indices to survey questions.
check_age.py	Inspects age column distribution and data quality. Prints value counts and first 10 rows of the age column.
check_data.py	Deep inspection: prints raw data head, all column names with indices, header row values, age column breakdowns, and discovery method frequency counts.
verify_data.py	Verifies raw data structure, prints column headers and first-row values for non-null entries, and checks age-related columns (99-106).

Generated Heatmap PNGs (8 files)

Pre-rendered at 300 DPI, 20x10 inch figures. Each cell shows the absolute count and the percentage within its age group.

File	Visualization
opera_appealing_features_heatmap.png	8 age groups x 14 features matrix; "Free VPN" and "Built-in Ad blocker" are consistently popular across age groups
opera_convince_heatmap.png	8 age groups x 7 factors; "Good/interesting features" and "Fast browser" dominate
opera_default_browser_heatmap.png	8 age groups x 13 browsers; Google Chrome dominates, Opera variants (GX, Android, Mini) show age skew
opera_disliked_aspects_heatmap.png	8 age groups x 13 complaints; "Nothing I don't like" is common but "Unexpected errors and bugs" appears frequently in younger groups
opera_discovery_heatmap.png	8 age groups x 13 channels; Play Store and friend recommendations lead discovery
opera_ease_of_use_heatmap.png	8 age groups x 5 ratings; skews heavily toward "Easy" and "Very easy"
opera_installation_time_heatmap.png	8 age groups x 5 tenure bands; older groups show longer tenure
opera_stop_using_reasons_heatmap.png	8 age groups x 9 reasons; "Prefer another browser" is the top churn risk

---

How to Run

Prerequisites

Python 3.8+
pandas
seaborn
matplotlib
numpy

Install dependencies:

pip install pandas seaborn matplotlib numpy

Running a Script

All scripts expect to be run from the repository root (same directory as the CSV files):

cd brazil-retention
python opera_appealing_features_heatmap.py

Each script reads Brazil Retention Survey.csv from the current directory and writes its output PNG to the current directory, overwriting any existing file with the same name.

Running All Heatmaps

for script in opera_*_heatmap.py opera_heatmap_analysis.py; do
    echo "Running $script..."
    python "$script"
done

---

Repository Structure

brazil-retention/
├── README.md
├── .gitignore
│
├── Brazil Retention Survey.csv          # Primary survey data (~9,730 responses)
├── OFI - New User Survey.csv            # New-user survey data (~24,650 responses)
│
├── opera_heatmap_analysis.py            # Discovery method heatmap
├── opera_appealing_features_heatmap.py  # Feature appeal heatmap
├── opera_convince_heatmap.py            # Convincing factors heatmap
├── opera_default_browser_heatmap.py     # Default browser heatmap
├── opera_disliked_aspects_heatmap.py    # Disliked aspects heatmap
├── opera_ease_of_use_heatmap.py         # Ease of use heatmap
├── opera_installation_time_heatmap.py   # Installation tenure heatmap
├── opera_stop_using_reasons_heatmap.py  # Stop-using reasons heatmap
│
├── check_columns.py                     # Utility: list CSV columns
├── check_age.py                         # Utility: inspect age data
├── check_data.py                        # Utility: deep data inspection
├── verify_data.py                       # Utility: verify data structure
│
├── opera_appealing_features_heatmap.png # Output: feature appeal
├── opera_convince_heatmap.png           # Output: convincing factors
├── opera_default_browser_heatmap.png    # Output: default browsers
├── opera_disliked_aspects_heatmap.png   # Output: complaints
├── opera_discovery_heatmap.png          # Output: discovery channels
├── opera_ease_of_use_heatmap.png        # Output: usability ratings
├── opera_installation_time_heatmap.png  # Output: install tenure
└── opera_stop_using_reasons_heatmap.png # Output: churn reasons

---

License

Internal Opera Software analysis. Not intended for public redistribution of the underlying survey data.