Remote developers earn 51% more. Here's what 65,000 developers revealed.

Stack Overflow Developer Survey 2024 — Analytics Deep Dive

Live Demo → Streamlit Cloud

---

63,023 valid responses. 20 analytical SQL queries. 6 dashboard sections.
One dataset that reveals what the tech industry actually looks like in 2024.

---

Key Findings

#	Finding	Data
1	Remote developers earn 51% more than in-person	$83k vs $55k median
2	SQL is used by 69% of data professionals	Most universal data skill
3	Python dominates data roles — R is irrelevant	82% vs 20% adoption
4	Hybrid is the new normal — office-only is dying	42.5% hybrid · only 19.6% in-person
5	Senior devs (20+ yrs) go remote at 2× the rate of juniors	46% vs 24%
6	PostgreSQL overtook MySQL as #1 database	24.4k vs 19.9k users

---

Screenshots

📊 Hero & KPI Row

🔍 Key Findings — 6 Data-Driven Insights

💰 Salary by Programming Language

📈 Salary by Experience Level

🌍 Salary by Country — Geographic Pay Gap

🏠 Remote vs On-site Salary Comparison

🛠️ Tech Stack Rankings — Languages & Databases

☁️ Top Cloud Platforms

🌐 Remote Work Trends

👤 Developer Profile

🎯 Data Roles Focus — SQL · Python · R · Salaries

🤖 AI Tool Adoption Among Developers

---

At a Glance

Metric	Value
Survey responses	65,437 (63,023 after salary filtering)
Countries represented	185
Salary data points	21,021 (filtered $10K–$2M)
Median developer salary	$72,673 USD
Fully remote developers	37.9%
Most used language	JavaScript (35.8k users)
Most used database	PostgreSQL (24.4k users)
Top cloud platform	AWS (21.3k users)
SQL queries	20 analytical queries
Dashboard sections	6
CI/CD checks	23 smoke tests — all passing

---

SQL Showcase

This project is a SQL-first analytics project — every chart is powered by a query with real business logic.

Salary by Language — Window Function + Aggregation

SELECT
    TRIM(lang)                          AS language,
    ROUND(MEDIAN(salary_usd))           AS median_salary,
    COUNT(*)                            AS respondents
FROM survey,
     UNNEST(STRING_SPLIT(LanguageHaveWorkedWith, ';')) AS t(lang)
WHERE salary_usd IS NOT NULL
  AND LanguageHaveWorkedWith IS NOT NULL
GROUP BY language
HAVING COUNT(*) >= 200
ORDER BY median_salary DESC
LIMIT 15

Remote Premium — Conditional Aggregation

SELECT
    RemoteWork,
    ROUND(MEDIAN(salary_usd))   AS median_salary,
    ROUND(AVG(salary_usd))      AS mean_salary,
    COUNT(*)                    AS respondents
FROM survey
WHERE salary_usd IS NOT NULL
  AND RemoteWork IS NOT NULL
GROUP BY RemoteWork
ORDER BY median_salary DESC

Experience × Remote Rate — FILTER Clause

SELECT
    CASE
        WHEN YearsCodePro < 2  THEN '0-1 yrs'
        WHEN YearsCodePro < 5  THEN '2-4 yrs'
        WHEN YearsCodePro < 10 THEN '5-9 yrs'
        WHEN YearsCodePro < 20 THEN '10-19 yrs'
        ELSE '20+ yrs'
    END AS exp_bucket,
    ROUND(
        COUNT(*) FILTER (WHERE RemoteWork = 'Remote') * 100.0 / COUNT(*), 1
    ) AS remote_pct
FROM survey
WHERE RemoteWork IS NOT NULL AND YearsCodePro IS NOT NULL
GROUP BY exp_bucket
ORDER BY MEDIAN(YearsCodePro)

SQL Adoption Among Data Professionals — Subquery + UNNEST

WITH data_folks AS (
    SELECT ResponseId, LanguageHaveWorkedWith
    FROM survey,
         UNNEST(STRING_SPLIT(DevType, ';')) AS t(role)
    WHERE LOWER(role) LIKE '%data%'
)
SELECT
    COUNT(*) AS total_data_pros,
    COUNT(*) FILTER (WHERE LanguageHaveWorkedWith LIKE '%SQL%') AS use_sql,
    ROUND(
        COUNT(*) FILTER (WHERE LanguageHaveWorkedWith LIKE '%SQL%')
        * 100.0 / COUNT(*), 1
    ) AS sql_pct
FROM data_folks

---

Architecture

Stack Overflow Annual Survey 2024 (Kaggle CSV · 65K rows)
        |
        v  src/load_data.py
   DuckDB — survey table (63K rows after salary filter)
        |
        v  src/queries.py — 20 analytical SQL queries
   Query layer (KPIs · Salary · Stack · Remote · Profile · DA/DE)
        |
        v  scripts/upload_db.py
   HuggingFace Dataset (evgeniimatveevusa/so-survey-db · ~3.4 MB)
        |
        v  dashboard/app.py — auto-download on cold start
   Streamlit + Plotly — 6-section dashboard
        |
   Streamlit Community Cloud (always-on)
        |
   GitHub Actions — smoke.yml
   23 SQL checks on every push

Key design decision: DuckDB binary (~3.4 MB) lives on a HuggingFace Dataset. The dashboard downloads it on startup — zero database hosting cost, instant cold start.

---

Data Pipeline

# 1. Load & clean (src/load_data.py)
df = pd.read_csv("data/survey_results_public.csv", usecols=KEEP_COLS)
df["salary_usd"] = pd.to_numeric(df["ConvertedCompYearly"], errors="coerce")
df = df[(df["salary_usd"] >= 10_000) & (df["salary_usd"] <= 2_000_000)]
# Result: 63,023 rows · 20 columns

# 2. Load into DuckDB
conn.execute("CREATE TABLE survey AS SELECT * FROM df")

# 3. Upload to HuggingFace Dataset
api.upload_file(path_or_fileobj="data/survey.duckdb", ...)

Salary outliers removed: 2,414 rows (implausible values like $0, $1B, etc.)

---

CI/CD — 23 Smoke Tests on Every Push

on:
  push:
    branches: [master]
  pull_request:
    branches: [master]
  workflow_dispatch:        # manual trigger from GitHub UI

Every push triggers 23 assertions — verifying every SQL query returns valid, non-empty results:

PASS  kpi_total_responses    PASS  top_languages
PASS  kpi_median_salary      PASS  top_databases
PASS  kpi_remote_pct         PASS  top_cloud_platforms
PASS  salary_by_language     PASS  remote_by_country
PASS  salary_by_experience   PASS  remote_by_experience
PASS  salary_by_country      PASS  remote_distribution
PASS  salary_remote_onsite   PASS  experience_dist
PASS  salary_by_org_size     PASS  education_vs_salary
PASS  language_salary_hmap   PASS  data_roles_salary
PASS  remote_by_country      PASS  sql_usage_rate
PASS  top_countries_volume   PASS  python_vs_r
                             PASS  ai_tool_adoption
23/23 checks passed

---

Tech Stack

Layer	Tool
Data source	Stack Overflow Annual Developer Survey 2024
SQL engine	DuckDB 1.5.3 (in-process, zero config)
Data processing	Python · pandas
Dashboard	Streamlit + Plotly
DB storage	HuggingFace Dataset (binary file, auto-downloaded)
CI/CD	GitHub Actions (smoke tests · Node.js 24)
Deployment	Streamlit Community Cloud

---

Quick Start

git clone https://github.com/evgeniimatveev/so-survey-analytics.git
cd so-survey-analytics
pip install -r requirements.txt

# Download survey CSV from Kaggle:
# "Stack Overflow Annual Developer Survey 2024"
# Place as: data/survey_results_public.csv

python src/load_data.py           # CSV → DuckDB
python scripts/smoke_test.py      # verify all 23 queries
streamlit run dashboard/app.py    # launch locally

---

Project Structure

so-survey-analytics/
├── src/
│   ├── load_data.py      # CSV → DuckDB (salary filter, type coercion)
│   └── queries.py        # 20 analytical SQL queries
├── dashboard/
│   └── app.py            # Streamlit — 6 sections + Key Findings
├── scripts/
│   ├── upload_db.py      # DuckDB → HuggingFace Dataset
│   ├── download_db.py    # HuggingFace Dataset → DuckDB
│   └── smoke_test.py     # 23 assertions
├── .github/
│   └── workflows/
│       └── smoke.yml     # CI/CD — Node.js 24
├── requirements.txt
├── runtime.txt           # python-3.11
└── .gitignore            # data/ excluded — DB lives on HuggingFace

---

Data: Stack Overflow Annual Developer Survey 2024 · 185 countries · 65,437 responses · Built by Evgenii Matveev