fast-exp-analytics 0.1.4

Python toolkit for AB/ABC experiment analytics, duration planning, reporting and exports

fast_exp_analytics

Fast experiment analytics toolkit for A/B and A/B/C tests, MDE estimation, experiment duration planning, reporting, and chat-ready summaries

fast_exp_analytics is a Python library for analysts and product teams who need to quickly analyze online experiments, calculate experiment metrics, prepare readable summaries, export results to Excel, and estimate the required duration of future tests

The library is designed for practical analytics workflows in Jupyter Notebook: prepare data, describe metrics in a simple config, run A/B or A/B/C analysis, generate a styled table, export the result, and prepare a short message for stakeholders

---

Examples

Chat-ready experiment summaries

A/B test summary

A/B/C test summary

Included example notebook and reports

The repository contains a full example notebook and several generated report examples:

File	Description
examples/fast_exp_analytics_example.ipynb	Full notebook with the main workflow: A/B analysis, A/B/C analysis, chat messages, Excel export, MDE and duration planning
examples/report_ab_exp_id_4242.xlsx	Example Excel report for an A/B test
examples/report_abc_id_4242.xlsx	Example Excel report for an A/B/C test
examples/duration_plan_ab.xlsx	Example duration planning report
examples/duration_plan_ab_custom.xlsx	Example custom duration planning report

---

What the library does

fast_exp_analytics helps with three main tasks:

1. Experiment result analysis

• A/B test analysis
• A/B/C test analysis
• Comparison of control and test groups
• Calculation of absolute and relative metric changes
• Statistical significance checks
• Multiple-comparison correction for A/B/C tests

2. Experiment reporting

• Styled result tables for notebooks
• Short summaries for chats and task comments
• Excel export with experiment results
• Optional LLM-based review of experiment results
• Optional message sending with an attached report

3. Experiment planning

• MDE estimation for different experiment durations
• Required number of days for target MDE
• Planning for A/B and A/B/C experiments
• Support for different rollout shares
• Aggregation of historical data to experiment-unit level

---

Main use cases

Use fast_exp_analytics when you need to:

• quickly calculate A/B test results
• compare several experiment groups in an A/B/C test
• prepare a short experiment summary for a manager, product owner, or team chat
• export experiment results to Excel
• estimate whether the experiment has enough sensitivity
• calculate MDE for 7, 14, 21, 28 or more days
• estimate how many days are needed to detect a target effect
• reuse the same experiment analysis logic across notebooks and projects

---

Installation

From PyPI

pip install fast-exp-analytics

From GitHub

pip install git+https://github.com/your-username/fast_exp_analytics.git

For local development

git clone https://github.com/your-username/fast_exp_analytics.git
cd fast_exp_analytics

python -m venv venv
source venv/bin/activate

pip install -e ".[dev]"

---

Quick start

A full usage example is available in the notebook:

examples/fast_exp_analytics_example.ipynb

The notebook shows the main workflow:

• loading or generating experiment data
• configuring metrics
• running A/B analysis
• running A/B/C analysis
• styling result tables
• building dashboard URLs
• generating chat messages
• adding optional LLM review
• exporting reports to Excel
• planning experiment duration and MDE

---

Basic workflow

import pandas as pd

from fast_exp_analytics import (
    run_ab_test,
    style_table_ab,
    build_ab_chat_message,
    export_ab_results_to_excel,
)

1. Prepare experiment data

The input data should usually be aggregated at the experiment unit level

Example:

user_id	exp_group	shows	clicks	amount	goals	is_create_ad
1001	A	1200	54	320.5	7	1
1002	B	980	41	210.0	3	0
1003	A	1500	73	460.7	9	1

The library does not force a specific data source. Data can come from SQL, YQL, ClickHouse, CSV, Excel, pandas transformations, or any other pipeline

---

2. Describe metrics

Metrics are configured with a simple pandas.DataFrame

metrics_df = pd.DataFrame.from_dict(
    {
        "shows": ["Shows", "additive", "shows", "shows", "positive"],
        "clicks": ["Clicks", "additive", "clicks", "clicks", "positive"],
        "amount": ["Amount", "additive", "amount", "amount", "positive"],
        "ctr": ["CTR", "ratio", "clicks", "shows", "positive"],
        "cpc": ["CPC", "ratio", "amount", "clicks", "negative"],
        "cpa": ["CPA", "ratio", "amount", "goals", "negative"],
        "cr_created": [
            "Created campaign conversion",
            "share",
            "is_create_ad",
            "is_in_exp",
            "positive",
        ],
    },
    orient="index",
    columns=["desc", "type", "num", "den", "direction"],
)

Metric config columns:

Column	Meaning
desc	Human-readable metric name
type	Metric type: additive, average, ratio, share
num	Numerator or value column
den	Denominator column
direction	Direction of improvement: positive or negative

---

3. Run A/B test

df_result_ab = run_ab_test(
    df=df,
    metrics_df=metrics_df,
    exp_start_date="2026-03-17",
    exp_end_date="2026-03-24",
    group_base="A",
    group_exp="B",
    alpha=0.05,
    power=0.80,
)

---

4. Display result table

style_table_ab(
    df_result_ab,
    comment="Experiment description",
    exp_start_date="2026-03-17",
    exp_end_date="2026-03-24",
    experiment_id=4242,
)

---

5. Export result to Excel

excel_path = export_ab_results_to_excel(
    df_result_ab=df_result_ab,
    output_path="report_ab_exp_id_4242.xlsx",
    experiment_desc="Experiment description",
    exp_id=4242,
    date_from="2026-03-17",
    date_to="2026-03-24",
)

---

6. Build short message

msg = build_ab_chat_message(
    df_result=df_result_ab,
    experiment_desc="Experiment description",
    exp_id=4242,
    date_from="2026-03-17",
    date_to="2026-03-24",
    dashboard_url="https://example.com/dashboard",
    max_metrics=10,
    show_days_more=True,
)

print(msg)

Example output:

AB test: Experiment description
Experiment: 4242
Period: 2026-03-17 - 2026-03-24

🟢 Amount: 1.61M → 8.15M | Δ 405.0% | p=<0.0001
🔴 CPA: 34.21 → 171.58 | Δ 401.6% | p=<0.0001
⚪ CTR: 4.969% → 5.010% | Δ 0.8% | p=0.7796

Dashboard:
https://example.com/dashboard

---

A/B/C analysis

For A/B/C experiments, use run_abc_test.

from fast_exp_analytics import (
    run_abc_test,
    style_table_abc,
    build_abc_chat_message,
    export_abc_results_to_excel,
)

df_result_abc = run_abc_test(
    df=df,
    metrics_df=metrics_df,
    exp_start_date="2026-03-17",
    exp_end_date="2026-03-24",
    include_bc=True,
    alpha=0.05,
    power=0.80,
    pvalue_adjust_method="holm",
)

Prepare a styled table:

style_table_abc(
    df_result_abc,
    caption="Experiment description: 2026-03-17 — 2026-03-24",
)

Generate a short summary:

msg = build_abc_chat_message(
    df_result=df_result_abc,
    experiment_desc="Experiment description",
    exp_id=4242,
    date_from="2026-03-17",
    date_to="2026-03-24",
    dashboard_url="https://example.com/dashboard",
    alpha=0.05,
    p_adjust_method="Holm",
    use_adjusted=True,
)

print(msg)

Export the report:

excel_path = export_abc_results_to_excel(
    df_result_abc=df_result_abc,
    output_path="report_abc_id_4242.xlsx",
    experiment_desc="Experiment description",
    exp_id=4242,
    date_from="2026-03-17",
    date_to="2026-03-24",
)

---

Experiment duration planning

The library can also help estimate experiment duration and MDE using historical data.

from fast_exp_analytics import (
    build_experiment_level,
    default_duration_metrics_config,
    mde_table_for_experiment_duration,
    required_days_for_target_mde_table,
    export_duration_results_to_excel,
)

Aggregate historical data to experiment-unit level

agg_df = build_experiment_level(
    df_raw,
    unit_id_col="user_id",
    date_col="date",
    amount_col="amount",
    shows_col="shows",
    clicks_col="clicks",
    goals_col="main_goals",
    entity_id_col="campaign_id",
)

Calculate MDE for different durations

metrics_duration_df = default_duration_metrics_config()

mde_by_days = mde_table_for_experiment_duration(
    df_raw=df_raw,
    unit_id_col="user_id",
    rollout_pct=0.5,
    exp_days=[7, 14, 21, 28, 35],
    experiment_type="ab",
    metrics_config=metrics_duration_df,
    alpha=0.05,
    power=0.80,
    date_col="date",
    amount_col="amount",
    shows_col="shows",
    clicks_col="clicks",
    goals_col="main_goals",
    entity_id_col="campaign_id",
)

Estimate required days for target MDE

days_for_target_mde = required_days_for_target_mde_table(
    df_raw=df_raw,
    unit_id_col="user_id",
    rollout_pct=0.5,
    target_mde_pct=[5, 7, 10],
    experiment_type="ab",
    metrics_config=metrics_duration_df,
    alpha=0.05,
    power=0.80,
    date_col="date",
    amount_col="amount",
    shows_col="shows",
    clicks_col="clicks",
    goals_col="main_goals",
    entity_id_col="campaign_id",
)

Export planning result

excel_path = export_duration_results_to_excel(
    output_path="duration_plan_ab.xlsx",
    mde_by_days_df=mde_by_days,
    days_for_target_mde_df=days_for_target_mde,
    experiment_name="Experiment name",
    experiment_type="ab",
    rollout_pct=50,
    recommended_days=21,
    comment="The estimate is based on historical data at user_id level.",
)

---

Core functions

The main public functions are grouped by workflow

A/B tests

Function	Purpose
run_ab_test	Runs A/B test analysis
style_table_ab	Builds a styled result table for notebooks
build_ab_chat_message	Generates a short text summary
build_dashboard_url_ab	Builds a dashboard URL with parameters
export_ab_results_to_excel	Exports A/B results to Excel

A/B/C tests

Function	Purpose
run_abc_test	Runs A/B/C or multi-group analysis
style_table_abc	Builds a styled result table for notebooks
build_abc_chat_message	Generates a short text summary
build_dashboard_url_abc	Builds a dashboard URL with parameters
export_abc_results_to_excel	Exports A/B/C results to Excel

Duration planning

Function	Purpose
build_experiment_level	Aggregates raw historical data to experiment-unit level
units_per_day_from_raw	Estimates available units per day from raw data
default_duration_metrics_config	Returns default metrics for duration planning
mde_table_for_experiment_duration	Calculates MDE for different experiment durations
required_days_for_target_mde_table	Estimates required days for target MDE
duration_plan_summary	Builds a full duration planning summary
export_duration_results_to_excel	Exports duration planning results to Excel

Optional integrations

Function / Config	Purpose
OpenAICompatConfig	Config for OpenAI-compatible LLM API
build_llm_review	Generates optional LLM review of experiment results
ChatSendConfig	Config for chat/message API
send_chat_message	Sends message and optional file to chat

---

Supported metric types

Type	Description	Examples
additive	Sum-based metrics.	revenue, amount, clicks, shows, goals
average	Average value per unit.	average amount per user, campaign days
ratio	Numerator divided by denominator.	CTR, CPC, CPA, CPM
share	Share/conversion metrics.	conversion to action, payer share

---

Project structure

fast_exp_analytics/
├── fast_exp_analytics/
├── examples/
├── tests/
├── README.md
├── CHANGELOG.md
├── pyproject.toml
├── MANIFEST.in
└── Makefile

---

License

MIT © 2026 Alexey Voronko