PostHog

PostHog A/B Testing

A/B test setup and ideation with PostHog Experiments, implemented via GTM or directly in your codebase. We review existing event data to inform test ideas, configure feature flags, set up exposure events, define success metrics, and validate consistent variant assignment.

TLDR

PostHog A/B testing via GTM or in-code: event data reviewed for ideation, feature flags configured, exposure events set up, success metrics defined, experiments running.

Best for

Teams with PostHog already collecting data who want a structured way to run experiments, without guessing what to test or misreading results.

  • 8-14 hours
  • Event data reviewed before ideation, not only after
  • Implemented via GTM or directly in your codebase

Depending on your existing setup, some steps below can be shortened or skipped entirely.

Delivery roadmap

How we deliver PostHog A/B Testing.

Step 01

Data review and ideation

We review your existing PostHog data (pageviews, funnels, drop-offs) to identify where a test is worth running. Hypotheses come from observed behaviour, not assumptions.

Step 02

Hypothesis and metric definition

Each test gets a defined hypothesis and a single success metric before any configuration starts.

Step 03

Feature flag and variant setup

PostHog feature flags are configured via GTM or in your codebase with variant assignment logic, holdout groups, and correct rollout percentages.

Step 04

Exposure event tracking

Exposure events are configured so PostHog records which users actually experienced each variant. Required for valid statistical analysis.

Step 05

Results, iterate

We deliver a guide for reading results correctly. Once a test concludes, findings feed back into the next round of ideation.

Overview

How to know if you need PostHog A/B Testing

  • You have PostHog collecting data but no structured process for using it to decide what to test.
  • You have test ideas but no way to define hypotheses, assign variants, and track exposure events correctly.
  • Previous tests produced unreliable results because exposure events were missing or variant assignment was inconsistent.
  • You need to know when a test has reached a valid conclusion rather than calling it early.

What PostHog A/B Testing covers

  • Existing PostHog event data reviewed to identify where tests are worth running before any ideation. Or events implementation to gather discovery data
  • Test hypotheses defined with a specific success metric for each experiment.
  • Results guide delivered so your team can read experiment data correctly and know when to call a test.

PostHog A/B Testing outcomes

  • Experiments running with correct variant assignment, exposure tracking, and success metric events.
  • A repeatable testing process grounded in PostHog data: ideate, implement, gather data, repeat.

PostHog A/B Testing scope and hours

  • 8-14 hours.
  • Traffic volume verified at scoping. We confirm statistical significance is achievable before setting up any test.
  • GTM implementation: flicker may occur for visual changes since GTM fires after page load. We mitigate it where possible but cannot always eliminate it entirely.
  • In-code implementation: no flicker.

What makes our PostHog A/B Testing setup different

  • We start with your existing PostHog data before suggesting what to test. Ideation grounded in observed behaviour produces better hypotheses than starting from assumptions.
  • Hypotheses are defined based on real data.
  • Tracking implementation is verified to ensure events are collected as expected.
  • Implementation is GTM or in-code based on your stack.

More PostHog services

View all PostHog services
FREE AUDITS

Want to try before you commit?

  • Looking to try out Beluacode but not sure about it? You can always ask us for a free audit.
  • Free, no strings attached. We deliver the results and wish you a great day.
  • You will not be forced to talk to us, no call needed to release the results :)
Free

Free Generic Tracking

A high-level review of your overall tracking setup: event coverage, obvious configuration errors, attribution quality, and the highest-impact quick wins across your analytics and ad platforms.

Free

Free GTM Container

A surface-level review of your Google Tag Manager container: tag count, obvious duplicates, missing consent triggers, broken references, and the highest-impact issues. Delivered as a written summary.

Free

Free Consent Mode

A basic review of your Google Consent Mode v2 implementation: whether it is configured, which mode is active (Basic or Advanced), and whether consent states appear to be respected by your key tags.

Free

Free Website Performance

A Lighthouse-based snapshot of your Core Web Vitals on key page templates, with a basic assessment of how your tag stack is affecting page speed. Delivered as a written summary with the top tag-related performance issues identified.

Free

Free GA4 Analytics

A spot-check of your GA4 property: event volume, obvious tracking gaps, and a comparison of your reported conversion numbers against expected behaviour. Delivered as a written summary with the top issues identified.

Frequently Asked Questions

What is a PostHog feature flag?
A PostHog configuration that controls which users see which variant. It evaluates at page load and returns a variant assignment that GTM or your application code acts on to show different content.
What is an exposure event?
An event that fires when a user actually sees the variant they were assigned. Without it, PostHog cannot separate users who were assigned a variant from users who experienced it, which invalidates the statistical analysis.
Does this have to go through GTM?
No. PostHog experiments can be implemented directly in your application code. That avoids the GTM flicker issue and is the better choice when developer access is available. GTM is the right route when the goal is to avoid code deployments for every test.
What is the GTM flicker issue?
GTM fires after the page loads, so users may briefly see the control before the variant is applied. For visual changes, this is the main downside of GTM-based testing. We mitigate it where possible but a small flicker may still be visible.
How long do tests need to run?
Until statistical significance is reached given your traffic volumes. We estimate this at scoping. Calling a test early is one of the most common causes of invalid conclusions.
Do I need PostHog already collecting data before this service?
Yes. PostHog needs to be correctly installed and capturing events before we can review data for ideation or configure experiments. The PostHog Setup service covers that if it is not already in place.

Request a Quote

Review your selected services and submit your quote request.
We'll get in touch within 24 business hours.

Book a call
Quote Request

Added to your quote

Fill in your details and we will get back to you within 24 hours to scope the work. Or keep browsing and add more services first.