
PostHog Session Replay: How to Use It Properly
Define the behavior question before capture, sample both experiment arms and outcomes, and check AI summaries against recordings and metrics; replay alone cannot establish cause or prevalence.
Read moreArticles, experiments and case studies from 99Ways.

Define the behavior question before capture, sample both experiment arms and outcomes, and check AI summaries against recordings and metrics; replay alone cannot establish cause or prevalence.
Read more
Click a funnel step to inspect the people who advanced or left, then see whether drop-off clusters in a segment without switching to user-based aggregation.
Read more
Route new-error and error-rate spike alerts to the Slack or Discord channels your team already watches, so issues surface without repeated dashboard checks.
Read more
Map trusted conversion events with ttclid and other matching signals; purchases need total order value, while Video Shopping Ads need product IDs and content type.
Read more
A promising segment may reflect chance, missing values, or a property changed by the treatment; reconcile groups to the total before planning a targeted follow-up test.
Read more
Preserve Reddit’s rdt_cid and a shared conversion ID when Pixel and API both fire; expect attributed totals to differ from PostHog because matching and attribution windows differ.
Read more
Variants can start collecting data while the measurement framework is unfinished; metrics can be added, changed, or removed later, but they still define how results are evaluated.
Read more
Use the latest GCLID for click-based uploads; without one, enhanced conversions need hashed customer data. Map each PostHog event to its intended Google conversion action.
Read more
The management view shows which saved insights depend on each variable, so you can check affected queries before renaming, replacing, or deleting a shared input.
Read more
This guide explains how to send conversion events such as CompleteRegistration, Lead, AddToCart, InitiateCheckout, and Purchase to Meta, along with all available customer context and event metadata, including email address, first and last name, phone number, address information, order value, currency, and other matching parameters that help improve attribution and event match quality.
Read more
Inline OR logic counts any selected event in one series, useful when naming drift left a single product outcome under several valid event names without creating a custom Action.
Read more
If the gap is understanding behavior and running tests, PostHog fits; if it is governing and routing customer data across tools, Segment fits. Some teams use both.
Read more