Session Replay for Product Managers

A funnel report can tell you that a share of the people who opened a new onboarding flow never finished it. It cannot tell you that many of them stalled on a date picker that opened off screen. Session replay for product managers closes that gap: the numbers say where, the recordings say why.

This page covers where replay fits next to the analytics you already have, what to do with it in the week after a launch, and how AnyReplay handles the two things that usually get in the way: volume and language.

The problem: analytics measure, they do not explain

Funnels, heatmaps and event dashboards are aggregate tools. A funnel counts users at each step; a heatmap shows where clicks land. Both are good at finding a step with a problem and bad at saying what the problem is. A drop between step two and step three could be a confusing label, a validation error, a slow request or a button that only fails on narrow screens.

The conventional answer is to instrument more events, ship, wait a week for enough data, and guess again. Each iteration costs a release cycle. Interviews and usability tests supply the qualitative side, but they observe five people in a lab rather than five hundred in production.

Replay is the qualitative tool that works at production scale. There is a longer comparison in session replay vs analytics; the short version is that they answer different questions and you want both.

How replay fits a product workflow

The most valuable hour with replay is the one after a launch. Ship the feature, wait for the first few dozen sessions, and watch them. You are looking for hesitation, back-and-forth between two screens, dead clicks on things that look interactive but are not, and rage clicks where something that should respond did not. AnyReplay flags chunks with hasRageClick and hasError, so you can start with the sessions that carry a flag rather than a random sample.

The second use is answering a specific question from analytics. When the funnel shows a drop at the plan-selection step, find sessions that reached it and did not proceed, and watch ten. Ten sessions are usually enough to find the pattern; a hundred are enough to be sure it is the pattern.

The third use is empathy at scale for a distributed team. A product manager who watches a user in Ankara fight an address form learns something a chart cannot carry into a roadmap meeting, and the customer support team can point at the same recording when the ticket arrives.

What AnyReplay does for product managers

Control volume with a sampling rate

Recording every visitor of a high-traffic product is more data than a product manager needs. The web SDK's sampleRate option takes a value from 0 to 1 and is decided once per visitor, so a recorded person stays recorded across pages and a sampled-out one is never half recorded. Set it to 0.1 to keep one visitor in ten; raise it for the week after a launch and lower it again afterwards. The sampling rate entry explains the trade-offs.

Watch users from any market in your own language

If you ship to several countries, the interface your users see is not the interface you designed in. Session translation translates the recorded UI on demand before playback, keeping labels as short as the originals and leaving product names and placeholders unchanged. Translations are stored per organisation, so once your product's strings are translated, the next hundred sessions from that market cost nothing extra. The translation memory entry has the details.

A dashboard your whole team can read

The dashboard ships in English, Turkish, German, Spanish and French, and every locale carries the same keys. A product team split across Istanbul and Madrid can share a session link and each person sees the dashboard in their own language while watching the same recording.

Web and mobile in one place

Web sessions are recorded with the @anyreplay/browser SDK, which wraps rrweb and records the DOM rather than video. React Native sessions are recorded with the React Native SDK as a semantic element tree. Both play back in the same dashboard, and both are translatable, because both are made of text rather than pixels.

What is never recorded, and what you can mask

Card numbers, CVVs, passwords and one-time codes are never recorded, whatever the configuration says, and masking happens in the browser so the originals never reach a server. Every other form value, email and phone included, is masked by default too. For a product manager that means the answer to "can we record the checkout?" is yes; add maskAllInputs: false only if you need to see what people typed. The privacy page lists the rules.

Getting started

1. Create a project. The Free plan includes one project and 500 sessions per month, which is enough for a first experiment on one feature; see pricing for what comes after.
2. Install the SDK snippet on the pages that carry the feature you want to watch, or run npm install @anyreplay/react-native for a mobile app.
3. Ship the feature with sampleRate: 1 for the first days so you have enough sessions to learn from.
4. Watch. Start with chunks flagged hasRageClick and hasError, then move to sessions that reached the step your funnel says is losing people.
5. Lower sampleRate once you have what you need, and write down what you changed so the next launch starts from a better baseline.

Frequently asked questions