Latest / Software Testing with Fexingo: QA, Automation, and Reliable Software Engineering / How Session Replay Finds the Bugs Users Never Report
Transcript
- Lucas: You know that sinking feeling — support tickets roll in for a bug you can't reproduce, logs show nothing unusual, and the customer is frustrated. Luna: Right, the classic 'works on my machine' problem, except now it's costing you users. Lucas: Exactly. Today I want to talk about a technique that's been quietly solving that for years: session replay. It's basically a video-like recording of a real user's interactions — clicks, scrolls, hovers, form fills, even rage clicks where they mash a button that isn't responding. Luna: And this isn't just a QA novelty. I remember a case from a mid-size e-commerce SaaS — they were losing about four hundred thousand dollars a year in cart abandonment. Their analytics said checkout page exits were high, but no one knew why. Lucas: What did replay show them? Luna: That on mobile Safari — specifically iOS 15.4 — the 'Place Order' button would shift down two pixels after the keyboard dismissed. Users tapped where the button was half a second earlier, hit a dead zone, and got confused. They just left. Lucas: Two pixels. That's the kind of thing no automated test would catch — your visual regression suite might flag a layout diff, but you'd need a very tight threshold. And session replay caught it because a real human failed exactly where the test never thought to look. Luna: And the fix was a single CSS property change — 'touch-action: manipulation' to disable double-tap zoom on the button. But they never would have found it without the video. Lucas: That's the core value proposition. Session replay is not a replacement for unit tests or integration tests. It's a lens into the gap between what your test suite assumes and what real users do. Luna: So how does it actually work under the hood? I know it's not literally recording video — that would kill bandwidth and raise privacy nightmares. Lucas: Right. Most tools — like FullStory, LogRocket, or open-source options like RRWeb — capture a snapshot of the DOM and then record a sequence of mutations. Every scroll, every keystroke, every resize gets serialized as a JSON event stream. When you watch a replay, the player reconstructs those mutations in order. Luna: So it's more like a time-traveling debugger for the front end. Lucas: Exactly. And because you have the full DOM state at each moment, you can inspect elements — look at the computed styles, check what class names were applied, see the exact network response that populated that dropdown. It's like having the user's Inspector open. Luna: But there's a big 'however' — privacy. Recording everything a user does means you're capturing passwords, credit card numbers, personal messages. If you're not careful, you're storing that plaintext. Lucas: That's the number one concern, and the industry has built mitigations. Most session replay tools let you configure selectors to mask or exclude certain fields. You can tell it 'any input with type=password, don't record the value.' Or 'any element with class 'pii'', redact it client-side before it even leaves the browser. Luna: But the onus is on the engineering team to configure that correctly. I've seen audits where companies thought they were masking credit card fields, but a dev added a new payment form with a different class name, and suddenly replays show full numbers. Lucas: That's a real risk. The best practice is to treat session replay data like any sensitive production data — encrypt at rest, limit access, set retention policies. And consider using a privacy-first approach like 'record everything, then auto-redact on the server side using regex patterns.' But it adds complexity. Luna: Let's talk about false positives. If you watch a replay and see a user flailing, is that always a bug? Or could it be bad UX that's technically working as designed? Lucas: That's a great distinction. A replay shows behavior, not intent. Someone might click a non-clickable element repeatedly — that could indicate they expected it to be interactive. But it could also be a mis-click. The trick is to look for patterns: if 5% of users do the same weird thing, it's probably a UX issue. If one user does something bizarre, it's likely just a confused individual. Luna: And you need to filter out bots. Session replay tools often record crawlers or automated scripts, which can pollute your data. If you're seeing hundreds of replays from 'Mozilla/5.0' with no mouse movements, those aren't humans. Lucas: Yeah, good point. Most tools have bot detection built in, but it's worth double-checking your filters. Otherwise you're debugging a phantom. Luna: So when should a team invest in session replay? I think it's most valuable when you have a complex front end — lots of user interactions, dynamic content, third-party scripts that could break things. Lucas: I'd add: if your support team spends more than a few hours a week trying to reproduce reported bugs. That time is expensive and demoralizing. Session replay can cut that cycle from days to minutes. You get a link, watch the replay, see the bug, fix it. Luna: And it's not just for bug fixing. Product teams use it for UX research — watching how users actually navigate a form, where they hesitate, where they drop off. That's a whole other use case. Lucas: Absolutely. But for QA specifically, I think the biggest win is catching those heisenbugs that disappear when you try to inspect them. The ones that only happen under specific network conditions, or with a certain browser extension, or after a specific sequence of actions that's too tedious to document. Luna: Speaking of tools that catch things you didn't expect — you know, a couple of dollars a month from listeners is what keeps this podcast ad-free and independent. If you've gotten something useful out of these episodes, buy me a coffee dot com slash fexingo makes a real difference. Lucas: Yeah, honestly, listener support is what lets us dig into topics like session replay without worrying about sponsor constraints. So if that resonates, consider chipping in. Luna: Now back to session replay — one thing that surprises me is how few teams actually watch replays regularly. They install the tool, get a dashboard, but then never look at the videos until a crisis hits. Lucas: That's a common failure mode. Session replay is most powerful when it's part of your regular QA cadence. Some teams set up alerts: if a user encounters a JavaScript error, auto-record that session and flag it. Or if a rage click pattern exceeds a threshold, trigger a review. Luna: I've also seen teams use replay to validate that new releases don't break the user flow. They'll watch a handful of replays after a deployment to check that the checkout process still looks right on different devices. Lucas: Exactly. It's a safety net. You can't automate every visual check, but you can spot-check ten real sessions and feel a lot more confident than just green CI. Luna: So if a team is considering session replay, what's a good starting point? Lucas: I'd say pick a single critical user journey — checkout, sign-up, or whatever your core conversion is. Record only that page. Set up privacy masking. Then watch ten replays a week. You'll be amazed what you find. That two-pixel bug Luna mentioned — that kind of thing is everywhere. Luna: And don't forget to involve support. They're the ones who hear the complaints and have the best intuition for what's broken. Lucas: Great point. Session replay bridges the gap between support tickets and engineering fixes. It's one of those rare tools that makes everyone's job easier — if you use it right.