Most support conversations burn their first two or three exchanges on evidence. The agent asks for a screenshot, the customer sends a cropped fragment, and a day passes before anyone looks closer. Then it turns out a screenshot was never going to cut it, because the bug only shows up while clicking. The fix is not more evidence, it is the right evidence on the first ask. Think of it as a decision tree with four branches, and getting the branch right the first time can shave days off a ticket.
When a screenshot is enough
Screenshots win when the problem is a state, not a sequence. An invoice total that is wrong, a typo in a modal, a layout break at a specific viewport, a single error message with the full page visible around it. If one frozen frame contains everything the engineer needs, do not ask for a video. The classic failure mode is the cropped screenshot: a zoomed-in error message with no URL bar, no surrounding page, no indication of what came before. Fix the ask, not the format. Ask for the full browser window, the URL visible, nothing cropped out. A good screenshot with context beats a bad video every time, and it costs the customer ten seconds.
When a sequence demands a screen recording
The moment a problem involves motion, timing, or steps, a screenshot stops telling the truth. The button that does nothing when clicked, the checkout flow that fails after step four, the dashboard widget that loads, flickers, and vanishes, the report that exports fine on the second try but never the first. A screenshot of a loading spinner tells you exactly nothing. A sixty-second recording shows the click target, the pause, the error, and the exact order of events. It also captures the things customers never think to describe, like a console warning flashing for half a second or a page that silently reloads mid-flow. For QA teams, recordings of the reproduction path are the closest thing to watching the bug happen in person. They turn vague reports into a deterministic script the developer can follow. If the customer can show the failure happening, stop asking for written steps to reproduce and ask for the recording instead. It is faster for them and far more useful for you.
The intermittent bug: always record
Some bugs only exist in the moment. Sometimes the save fails. It was broken this morning, it works now. By the time an agent opens the ticket, the state is gone, and no screenshot taken later will recover it. For intermittent issues, a screen recording is the only reliable evidence, and the window to capture it is small. Keep a recording link in your macros and ticket templates so the customer can capture the bug the next time it happens, not describe it from memory. One more detail that matters here: ask the customer to record the thirty seconds before the failure too, not just the error state. The setup is usually where the clue lives.
When only a live session will do
Live screen shares are the most expensive evidence you can collect. They consume an agent, they need scheduling, and they put a frustrated customer on the spot. Save them for the problems that genuinely need interaction: permission and SSO issues tied to the customer's own identity provider, bugs that only reproduce inside the customer's network or device fleet, and integration failures where you need to poke at their environment in real time. Even here, start with a recording. Half the time, the recording reveals that the environment issue was a sequence issue all along, and the live session gets cancelled. When a live session is unavoidable, the recording still gives the agent a head start instead of a cold start.
The combo ask for backend bugs
Some tickets need two kinds of evidence at once: the frontend behavior and the backend signal. A form that submits and silently fails, an export that returns the wrong data, a webhook that never fires. A recording alone shows the symptom, server logs show the cause, and neither is enough without the other. Make it one ask, not three. Tell the customer the exact two things you need: a short recording of the reproduction, plus a screenshot of the browser console and network tab from the same session. Specificity is what makes customers comply. Asking for more info gets ignored. Asking them to record the checkout and screenshot the console gets done.
A cheat sheet for the whole team
Put this in your internal wiki:
- Static, single state (wrong value, typo, layout): full-window screenshot, nothing cropped.
- Sequence, motion, timing (dead button, failing flow, flicker): screen recording.
- Intermittent (sometimes broken, gone by morning): a recording captured the moment it happens, plus the thirty seconds before.
- Account or environment specific (SSO, permissions, customer network): live session, only after a recording rules out the easy stuff.
- Backend or data issue (silent failure, wrong export): recording plus console screenshot, in one ask.
Make the recording ask friction-free
The real reason agents underuse recordings is not ignorance, it is friction. The typical recording workflow asks the customer to install software, create an account, or add a browser extension, right in the middle of a problem. Most customers quietly give up, and the agent falls back to screenshots. The fix is to remove the ask entirely: send the customer a link, they press record in their browser, they send it back. No install, no extension, no account. That is how ScreenFlowr works: the customer opens a link and records right in the browser, so the request gets answered instead of ignored. Keep recording asks in your saved replies so the decision tree becomes a reflex: could you record what happens when you click Export? Here is a link, it takes about a minute. Pair that with the cheat sheet above, and your team stops negotiating for evidence and starts diagnosing.
Close the loop faster
Evidence is the cheapest lever in support. The wrong ask costs days of back-and-forth, the right ask costs the customer a minute and hands your team everything they need. Pick the branch, send the right link, and close the ticket. If your team is still trading screenshots for problems that move, try ScreenFlowr on your next moving-bug ticket: one link, one recording, no installs for your customer.