A user feedback widget is the smallest surface in most products and often the least useful one. Someone types “checkout broke”, hits send, and the inbox gets a sentence with no browser, no build, no page and no way to reproduce what happened. The message is honest, the context is missing, and the ticket becomes a guessing game between support and engineering.
A bigger form does not fix that. The fix is a link between the message and the recording of the session in which the person wrote it. When the widget submits, it carries a session identifier; when the message arrives, your team opens the replay from the same screen. In Prynt Replay the feedback widget is part of the recorder, so every submission points at the exact session it came from.
Here is how to build that loop: what to ask, what to attach, how to keep fields masked, and how to triage without watching an hour of video for a two-line bug report.
Decide what the widget is for before you add fields
Most teams pour bug reports, feature requests and sentiment into one box. Those are three jobs.
- Bug report: session, route, app version and the last few actions before submission.
- Feature request: who is asking, on which plan, and the page they were on when they wrote it.
- Sentiment: a rating and an optional comment.
Ask one open question, and drop the email field when the person is already identified in your app; asking again adds friction and reduces replies. The message itself should be the only required field.
Attach the session and the context you can query
Timing matters. Attach the session when the page loads, not when the widget opens: feedback about a page that broke two clicks earlier then lands on the right session, because that session is already recording.

Identify the user early and pass the account so sessions can be grouped later:
prynt.identify(userId, { accountId: "acct_123" })
From there, the widget submits with the session id and a timestamp. The message is free text; the context should be structured fields you can filter on: sessionId, timestamp, route, appVersion, userId, accountId, deviceClass (desktop, mobile web, iOS, Android) and category. Two of them do most of the work in triage: a cluster of reports on one route usually means one regression, and appVersion tells you whether it started with the last release. Our documentation covers the SDK calls, the secret-key endpoints for listing sessions and the embedded player.
Keep the widget from becoming a privacy hole
Feedback forms collect free text, and free text contains names, order numbers and card fragments. Replay masks every form field in the browser and again on the server, with no unmask option, so recorded sessions never show what was typed into the widget.
That boundary is deliberate. If you want to read the message, capture it through your own submission path as a product event you control, and tell people what you keep. Recording how someone moved through a page is not the same as recording the values in it. The same reasoning runs across the suite, from the shared agent behind every product to the terms in our glossary.
Triage the session, not the sentence
Open the replay before you open the ticket. Three signals usually explain a vague report faster than reading the message twice.

Rage clicks and dead clicks
Repeated clicks on the same element mean the interface did not respond. When the report says “the button doesn’t work”, the replay shows whether the element was disabled, covered by an overlay or slow to react.
JavaScript errors and Web Vitals
A JS error captured in the session often names the failing component. Web Vitals on the same page tell you whether the complaint is really about latency: a checkout that takes seconds to become interactive feels broken even when the request succeeds.
Py, the assistant
Py summarizes a session, explains what went wrong and sends a daily brief. It is included with Replay under a monthly limit per plan, and it turns a long session into two sentences an engineer can act on. From there, share links and comments carry the evidence to engineering, while the Zendesk, Intercom and Sentry integrations keep the session next to the ticket instead of in a separate tool.
Put the session where the team already works, then measure the loop
If your product is multi-tenant, embed the player in your own admin panel: identify with accountId, list sessions with the secret key and open them per company or per user. Support starts from the account the customer names, not from a timestamp.
Then watch three numbers: the share of submissions that arrive with a session attached, the time from submission to first session view, and the share of submissions whose session contains a JS error or a rage click. If the first is not close to all of them, the widget is firing before the identify call. Bot sessions do not count against replay quota, so a spike in recorded sessions reflects real traffic rather than noise.
FAQ

Does linking a user feedback widget to a session mean recording what people type?
No. Every form field is masked in the browser and again on the server, and there is no unmask option. You keep the interaction — pages, clicks, navigation, errors — not the values entered into fields.
Can we open the replay inside our own admin panel?
Yes. Identify the user with an accountId, list their sessions with the secret key and open them in the embedded player, per company or per user. The calls are in our docs.
What if the person is not logged in when they send feedback?
The session is still recorded and the widget can attach the anonymous session id. You lose the account link but keep the reproduction, which is usually enough to fix the bug.
Start with the free plan
Replay includes the feedback widget, heatmaps, JS errors and Py on the free plan: 1,000 sessions a month, 14-day retention and one site. Start at pryntid.com and attach a session to the first report you get.
Try it free
Prynt is device intelligence with a free plan: 5,000 identifications a month, visitor IDs with confidence and bot detection — no credit card. Upgrade to Pro for VPN, proxy, tampering and behavioral signals.