You can record a bug without a connection, as long as the recording is saved on the device first and uploaded later. The report should never depend on the network at the moment you press Send. Capture the evidence locally, let a queue retry the upload when you reconnect, and give every report a unique ID so retries do not create duplicate tickets. This guide covers what to capture, where it lives while you are offline, how to confirm the sync, and what to do when an upload stalls.
TL;DR
- Save first, send second. A report that is stored on the device survives a dropped connection, a closed app or a dead battery.
- Capture the evidence that matters while the bug is fresh: a screen recording, a short voice or written description, screenshots and the device’s diagnostic logs.
- Give each report an ID when it is saved, so every retry carries the same one and the backend can discard repeats.
- If an upload sits in the queue, check the network, the app’s background permissions and available storage, then reopen the app to restart the job.
- Before a synced report becomes a backlog item, check it against existing tickets. Offline queues are a common source of duplicates.
Table of Contents
- Why the network is the wrong thing to depend on
- What you can capture without a connection
- Where an offline report lives and how to tell it synced
- How the upload works after you reconnect
- If a report stays queued or the upload fails
- Avoiding lost reports and duplicate tickets
- A field routine for offline UAT sessions
- Our perspective
- What a clean intake step looks like
- Capture and deduplicate with Wezard
- FAQ
- Sources
Why the network is the wrong thing to depend on
People report bugs when something is going wrong, and a bad connection is often part of what is going wrong. Menelaos Vergis makes this point in a write-up on building a feedback screen with no server: the obvious design, which sends the report at the moment the user taps Send, is the wrong time to rely on the network. The better design stores the report on the device first, then tries to send it, so the button never waits on a connection.
That is the rule behind everything below. For a UAT tester sitting in a warehouse, a train, a client site with guest Wi-Fi or a locked-down test lab, the question is not whether the connection will fail. It is whether the evidence will still be there when it comes back.
What you can capture without a connection
Recording the screen, writing notes and taking screenshots all happen on the device, so none of them need the internet. What changes offline is anything that needs a server: live transcription in the cloud, attaching the report to a ticket, and checking for duplicates.
Here is a practical split.
| Evidence | Works offline? | Notes |
|---|---|---|
| Screen recording of the failure | Yes | Stored locally as a video file; keep clips short so they upload faster later |
| Screenshots and annotations | Yes | Small files, easy to queue |
| Typed steps and expected vs. actual result | Yes | Write these while you remember the details |
| Device diagnostics and logs | Yes | Generated on the device itself |
| Cloud transcription of voice notes | Usually no | Needs a service, so it typically runs after reconnecting |
| Ticket creation in your tracker | No | Needs the tracker’s API |
| Duplicate check against the backlog | No | Needs access to existing tickets |
Diagnostic logs on Android
If you test on Android, the platform produces a rich bundle of evidence locally. According to Android Developers documentation, bug reports are saved by default in the /bugreports folder on the device and can be listed with adb shell ls /bugreports/. The zip file contains diagnostic output for system services, error logs and system message logs, including stack traces when the device throws an error.
Managed device fleets work similarly. Zebra’s StageNow documentation describes bug reports stored in a time-stamped zip file in device flash memory, which can also be e-mailed to a designated address or uploaded to the Zebra Cloud. The report includes the standard Android bug report text file and a screenshot. The same page notes that these files can sit in fixed or removable flash, limited only by available memory, and that the storage path you specify has to be valid when the tool runs. If you test on rugged handhelds, confirm that path before you head into the field.
Capture the story while it is fresh
A recording shows what happened. It does not always say what you expected. Spend 20 seconds narrating or typing three things:
- What you were trying to do.
- What you expected to happen.
- What actually happened, including any error text.
Our guide on submitting clear bug reports to developers goes deeper on the wording. If you are deciding between a still image and a clip, screenshot vs. video bug reports lays out when each one earns its place.
Where an offline report lives and how to tell it synced
While you are offline, the report exists in one place only: local storage on your device. It is not in Jira, not in Azure DevOps and not on anyone else’s screen. That makes the local copy the single source of truth, so treat it carefully.
A well-built offline flow gives you three visible states:
- Saved on device. The report is stored and waiting.
- Uploading. The queue has a connection and is sending it.
- Sent. The server has confirmed receipt, and the item now shows in the shared list or backlog.
Only the third state means the report has synced. A confirmation toast on your own screen is not enough, because it can appear when the report is merely saved. The reliable check is to open the destination, whether that is the feedback list or the tracker, and find the ticket with its attachments and recording playable.
Before you leave a reliable connection, do one more thing: note how many reports you recorded offline. When you reconnect, count what arrived. A mismatch tells you something is still queued.
How the upload works after you reconnect
Most mobile apps handle this with a background job that waits for a network condition and retries on failure. On Android, the WorkManager documentation explains that work can carry constraints such as requiring a network, that it runs only when all constraints are met, and that a worker can ask to be retried according to a linear or exponential backoff policy. If the connection drops mid-upload, the work is stopped and retried once the constraints are met again.
The practical consequences for a tester:
- Uploads are not instant. Backoff means each failed attempt waits longer than the last. A flaky connection can leave a report queued for a while even after the signal looks fine.
- The app may need to be opened. In the design Vergis describes, a force stop cancels scheduled jobs until the user launches the app again. If you swipe the app away to save battery, the queue may pause.
- Retries have a limit. That same sample gives up after a fixed number of tries, so a report can end in a failed state rather than retrying forever.
These details come from one implementation, and other apps behave differently. The principle holds across them: an upload queue retries on its own, but it is not guaranteed to finish without you.
If a report stays queued or the upload fails
Work through the likely causes in order, from cheapest to check to most disruptive.
- Confirm the connection actually works. Open a web page or a different app. Captive portals on hotel and guest Wi-Fi can show a connected icon while blocking traffic.
- Reopen the app. This restarts a paused queue, especially after a force stop.
- Check background permissions. Battery savers and data savers can stop a background upload. Allow the app to run in the background and use mobile data if you need it.
- Check free storage. A full device can fail while saving a recording, and a large recording can take long to send. Shorter clips help.
- Move to stronger Wi-Fi. Video is the heaviest attachment. Upload large recordings on a stable connection rather than a weak mobile signal.
- Keep the local copy until you have verified the ticket. Do not clear the app’s data or uninstall it while anything shows as queued.
- Resubmit only as a last resort. If you do, check first that the first attempt did not arrive. Otherwise you create the very duplicate you were trying to avoid.
If the report still will not leave the device after these steps, tell your QA lead or the person who administers the tool. They can see server-side whether anything arrived and whether a corporate firewall is blocking the endpoint.
Avoiding lost reports and duplicate tickets
Offline queues create two opposite risks: losing a report and sending it twice. The two fixes work together.
Do not lose it
Save to local storage before any attempt to send. This is the same principle described in the feedback screen write-up, and it is why the Send button can respond immediately even with no signal.
Do not send it twice
A queue that retries delivers a message at least once, which means it can deliver it more than once. The same article describes the fix: create an ID on the device at the moment the report is saved, and attach that same ID to every retry. The first request with that ID is stored as usual. A repeat gets a normal success reply flagged as a duplicate, and nothing new is stored. This technique is called an idempotency key, and if you build or buy a reporting tool, it is worth knowing about.
That protects against the same report arriving twice. It does not protect against two different testers reporting the same defect, which is the more common problem in UAT. When five people hit the same broken checkout button during a session and all of them sync at 5 p.m., the backlog gets five tickets. Our articles on why duplicate bug reports exist and on preventing duplicate Jira tickets at capture explain how to catch them before they pile up.
A field routine for offline UAT sessions
If your testers will be away from a reliable network, set a short routine before the session starts.
Before the session
- Confirm the testing app is installed, signed in and updated while you still have a connection.
- Free up device storage so recordings are not rejected.
- Allow background activity and notifications for the app.
- Agree on a naming habit, such as the test case ID at the start of every description.
During the session
- Record the failure as soon as it happens, then add the expected and actual result.
- Keep each recording to the single issue it shows. Our note on partial screen recording explains why a focused clip speeds up triage.
- Do not clear or reinstall anything while reports are pending.
After the session
- Connect to stable Wi-Fi and keep the app open until the queue empties.
- Check the destination to confirm every report arrived, with its recording and attachments.
- Review new tickets for duplicates before they reach developers.
For teams running structured acceptance cycles, the UAT testing feedback page shows how reports, recordings and tracking come together in one place.
Our perspective
Most teams treat connectivity as a tester problem. We see it as a workflow problem. The evidence of a bug is perishable: the tester remembers the steps today and not next week, and a recording made in the moment beats any reconstruction after the fact.
That is why we built Wezard around capture. Testers record the screen, describe the issue by voice or text, and get a transcript of what they said in real time, so the report is rich before anyone has typed a ticket. Enterprise teams also need what happens next to be tidy. Our AI checks for duplicate tickets before they reach your Jira or Azure DevOps backlog, and we designed this to cut duplicate noise by up to 40% compared with manual processes.
Whether a report is recorded in a meeting room or a basement, the standard should be the same: complete evidence, one ticket per issue, nothing that costs a developer an hour of guessing. A clean intake step is what makes late-arriving reports safe to accept.
Marketing
What a clean intake step looks like
When reports arrive in a burst after reconnecting, the intake step decides whether your backlog stays usable. A good one does four things:
- Attaches the evidence. Recording, screenshots, transcript and environment details come with the ticket, not in a follow-up thread.
- Checks for matches. Before a new ticket is created, the tool compares it to open issues. Our piece on duplicate bug detection explains why retrieval accuracy alone is not enough.
- Routes to the right place. Tickets land in the project and issue type your team already uses, which is easier with a direct Jira integration.
- Lets the tester fix the words. Transcripts of spoken notes contain errors, and a quick edit pass keeps the ticket accurate. See editing transcriptions for bug reports.
If your backlog is already crowded, cleaning up your Jira backlog with AI intake is a good starting point.
Capture and deduplicate with Wezard
Wezard is user acceptance testing software that records the screen, transcribes the issue in real time and uses AI to detect duplicate tickets before they reach Jira, Azure DevOps or ServiceNow. Testers can create tickets by voice or video, use guided submission, and add screenshots and files. Security is built for enterprise use, with SOC 2 certification and SSO.
Plans are published on the pricing page: Starter is $25 per month, Team is $70 per month and Business is $160 per month. Enterprise has no published price.
If you manage UAT cycles and want to see how reports move from recording to a clean backlog item, start with the use cases or pick a plan that fits your team.
FAQ
Can I record a bug with no internet connection?
Yes, if the tool saves the recording on the device first. Screen recording, screenshots and typed notes are all local actions. Steps that need a server, such as ticket creation and duplicate checks, happen after you reconnect.
How do I know my offline report uploaded successfully?
The only reliable proof is finding the ticket in its destination with the recording playable and attachments present. A message on your own screen can mean the report was merely saved. Count your offline reports and compare with what arrived.
Why is my report still queued after I reconnected?
Common causes are a connection that is not really working, an app that was force-closed, background restrictions from battery saving, or low storage. Reopen the app, allow background activity and move to stable Wi-Fi. Android’s background work system retries failed jobs with growing delays, so a stalled queue can take time to resume.
Will an offline report create duplicate tickets?
The same report should not arrive twice if the tool gives each report a unique ID when it is saved and reuses it on every retry. Two different testers reporting the same defect is a separate problem, solved by checking new reports against open tickets before they enter the backlog.
What should I capture for a bug I cannot reproduce later?
A screen recording, the expected and actual result, the exact steps, and device diagnostics where available. On Android, the platform’s bug report zip includes system logs and stack traces, which developers can use directly.
Sources
- A feedback screen for Android and Flutter with no server (DEV Community)
- Capture and read bug reports (Android Developers)
- Bug Report Manager (Zebra Technologies TechDocs)
- Define work requests (Android Developers)



