Why Engineering Teams Are Moving to BYO Storage for Bug Reporting

Every bug report that contains a screen recording is, by definition, a copy of something that happened inside a real user's session. It may show a checkout form with a half-typed card number, an internal admin panel with customer emails in the table, or a staging dashboard that reveals a feature nobody has announced yet. Teams have been willing to ignore this for years because the recording tools were so convenient. That tolerance is ending. Security questionnaires now ask where screenshots are stored, how long they live, and who at the vendor can read them.
This article walks through the architecture decision behind bring-your-own-storage (BYO storage) for bug reporting. It explains what changes when the media never touches the vendor's servers, what risks remain, and how an engineering team can evaluate the model before rolling it out. The examples reference BugSnap, the Chrome extension we build, because the design choices are concrete enough to inspect. The reasoning applies to any tool that captures media from a browser.
Why the default architecture is a problem
Most visual bug reporters follow the same pattern. The browser uploads a WebM video or a PNG to a multi-tenant object store owned by the vendor, the vendor generates a share link, and the metadata lands in the vendor's relational database. That pattern is simple to build and easy to scale. It also creates three liabilities that grow with every report:
- Data sprawl. Sensitive values such as account numbers, session tokens, and internal feature names end up in a bucket you do not control, with a retention policy you did not negotiate.
- Blast radius. A misconfigured share link or a vendor-side breach exposes every recording the vendor has ever stored, not only the ones your team created.
- Exit cost. When you leave the vendor, the archive is either unavailable or locked behind an export queue. Your evidence lives somewhere else.
None of these problems is unusual to a security reviewer. They appear in every vendor assessment for SOC 2 Type II, ISO 27001, and HIPAA-scoped engineering work. The question for QA leads is whether the tool can answer them without a long procurement exception.
What BYO storage changes
In a BYO model the file path is different. The extension records the screen inside the tester's browser, and the resulting file is uploaded directly to a Google Drive folder that belongs to the tester's organization, using the tester's own Google account and an OAuth grant. The vendor's backend coordinates the metadata: a title, a timestamp, the page URL, the browser and operating system, and a pointer to the Drive file. It does not receive the video bytes.
Three details matter for an engineer reading the design:
- Scope. The OAuth request asks for the narrowest Drive scope that still lets the extension create and update files it made itself. Google calls this scope
drive.file. It does not grant access to the rest of the Drive, which is also what makes the consent screen easier for security teams to approve. - Ownership. The file lives in the organization's Drive. Retention, sharing, deletion, and audit logging come from Google Workspace, which most teams already govern.
- Metadata only on the vendor side. A capture row in the vendor database is a small record. In our production measurements the average row for a capture is roughly one to two kilobytes, because the heavy artifacts live in Drive.
The data flow, step by step
Here is the path a single bug report takes in a BYO design, written as a checklist an auditor can follow:
- The tester presses record. The extension captures frames and the event stream locally in the browser.
- Sensitive values are redacted in the browser before anything is serialized. Known token shapes, cookies, and authorization headers are replaced with placeholders.
- The video or screenshot is uploaded to the tester's Drive folder. The browser refreshes its access token when needed.
- The developer-facing telemetry (console errors, network failures, user steps) is uploaded as a JSON file into a
logssubfolder next to the media. Only a pointer to that file is stored in the vendor database. - The extension calls the vendor API with the Drive link and metadata. The API validates the payload and stores the row.
- When someone opens the share page, the server issues a short-lived signed download URL that points back to the Drive file. The media is streamed through a signed link, not copied into vendor storage.
Step two deserves emphasis. Redaction that happens after the data leaves the browser is only a promise. Redaction that happens in the browser means the vendor never sees the secret in the first place. Write a test that feeds known secret shapes through the redaction function and asserts they are gone before the payload is built. This is the kind of control that survives an audit.
Threat model: what BYO does and does not solve
BYO storage is a reduction of risk, not an elimination. A short threat model keeps the conversation honest:
- Solved: vendor-side bucket exposure, long-lived vendor archives, vendor employees reading recordings, and most cross-tenant leaks.
- Reduced: the amount of PII that can leave the browser, because redaction runs locally and the redaction rules are testable.
- Not solved: a tester who records a sensitive screen and shares the Drive link with the wrong person. Drive sharing settings and Workspace DLP policies remain the control here, and the organization owns them.
- Not solved: a compromised browser profile with an active Drive session. The fix is the same as for any Google account: device management, session policies, and revocation.
Writing this table into the rollout document is worth the time. Security reviewers trust a vendor that names its own limits.
How to evaluate a BYO bug reporter in one afternoon
If you are comparing tools, run this short test plan. It takes less time than reading a vendor's security whitepaper and tells you more:
- Where do bytes go? Open the browser network panel during a recording. Confirm that the video upload targets the Drive API, not the vendor domain.
- What scope does consent ask for? Read the OAuth consent screen. A narrow file-level scope is a good sign. Broad Drive access is a reason to ask questions.
- What does the vendor store? Ask for a sample row. Count the columns. If the row contains large text blobs or base64 images, the storage claim needs a second look.
- Can you leave? Disconnect the integration and check whether your recordings still open in your own Drive. Confirm the share links keep working or fail cleanly.
- Does redaction run where you think? Put a fake token in a page header during a test session and check that the uploaded log file does not contain it.
Rollout checklist for engineering teams
- Create a dedicated Drive folder for bug evidence, with a Workspace group as the owner rather than a personal account.
- Set a retention rule. Many teams keep recordings for ninety days and keep only the telemetry JSON for longer.
- Enable audit logging in the Google Workspace Admin console and review who shares evidence externally.
- Document which fields the telemetry captures and which ones it redacts. Keep the list next to the threat model.
- Add one automated test that fails the build if a known secret shape appears in a generated payload.
- Train testers to avoid recording payment forms and identity documents. Redaction helps, but it is not a substitute for judgment.
Where this leaves QA teams
The argument for BYO storage is not that vendors are careless. It is that the cheapest place to remove a risk is the place where the data never gets copied. A bug report should carry enough context to reproduce the problem, and that context does not require a permanent copy of the customer's screen on someone else's cloud. When the architecture makes that copy unnecessary, the security conversation becomes shorter, and QA can get back to finding bugs.
Found this article valuable? Share it with your team
Help fellow developers and QA engineers triage bugs faster and maintain reliable web apps.