The short answer: Dovetail will give your data back, but not in one piece. A whole-project export arrives as an encrypted .dovetailproject file that only Dovetail can open. The spreadsheet export that any other tool can read does not preserve formatting, layout, highlights, or tags, which is precisely the analysis layer you spent two years building. Comments are excluded entirely. And any URLs inside your exported project data stop working three days after you run the export. Plan the extraction as a one-week project with a hard deadline, not an afternoon of clicking Download.
Most teams discover this in the worst possible order: they give notice, then try to get their data out. Do it the other way around.
Why Dovetail is harder to leave than it looks
A research repository is not a tool you use, it is a tool you accumulate into. The raw material - transcripts, recordings, notes - is usually the easy part, because you had it before Dovetail did. What you cannot easily recreate is the interpretation layer: the highlights your team pulled, the tag taxonomy you argued over, the insights that link back to evidence, and the comment threads where someone explained why a theme mattered.
That interpretation layer is the asset. It is also the part that travels worst.
Every platform fails this differently, which is why a generic migration checklist does not help much. UserTesting hands your data back in portable formats but as disconnected pieces you have to reassemble by hand. dscout exports almost everything cleanly but gates the export behind a role. Dovetail's failure mode is format: the complete version is locked, and the readable version is lossy.
This matters more in 2026 than it did two years ago, because teams are actively consolidating. In Perspective's 2026 mid-year tooling report, which surveyed 500 product, research, and CX leaders between February and April 2026 about line-item spend against their 2025 baseline, customer research tooling spend was down 12 percent year on year, legacy CXM seat counts dropped 38 percent on average, and 22 percent of enterprise buyers in the cohort had begun a full migration off their incumbent platform. Treat those figures as one vendor's survey of its own market rather than an audited census, but the direction is consistent with what the rest of the category reports.
The AI shift is the other pressure. Maze's Future of User Research Report 2026 found that 69 percent of respondents now use AI in at least some of their research projects, up 19 percent year on year. A repository designed around manual tagging is a worse fit for that workflow every quarter.
What actually comes out of Dovetail
Per Dovetail's own help documentation on downloading project data, these are the routes out:
| What you want | Format you get | What survives |
|---|---|---|
| A single video, highlight video, or highlight reel | .mp4 | The media only |
| A transcript from a video or audio file | VTT | Timed text |
| Raw data, highlights, tags, and Docs for a whole project | Spreadsheet (CSV) | Text and fields, not structure |
| An entire project, by a Manager or Contributor | Encrypted .dovetailproject | Everything - but Dovetail states the file is encrypted and cannot be opened or read outside of Dovetail |
| An individual note | Selectable text, formatting, layout, embedded images and files, and highlights |
Read that table twice, because the trap is in the gap between rows four and three. Dovetail does not hide this: its documentation states plainly that the project file is encrypted and cannot be opened or read outside of Dovetail. The complete export is the encrypted project file. The portable export is the spreadsheet. There is no row that is both complete and portable.
Dovetail is straightforward about the consequence: the spreadsheet export does not preserve formatting, layout, highlights, and tags. Its documentation describes the per-note PDF as the only semi-standard format that keeps selectable text together with formatting, layout, embedded images and files, and highlights - which is a reasonable claim, and also an admission that the structured, machine-readable path loses the structure.
What does not come out at all
Three exclusions are worth putting in your migration plan explicitly:
- Comments. The export does not include comments added to data within projects. Every "I think this is actually a pricing objection, not a usability problem" lives in comments. None of it ships.
- Channels data. Channels content is not included in the export.
- Working URLs. Any URLs found in your project data are only valid for three days from the day you exported your project.
That third one is the single most expensive detail in this article. It means a Dovetail export is a perishable artifact. If you export in week one of a 90-day wind-down and open the archive in week six to start the real migration, the links inside it have been dead for over a month. Export late and work fast, or export twice.
The extraction sequence that works
Run this before you give notice, while you are still a paying customer with full access and a responsive support contact.
Step 1 - Inventory before you extract. List every project, and for each one record how much of its value is raw data versus interpretation. Projects that are mostly transcripts are cheap to move. Projects that are mostly tags, highlights, and insights are the expensive ones, and they are where you should spend your effort.
Step 2 - Take the encrypted project export anyway. It is your insurance policy. You cannot read it without Dovetail, but if a dispute or an audit comes up later, having the complete archive beats having a lossy spreadsheet. Store it somewhere you will still have in three years.
Step 3 - Take the spreadsheet export for everything. This is the version your next tool can actually ingest. Accept up front that highlights and tags will arrive as text rather than as structure, and plan to re-establish the taxonomy in the new system rather than import it.
Step 4 - PDF the notes that carry judgement. For the subset of notes where the formatting and the highlights are the finding - the synthesis documents, the readouts, the insight write-ups - the per-note PDF is the only format that keeps them intact. This is manual, so scope it: pick the notes a future colleague would actually reopen.
Step 5 - Pull media and transcripts separately. Videos come out as .mp4 and transcripts as VTT, one route at a time. If your next platform transcribes on ingest, you may only need the media.
Step 6 - Screenshot or re-document the comment threads you cannot lose. Since comments do not export, the only way to keep a consequential decision trail is to copy it out by hand before the account closes. Do this for the three or four studies that drove real product decisions, not for everything.
Step 7 - Re-export immediately before cutover. Because of the three-day URL window, the export you migrate from should be days old, not months old.
For the contract and notice-period side of this - when to give notice, how to time the overlap, what to get in writing - our guide on how to switch user research platforms covers the 90-day sequence. If a data protection review is part of your exit, the research vendor DPA review guide covers the Article 28 clauses that govern deletion and return of data, and research data residency and international transfers covers where the copies actually live.
What to demand from the replacement
Having been through one lossy migration, write the exit terms into the next contract. Four questions separate a tool you can leave from a tool you cannot:
- Is the complete export readable without you? An encrypted vendor-specific archive is a backup, not portability. Ask for CSV and JSON.
- Does the structured export keep the analysis layer? Tags, themes, and highlights should come out as fields, not as flattened text.
- Do export artifacts expire? Perishable links and email-only delivery turn an export into a deadline.
- Is export a permission or a right? Confirm who on your team can run an export without filing a request.
Ask these during evaluation, when you have leverage, rather than during offboarding, when you have none.
Why teams move to Koji
Koji was built on the opposite premise: the research platform should not be the place your data gets stuck. Koji runs AI-moderated voice interviews, does the thematic analysis automatically, and produces a shareable report - so the interpretation layer is generated rather than hand-built, which also means it can be regenerated rather than rescued.
On portability specifically, Koji exports study data to CSV as an immediate download - no queue, no emailed link, no expiry window - and exposes the full structured analysis as JSON through the API, which is the same data Koji uses internally to build its reports. Transcripts are accessible per participant. There is no proprietary encrypted archive format, because there is no step where your data has to be trapped to be complete. The guide to exporting research data from Koji documents every route, including the API endpoint and the field list.
The structural difference is where the analysis lives. In a repository, your insight is a hand-made artifact stored in the vendor's schema, which is exactly why it does not travel. In Koji, interviews are built from six structured question types - open_ended, scale, single_choice, multiple_choice, ranking, and yes_no - so every answer already has a type, a stable question ID, and a known shape. An open-ended answer gets AI follow-up probing; a scale question becomes a distribution; a ranking becomes an average position. That typing is what makes a clean export possible, and the structured questions guide explains how to design a study that way.
For teams comparing the two directly, Koji vs Dovetail sets out the end-to-end-research versus analysis-only-repository distinction, and Dovetail alternatives covers the wider field.
The practical argument for moving now rather than at renewal: the Condens State of AI in User Research Analysis survey of 332 practitioners found that 71 percent agreed AI lets them analyse data significantly faster, while 71 percent also agreed that validating AI outputs still requires significant time, and only 21 percent trust AI output with minimal review. The Lyssna Research Synthesis Report 2025, covering 300 professionals, found 60.3 percent naming time-consuming manual work as their biggest synthesis challenge. A repository does not fix that; it stores the output of it. Koji compresses the question-to-insight loop to hours rather than weeks, without requiring research expertise to operate - and it does not hold the result hostage on the way out.
Ready to stop renting your own research? Start a Koji study, run real AI-moderated interviews this week, and export everything on day one to confirm it comes out clean before you commit a single project to it.
Frequently asked questions
Can I export everything from Dovetail in one file?
Yes, but not in a format you can use elsewhere. Managers and Contributors can export an entire project as an encrypted .dovetailproject file, which is complete but readable only by Dovetail. The portable alternative is the project spreadsheet export, which Dovetail notes does not preserve formatting, layout, highlights, and tags. There is no single export that is both complete and portable, so most teams take both.
Do Dovetail exports expire?
The export file itself does not disappear, but any URLs found in your project data are only valid for three days from the day you exported the project. In practice that makes an export perishable: if you migrate from an archive created weeks earlier, the links inside it will already be dead. Export immediately before you cut over, and re-export if your timeline slips.
Do highlights and tags survive a Dovetail export?
Not in the spreadsheet route. Highlights and tags are included in what you download, but not preserved as structure - they arrive flattened rather than as usable fields. The per-note PDF is the only format Dovetail describes as preserving highlights along with formatting and layout, but a PDF is not machine-readable for re-analysis. Expect to rebuild your tag taxonomy in the new platform.
Will my team's comments come across?
No. Dovetail states that the export does not include any comments added to data within projects. Channels data is also excluded. Since comments are often where the reasoning behind a theme was recorded, copy out the threads attached to your most consequential studies manually before the account closes.
How long should a Dovetail migration take?
Budget one focused week for extraction inside a longer wind-down. The extraction itself is not slow, but it is multi-route - spreadsheets, media, transcripts, and selective PDFs are four separate passes - and the three-day URL window means the final export has to happen close to cutover. Run a real study in the replacement platform in parallel before you cancel anything.
Is it worth migrating historical research at all?
Often only partially. Projects that are mostly raw transcripts are cheap to move and rarely reopened. Projects that drove real product decisions are worth the manual effort of PDFs and copied comment threads. A useful test: if nobody has opened a project in twelve months, archive the spreadsheet export and move on rather than reconstructing its tag structure by hand.
