{"site":{"name":"Koji","description":"AI-native customer research platform that helps teams conduct, analyze, and synthesize customer interviews at scale.","url":"https://www.koji.so","contentTypes":["blog","documentation"],"lastUpdated":"2026-10-06T14:13:33.695Z"},"content":[{"type":"blog","id":"e6a13f13-0f57-4ba7-ac5b-19886b55ac1c","slug":"migrate-off-usertesting-2026","title":"How to Migrate Off UserTesting (2026): The Export Is Real, the Reassembly Is Manual","url":"https://www.koji.so/blog/migrate-off-usertesting-2026","summary":"UserTesting exports are accessible and in ordinary formats but arrive as disconnected artifacts: an Excel workbook of metrics, separately downloaded .mp4 videos, and transcripts as text, with no field linking a video to its metrics row. Condens, which documents importing UserTesting data, instructs users to delete unused sheets, remove header rows above the Username row, and remove calculated mean and median columns before import. Koji exports participant metadata, session data, summaries and quality scores in one CSV and full structured analysis as JSON via API.","content":"**The short answer:** UserTesting does let you export your data, and that is not the problem. The problem is that your research comes out as three unrelated piles - an Excel workbook of metrics, a folder of `.mp4` files, and transcripts as text - with nothing connecting a video to the row that describes it except the order you downloaded them in. The reassembly is manual, and it is the step every migration plan underestimates. Budget for the join, not the download.\n\n## The lock-in is assembly, not access\n\nResearch platforms get stuck in three different ways. Some trap your data in a proprietary format, as [Dovetail does with its encrypted project archive](/blog/dovetail-data-export-migration-2026). Some put export behind a permission, as [dscout does by making it a Researcher capability](/blog/dscout-data-export-migration-2026). UserTesting does neither: the exports exist, they are documented, and they are in ordinary formats. What UserTesting does not give you is a single artifact that keeps a session's video, its transcript, its task metrics, and your notes together as one object.\n\nThat sounds like a minor inconvenience until you multiply it by a few hundred sessions.\n\nThe best evidence for how much manual work this actually takes does not come from UserTesting. It comes from Condens, a research repository that publishes step-by-step instructions for importing UserTesting data - essentially a competitor's migration runbook. Those instructions are worth reading in full before you plan your own move, because they are specific about the cleanup.\n\nPer Condens's integration documentation, getting UserTesting metrics into another tool goes like this: open a test in UserTesting, check the box for each session you want to export, and from the menu icon on the lower left select Export to Excel, choosing the option that exports all sessions into one Excel workbook. Then the cleanup begins. Condens recommends moving the Metrics sheet to the first position or deleting the other sheets, removing the first few rows above the one containing the Username, and removing the columns that contain calculations such as mean and median. Only then is the file ready for a CSV or Excel import.\n\nVideos are a separate pass: open the test, check the box for each session you want, and select Download videos. Transcripts are a third route - UserTesting generates transcripts which can be imported as text. Session-level detail lives on a different sheet again, named Session details.\n\nSo the honest summary is four operations, one of which is spreadsheet surgery, and none of which produce a combined record. A tool that imports UserTesting data for a living still tells its customers to hand-edit the file first.\n\n## What this costs in practice\n\nThe reason this matters now is that the category is consolidating and these migrations are no longer rare. Perspective's 2026 mid-year tooling report surveyed 500 product, research, and CX leaders between February and April 2026 on their line-item spend against a 2025 baseline. It found customer research tooling spend down 12 percent year on year, average legacy CXM seat counts down 38 percent, 22 percent of enterprise buyers beginning a full migration off their incumbent, and 47 percent of teams cancelling or downgrading NPS-only tools. That is one vendor's survey of its own market rather than an audited census, so weight it accordingly - but a lot of teams are doing this work at once, and most of them are doing the reassembly by hand.\n\nThe second number worth holding onto: 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. If your next platform analyses sessions automatically, a large share of what you are about to hand-reassemble will be re-derived anyway. That changes what is worth migrating.\n\n## What actually exports\n\n| What you want | How it comes out | What you have to do after |\n|---|---|---|\n| Task metrics, written responses, rating scales | Excel workbook, optionally all sessions in one file | Delete unused sheets, strip header rows above Username, remove mean and median columns |\n| Session videos | `.mp4` files, downloaded per selected session | Match each file to its row by hand |\n| Transcripts | Text, generated by UserTesting | Re-associate with the session and the video |\n| Session-level details | A separate sheet named Session details | Join to the metrics sheet |\n\nOne caveat on sourcing, stated plainly: UserTesting's help centre is not publicly readable to automated tools, so the export mechanics above are sourced from Condens's published import instructions rather than from UserTesting's own pages. Plan terms, available export routes, and API access vary by edition and by contract. Confirm the specifics against your own account and your own agreement before you commit to a timeline - and if your account includes API access, ask your account team whether video downloads can be driven programmatically, because that is the difference between an afternoon and a fortnight.\n\n## The extraction sequence\n\nDo all of this while you are still a customer in good standing, before you give notice.\n\n**Step 1 - Decide what deserves to move.** Sort tests into three buckets: studies that still inform a live decision, studies that are reference material, and studies nobody has opened in a year. Only the first bucket justifies full reassembly. For the second, the Excel workbook alone is usually enough. For the third, do nothing.\n\n**Step 2 - Export metrics for everything in one pass.** Use the all-sessions-in-one-workbook option. One file per test is how migrations turn into months.\n\n**Step 3 - Clean the workbook once, then template it.** The cleanup Condens describes - sheet selection, header rows, calculated columns - is identical every time. Do it once, write down the exact steps, and hand the repetition to someone else or to a script.\n\n**Step 4 - Download videos only for bucket one.** This is the slow, storage-hungry part. Be ruthless. A video nobody will rewatch is a cost, not an asset.\n\n**Step 5 - Build the join key before you need it.** Username plus session plus test is usually enough to reconnect a video file to its metrics row. Decide the key and apply it to filenames during the download, not after.\n\n**Step 6 - Re-run one real study in the new platform in parallel.** Not a sandbox test - a study with a real decision attached. This is how you find out whether the new tool actually replaces UserTesting for your use case while you still have both.\n\n**Step 7 - Then give notice.** On timing, notice periods, and what to get in writing, see [how to switch user research platforms](/blog/switch-user-research-platform-2026). For the pricing side of the decision, [UserTesting pricing](/blog/usertesting-pricing-2026) breaks down seats, credits, and contracts, and [UserTesting alternatives](/blog/usertesting-alternatives-2026) covers the field. If your organisation is also folding in a UserZoom instance after the merger, [UserZoom alternatives](/docs/userzoom-alternatives-2026) is the companion piece.\n\n## What to require from the next platform\n\nWrite portability into the contract this time. The questions that matter:\n\n1. **Does one export contain one session?** Video, transcript, structured answers, and analysis in a single record beats four files you have to marry.\n2. **Is there an API, and does it cover media?** A UI-only export caps your migration speed at how fast a person can click.\n3. **Are structured answers typed?** Free text in a spreadsheet column is not a scale question. Typed fields survive a move; prose does not.\n4. **Who is allowed to export?** Confirm it is not gated behind an admin request.\n5. **Do exports expire?** Emailed links and short-lived URLs convert an export into a deadline.\n\nIf a vendor cannot answer all five in writing during evaluation, assume the answer you will get at offboarding is worse. [Enterprise security for AI research platforms](/docs/enterprise-security-ai-research-platforms) covers how these questions usually surface in a SOC 2 and vendor review, which is the right moment to ask them.\n\n## Why teams move to Koji\n\nKoji is built so that the record and the export are the same shape. Interviews are AI-moderated voice conversations, analysed automatically into themes with supporting quotes, and the output is a report you can share in one click - which means the analysis is generated, not hand-assembled, and therefore can be regenerated rather than reconstructed.\n\nOn the specific failure described in this article, the structural fix is typing. A Koji study is built from six structured question types - open_ended, scale, single_choice, multiple_choice, ranking, and yes_no - and each one has a stable question ID and a defined shape. An open-ended answer carries AI follow-up probing; a scale question produces a distribution; a ranking produces an average position. Because every answer is already typed against a known question, the export is a join that has already been done: the CSV export carries participant metadata, session data, summaries, and quality scores together, and the JSON export via the API carries the full structured analysis - the same data Koji uses internally to generate reports. The [exporting research data guide](/docs/exporting-research-data) lists the endpoint and the fields, and the [structured questions guide](/docs/structured-questions-guide) covers designing studies that export cleanly.\n\nThe CSV downloads immediately rather than arriving as a time-limited email link, and there is no proprietary archive format in the path. You can verify all of this on day one of a trial, which is the only honest way to evaluate a portability claim.\n\nThere is also no moderator in the loop to schedule, which is the other half of why UserTesting migrations get postponed: the incumbent is painful but the replacement sounds like work. The Condens State of AI in User Research Analysis survey of 332 practitioners found 71 percent agreeing that AI lets them analyse data significantly faster, alongside 71 percent agreeing that validating AI output still takes significant time and only 21 percent trusting it with minimal review - so the right standard for a replacement is not \"AI touched it\" but whether the evidence behind every theme is one click away. In Koji every theme traces to the quotes and the participants that produced it, with a one-to-five quality score on each interview so you can filter out thin sessions before they reach the report. For a direct feature comparison, see [Koji vs UserTesting](/docs/koji-vs-usertesting).\n\n**Stop paying for research you cannot move.** Run a Koji study this week, export it on day one, and check that what comes out is a single clean record before you renew anything.\n\n## Frequently asked questions\n\n### Can I export my UserTesting data?\n\nYes. Metrics, written responses, and rating-scale answers export to an Excel workbook, with an option to put all sessions into one file; videos download as `.mp4` per selected session; and UserTesting generates transcripts that can be imported elsewhere as text. The limitation is not access, it is that these arrive as separate artifacts with no field linking a video file to its metrics row.\n\n### Does a UserTesting export keep videos and data together?\n\nNo. Videos and metrics are separate downloads and have to be rejoined manually. Decide on a join key - username plus session plus test usually works - and apply it to video filenames as you download, rather than trying to reconstruct the mapping afterwards from timestamps.\n\n### How much cleanup does the Excel export need?\n\nMore than you would expect. Condens, which documents importing UserTesting data into its own platform, advises moving the Metrics sheet to the first position or deleting the other sheets, removing the first few rows above the one containing the Username, and removing columns containing calculations such as mean and median. Do that cleanup once, document the steps, then repeat them mechanically.\n\n### Should I migrate all my historical UserTesting sessions?\n\nUsually not. Sort studies into live decisions, reference material, and dormant, and only fully reassemble the first group. Video is the expensive part to move and the least likely to be rewatched. A spreadsheet-only export is adequate for reference material, and dormant studies rarely justify any work at all.\n\n### How long does migrating off UserTesting take?\n\nThe downloads are fast; the reassembly is not. For a few hundred sessions, plan a focused week or two with the cleanup templated, inside a longer wind-down that includes running a real study in the replacement platform in parallel. Do the extraction before giving notice, while your access and your support relationship are both intact.\n\n### Can I automate the export instead of clicking through it?\n\nThat depends on your edition and contract. UserTesting documents API access for retrieving results, including video downloads, but availability varies by plan - so ask your account team directly whether your agreement covers it. If it does, automation is the single biggest time saver in the whole migration; if it does not, scope the manual work honestly before you set a cutover date.\n","category":"Comparisons","lastModified":"2026-10-06T07:51:05.905789+00:00","metaTitle":"How to Migrate Off UserTesting (2026): Export, Videos & Reassembly","metaDescription":"UserTesting exports metrics, videos and transcripts separately, with nothing linking them. The cleanup steps and a 7-step extraction sequence.","keywords":["migrate off usertesting","usertesting data export","export usertesting videos","leave usertesting","usertesting migration","usertesting alternatives","usertesting excel export"],"aiSummary":"UserTesting exports are accessible and in ordinary formats but arrive as disconnected artifacts: an Excel workbook of metrics, separately downloaded .mp4 videos, and transcripts as text, with no field linking a video to its metrics row. Condens, which documents importing UserTesting data, instructs users to delete unused sheets, remove header rows above the Username row, and remove calculated mean and median columns before import. Koji exports participant metadata, session data, summaries and quality scores in one CSV and full structured analysis as JSON via API.","aiKeywords":["migrate off usertesting","usertesting data export","export usertesting videos","leave usertesting","usertesting migration","usertesting alternatives","usertesting excel export"],"aiContentType":"guide","faqItems":[{"answer":"Yes. Metrics, written responses, and rating-scale answers export to an Excel workbook, with an option to put all sessions into one file; videos download as `.mp4` per selected session; and UserTesting generates transcripts that can be imported elsewhere as text. The limitation is not access, it is that these arrive as separate artifacts with no field linking a video file to its metrics row.","question":"Can I export my UserTesting data?"},{"answer":"No. Videos and metrics are separate downloads and have to be rejoined manually. Decide on a join key - username plus session plus test usually works - and apply it to video filenames as you download, rather than trying to reconstruct the mapping afterwards from timestamps.","question":"Does a UserTesting export keep videos and data together?"},{"answer":"More than you would expect. Condens, which documents importing UserTesting data into its own platform, advises moving the Metrics sheet to the first position or deleting the other sheets, removing the first few rows above the one containing the Username, and removing columns containing calculations such as mean and median. Do that cleanup once, document the steps, then repeat them mechanically.","question":"How much cleanup does the Excel export need?"},{"answer":"Usually not. Sort studies into live decisions, reference material, and dormant, and only fully reassemble the first group. Video is the expensive part to move and the least likely to be rewatched. A spreadsheet-only export is adequate for reference material, and dormant studies rarely justify any work at all.","question":"Should I migrate all my historical UserTesting sessions?"},{"answer":"The downloads are fast; the reassembly is not. For a few hundred sessions, plan a focused week or two with the cleanup templated, inside a longer wind-down that includes running a real study in the replacement platform in parallel. Do the extraction before giving notice, while your access and your support relationship are both intact.","question":"How long does migrating off UserTesting take?"},{"answer":"That depends on your edition and contract. UserTesting documents API access for retrieving results, including video downloads, but availability varies by plan - so ask your account team directly whether your agreement covers it. If it does, automation is the single biggest time saver in the whole migration; if it does not, scope the manual work honestly before you set a cutover date.","question":"Can I automate the export instead of clicking through it?"}],"relatedTopics":["usertesting","data export","vendor migration","usability testing","data portability","vendor lock-in"]}],"pagination":{"total":1,"returned":1,"offset":0}}