Our planning estimate for a data deal is two to three weeks when the data is clean and your rights are clear, and up to several months when customer contracts need written permission first. Rights decide the timeline more than anything else, so the first call starts there.
These are our own estimates for planning, not a promise. Buyer review cycles vary, and some deals stall at the offer stage because no lab needs that data right now.
The stages, in order
Every deal moves through the same stages. Each stage change is logged with a timestamp in your deal room, so you can see where things stand at any time.
| Stage | What happens | Who does the work |
|---|---|---|
| Applied | You fill in the application or calculator | You |
| NDA sent | We send a mutual NDA, usually the same day as the first call | Us |
| NDA signed | Both sides sign | You and us |
| Rights review | We read your customer agreements, DPAs and confidentiality clauses | Us, with your documents |
| Sample received | You send a sample export | You |
| Readiness report | We scrub the sample and report what can be sold | Us |
| Buyer interest | Labs review a redacted sample under NDA | Buyers |
| Offer | A buyer names terms and a price | Buyer and us |
| Licensed | You approve the buyer and sign | You |
| Paid | The buyer pays and your share goes out | Us |
| Recurring | Monthly refresh batches start | The SDK |
Week one: call, NDA and rights
The first call takes 30 minutes. We ask about rights first, then systems, volume, years of history and who signs. If the answers point to a fit, we send a mutual NDA the same day and agree a date for the sample.
Rights review starts as soon as the NDA is signed. We read the agreements that govern the data: customer contracts, data processing agreements and any confidentiality clauses. A clause saying data is "used only to provide the service" blocks a sale even after scrubbing. If you are your customers' processor, you usually need their written permission. Our guide on selling data you hold for customers explains the test.
This is the stage that turns two weeks into two months. If your contracts are clear, rights review takes days. If each customer has to agree in writing, the timeline depends on how fast they answer, and some never will. We leave out any data where we can't get a clear answer.
Week one to two: sample and readiness report
You send a sample export under NDA. Our guide to preparing an export covers formats and what to include. We run our scrubbing tool on the sample and send back a readiness report: which fields are sellable, which have to go, and the result of a re-identification test. You keep the report even if you never sell.
If you upload the sample through the portal, our server scrubs it, keeps only the scrubbed output and report, and deletes the raw file.
Week two to three: buyers review
Labs review a redacted sample before they commit, under their own NDA. We pass on an indicative price range as soon as a buyer has looked at the sample, so you know early whether the deal is worth finishing.
This step depends on demand. Labs buy in bursts tied to their training plans, so one buyer may move in days while another waits for its next cycle. Our guide on what AI labs pay for data explains how buyers think about price.
Offer, license and payment
An offer sets the uses (training, fine-tuning or evaluation), the length, any exclusivity, a price for the history and, where the buyer wants ongoing data, a price per monthly refresh. Our guide to data licensing terms explains each one.
You approve the buyer before anything is signed and can rule out any company. Once the license is signed, the buyer pays and your share is paid out. We don't promise payout timing before a deal exists, because it depends on the buyer's payment terms.
After the first payout
The first license sells your history. If the license includes a refresh term, we set up the SDK on your machine to run monthly. It pulls only records created since the last run, scrubs them locally and uploads a new batch. Each accepted batch pays the refresh price, with no further work from your team. The recurring data revenue page shows how this builds over time, and the docs cover setup.
What slows deals down
- Customer contracts that need written permission
- Data held for clients, such as a bookkeeper's client ledgers or a call center's recordings
- Free text that has to be reviewed by hand before it can go out
- Sector rules: health, finance, government and export-controlled data need their own review or are excluded
- No current buyer demand for that data type
What speeds them up
- One person who can sign and answer contract questions
- Exports ready from the systems you use, with field names intact
- Clear ownership of the data, such as your own operational records
- Several years of history in one system
Keeping you informed
Data deals can feel slow because the work is out of sight. Each seller gets a deal room page showing the stage, the open checklist and a named owner, and we send a short update every Friday, even in weeks when nothing moved.
Start the clock
The fastest first step is the valuation calculator, which takes a few minutes and gives a range for your data. The seller hub explains the process from application to recurring payouts.
Frequently asked questions
What is the fastest a data deal can close?
In our planning, two to three weeks for clean data with clear rights and a buyer who already wants that data type. We treat that as an estimate. Buyer timing is outside our control.
Why would a deal take months?
Usually because customer contracts require written permission before the data can be licensed. Asking each customer takes time, and data with no clear answer is left out.
How much of my team's time does it take?
A 30-minute call, a contract review with us, a sample export and an approval for each buyer. We prepare the data, find buyers and negotiate.
Can I stop partway through?
Yes. Nothing goes to a buyer until you approve it, and you keep the readiness report either way.