What drives value: the factors that move a data license up or down the range

Owners ask what their data is worth and expect a rate per gigabyte. There is not one. Two companies with the same headcount and the same systems can get offers that differ by a factor of five, because buyers are paying for signal, and signal depends on things a file size does not capture. Published buyer programs quote ranges from the low six figures into the millions for a single company's records, and the spread inside that range is explained almost entirely by the six factors below.

Every figure here is indicative. Actual value depends on buyer demand on the day, which is why we benchmark a manifest against live demand rather than against a formula.

The single biggest driver. Buyers want to see a process repeated enough times that a model can learn the pattern and its exceptions. Three years is the current threshold; five to ten years in the same systems is where the number climbs. Continuity matters more than length. A migration in year four that dropped the internal notes leaves you with four good years and four thin ones, and buyers price the two periods differently.

Our estimator counts up to ten years. Beyond that, buyers have usually seen enough of the pattern, and old records increasingly reflect tools and practices that have changed.

A ticket that ends in a documented fix is worth several times a ticket that ends in nothing. A call tagged with its disposition is worth more than an untagged recording. A deal that closed won or lost with a reason is worth more than one left open. Outcomes turn a record into a labelled example, and labelled examples are what buyers are short of.

Practically, this is why support queues with high resolution rates, CRMs where reps record lost reasons, and call centers with QA dispositions land toward the top of the range. If your team has been lazy about closing out records, the manifest will show it and the number will too. Starting to close records properly now improves the value of everything you produce from here on.

Volume matters, but density matters more. A 30-person company whose records are almost entirely about work outranks a 100-person company whose chat is mostly social and whose email is mostly notifications. Buyers screen samples for the share of records that carry workflow signal and discount the rest.

The record types with the highest demand right now are calls with outcomes and support tickets, followed by engineering records and team chat. Email, CRM, documents, and operational systems carry steady demand. Having several of these in the same company compounds the value, which brings us to linkage.

The most valuable thing a company can offer is a single piece of work that can be followed across systems: the email that became a ticket, the ticket that became a bug, the bug that became a release, the release that closed three more tickets. Each record alone is a snapshot. Linked, they are a trajectory, and trajectories are what agentic models are trained on.

This is why the qualification threshold asks for records in three or more systems. It is also why the manifest records the identifiers that connect them: ticket numbers in chat, deal IDs in email, customer IDs across CRM and billing. You do not need the systems to be integrated. You need the references to exist.

A set with clear rights and sensible exclusions closes. A larger set with open questions stalls in diligence and often closes smaller, later, or not at all. Buyers price in the risk, so the same records are worth more when the rights checklist has been worked through before the manifest goes out. Regulated and client-owned classes excluded cleanly cost less value than you would expect, because buyers were going to discount them anyway.

The terms you grant change the price more than any property of the data.

• Non-exclusive is the default and lets you license the same records to a second buyer. Most deals are non-exclusive.
• Exclusive licenses command a premium, often a large one, because the buyer is paying for everyone else not to have the records. Exclusivity should be limited in time and field, and priced separately so you can see what you are being paid for it.
• Field of use narrows what the buyer can do. A license limited to model training is worth less to the buyer than one that also permits product development, and the price should reflect which you grant.
• Refresh rights, where the buyer pays again for new records on a schedule, turn a one-time license into recurring revenue and are worth negotiating for if your records keep accumulating.

Competing offers are what make these terms negotiable. One buyer presents its template; three buyers present terms you can play against each other.

A few things owners expect to matter that mostly do not. Your industry's prestige: buyers want workflow, and a plumbing dispatcher's exception handling can be worth as much as a consultancy's. Your company's revenue: unrelated to the data. Raw storage size: a terabyte of attachments is worth less than a gigabyte of tickets. Whether you are a technology company: the market is explicitly looking for non-technology operational records. The industry pages on this site describe where each sector tends to land and why, without pretending that sector alone decides it.

The estimator gives a rough range from headcount, years, systems, and record types in about a minute, and shows its factors. It is a heuristic calibrated to published buyer ranges, not a quote. A real range takes a manifest, which takes a day, and comes back benchmarked against live buyer demand within 24 to 48 hours. Binding terms follow sample review. If you want to know what your records are worth, that is the shortest honest path to the answer.