How AI data buyers calculate an offer: the estimators, compared

Most AI data buyers' online estimators turn two numbers, how many people a company has and how many years of records it keeps, into a range, and three of them share one anchor: micro1, Handshake and Nyne run the same formula, a fixed $140,000 plus $2,333 a person plus $11,667 a year of records. Their outputs still differ widely for the same company: micro1's estimator showed $379,290 to $888,088 for 100 people with 10 years of records in the US, Troveo's $800K to $1M and Grepped's $37.5K to $54K for a company of that size. An estimate is not an offer, and an offer is not a sale. Across our price board the middle of what buyers offer for one company's records is $316K per company, while the median of the few sales that closed and were made public is $28.3K per company, a figure that rests on only four prices.

This guide is for the founder, CEO or CFO of a company that has typed its head count into a buyer's estimator, or has been sent a number, and wants to know how that number was produced and what it is worth. It sets out every buyer estimator we have recorded, the inputs each one takes and what moves its result, the same company sizes run through each of them side by side, why an estimate sits so far from an offer and an offer so far from a sale, how to use an estimate without being anchored by it, and how AIData.Broker weighs the same evidence in its own valuation. The wider question of what company records are worth, with worked examples, is in how much is your company's data worth to AI companies; this guide is about the buyers' calculators and the numbers they print.

A word on who is writing. AIData.Broker keeps the price board and the rates this guide draws on, and it also acts as the agent of holders who license their records. That makes us a party in this market. Every estimator output below links to the record that holds it, with the inputs used and the day it was read, so you can rerun any of them yourself.

This guide is research from public sources, not legal, tax or investment advice.

Contents

1. At a glance: the estimators we have recorded
2. The shared anchor: one formula behind three buyers
3. Each estimator: what it asks and what moves its number
4. The same company, side by side
5. Buyers that publish no estimator
6. Why an estimate is not an offer, and an offer is not a sale
7. How to use an estimate you have run
8. How our valuation weighs the same evidence
9. Questions people ask
10. What this means for you

1. At a glance: the estimators we have recorded

A buyer's estimator is a page on its own site with a few sliders or menus and a range that appears when you move them. It is the first number most holders ever see, and for many it becomes the number they carry into every later conversation. That is exactly why it is worth knowing what is inside one.

We ran every public estimator we could find on 11 October 2026 for a set of company sizes, recorded each output with the inputs that produced it, and reran several of them on 17 October 2026. The table lists what each estimator asks, what it does not ask, and one output for the same mid-sized company, about 100 people with about 10 years of records in the US, as close as each tool allows. We call it the mid-sized company from here on.

Three things stand out before any detail. First, almost every estimator is driven by head count and years, and only two (Grepped's and Ooak's) ask about anything that describes the company beyond its size, revenue in one case and the tools it uses in the other. Second, the outputs for the same company span more than an order of magnitude, from Grepped's estimator at the bottom to Troveo's at the top: their low ends differ about twentyfold. Third, three of the outputs (micro1's, Handshake's and Pewter's) agree with each other to within about one percent, which is not a coincidence: section 2 explains why.

The rest of this guide takes these in turn. Where a buyer publishes nothing that would let us say how it calculates, we say so rather than guess.

2. The shared anchor: one formula behind three buyers

The most useful single fact about buyer estimators is that three of them are, underneath, one calculation. We read the outputs of micro1's, Handshake's and Nyne's public estimators for many company sizes and found that every one of them fits the same formula: a fixed part of $140,000, plus $2,333 for each person, plus $11,667 for each year of records, times a factor for where the company is based. We keep that finding as a record of its own, and each of the three buyers' outputs listed on our price board was compared against it.

For Nyne the anchor is not only inferred. Its estimator page carries the three amounts in its own page code, as variables named for a base fee, a per-employee amount and a per-year amount, and applies them to its own size and years brackets. For micro1 and Handshake the anchor is inferred from their outputs, and the fit is exact or close enough that the remaining gap is explained by rounding and by the way each turns the central figure into a range.

How each one turns the anchor into a range

What follows is our reconstruction, worked out from the outputs these estimators rendered on 11 October 2026. None of the three buyers publishes its formula as a formula, and none says how it turns a central figure into a low end and a high end. The reconstruction is how the comparison in our offer tool places a buyer's number against that buyer's own estimator, and its notes say for each buyer how close the fit is.

• micro1. The reconstruction is exact. The estimator takes the anchor, adds an assumed volume of records priced per gigabyte, and scales the result down for the low end and up for the high end, so its high end is a little more than twice its low end at mid sizes and much more at the largest. For a company outside the US it multiplies the whole range by a regional factor, which is why its output for the mid-sized company in Europe is $284,468 to $666,066 and its output for the rest of the world $113,787 to $266,426, against $379,290 to $888,088 for the US. micro1 calls the result an indicative estimate only.
• Handshake AI. For companies of 50 people or more, Handshake's estimator shows micro1's figure plus one percent: $383,083 to $896,969 for 100 people and 10 years in the US. Its regional outputs follow the same ratios as micro1's: $287,313 to $672,727 for Europe and $114,925 to $269,091 elsewhere. For smaller companies its low end runs lower than micro1's and stops at a floor: 20 people and 3 years in the US shows $100,000 to $296,551, where micro1 shows $167,427 to $293,614. Its sliders stop at a top setting, 200 people and 20 years, which shows $683,978 to $2,173,950+, the plus sign marking the cap.
• Nyne. Nyne applies the anchor to brackets rather than exact numbers, so every company in a bracket gets the same output: $304,500 to $657,000 for any US company of 50 to 100 people with 8 to 15 years of records. Its regional factors combine a location factor with an assumed share of English-language records, which puts Canada and Europe at $194,000 to $419,000 and the rest of the world at $64,000 to $138,000 for the same bracket.

A fourth tool that lands on the same number

Pewter's home page estimator shows $385K to $895K for 100 employees and 10 years in business at a US firm. Pewter describes it as its licensing firm's payout model, without naming that firm, and says the figure may be lower, higher or unavailable after its review of the records. The output sits within about one percent of micro1's for the same inputs. That closeness is our inference from the two numbers, not a statement by either site, but it means that four of the ten estimators in the table above are, for practical purposes, printing the same number.

What this means for a holder

If you ran micro1's estimator, then Handshake's, then Pewter's, and saw three ranges that agree, you have not collected three independent opinions of your company. You have read one formula three times. The agreement says nothing about what your records are worth; it says these firms start from the same arithmetic. It also explains a pattern holders report: an offer that arrives "in line with our estimator" may simply be the anchor with a discount applied, and the anchor never asked what your records contain.

3. Each estimator: what it asks and what moves its number

The anchor estimators move on head count, years and region alone. The others take different inputs, and those inputs are the part worth understanding, because they show what each firm thinks matters.

micro1, Handshake and Nyne: head count and years

Within the shared formula, head count is the larger lever at most sizes. In micro1's estimator, doubling the head count at 10 years of records in the US moves the range from $285,896 to $604,463 at 50 people to $379,290 to $888,088 at 100. Going from 5 years to 10 years at 50 people moves it from $239,192 to $462,638 to $285,896 to $604,463. At the large end the range opens up sharply: $1,273,005 to $4,681,050 for 500 people and 20 years, and $2,471,633 to $11,797,513 for 1,000 people and 30 years, because the assumed data volume grows with size and the high end multiplies it.

Handshake follows micro1 closely across the same sizes: $288,755 to $610,508 for 50 people and 10 years, $158,979 to $380,518 for 30 people and 5 years, and $530,547 to $1,465,864 for 150 people and 15 years. Nyne's brackets step rather than slide: $202,000 to $373,500 for 25 to 49 people with 3 to 7 years, $243,000 to $479,000 for the same size with 8 to 15 years, and $777,500 to $2,217,000 for 250 people or more with over 15 years.

None of the three asks what the records are. Two companies of the same size, one whose records are mostly marketing assets and one with ten years of linked support tickets, code reviews and finance records, get the same range. That is the single biggest limit of the anchor estimators, and the reason their outputs should be read as a starting point a buyer chose, not as a measurement of your company.

Mode (BizBuyData): staff type, years and region

Mode's estimator separates desk staff from non-desk staff and asks for years of operational records and where the team is based. We read the estimator's own published script on 17 October 2026. It holds data volume and the number of sources fixed for every company, so only head count, years and region move the result. It renders $290K to $570K for 100 desk staff and 10 years in the US, $220K to $440K for 50 desk staff and 10 years and $130K to $270K for 20 desk staff and 3 years.

For the same 50-person, 10-year company, Mode's range sits about a quarter below micro1's $285,896 to $604,463. Its regional factors match the ratios micro1 and Handshake use: $210K to $430K for a team in the EU, Canada or Australia, and $90K to $170K elsewhere, with a note on its page that companies outside the US are reviewed case by case. Mode's page calls the figure indicative, not an offer.

Troveo: employees and years, with ranges far above the buyers'

Troveo's estimator takes employees and years of operating and returns narrow ranges that run well above every buyer's estimator: $300K to $400K for 20 employees with 3 to 7 years, $800K to $1M for 100 employees with 8 to 15 years, $1.2M to $1.5M for 250, $1.6M to $2M for 500, and $3M to $3.6M for over 1,000 employees with over 15 years. Troveo is paid by a share of each licence it arranges, and its page labels these figures directional.

The difference in shape matters as much as the level. Troveo's high end is only a quarter to a third above its low end at every size, where micro1's is more than double. A narrow range reads as precise. It is not more precise; it is a different presentation choice, and it sits on top of a level that no recorded sale supports. Section 6 comes back to this.

Grepped: revenue moves it, head count barely does

Grepped's estimator for companies asks for team size, the year the company was founded and a revenue band, and it calls the result "a rough guide to start the conversation, not an offer." It produced the lowest number of any public estimator for the mid-sized company: $37.5K to $54K, founded in 2016, with revenue not given.

What moves it is revenue, not head count. The same company with revenue entered shows $71.5K to $102.5K at 10 to 50 million dollars of revenue, nearly double. Doubling the team with revenue still not given barely moves it, to $41.5K to $59.5K for 200 people. At the large end it shows $391.5K to $560K for 1,000 employees, founded in 2001, with over 200 million dollars of revenue, and at the small end $14K to $20.5K for 10 employees, founded in 2020, with under a million.

Grepped also publishes where its number comes from, in its own site settings. Its instant estimate was calibrated on one offer it says it received for a small company's records: about $50K for a team of about eight people with email, chat, code, calls and documents over three years. That is an offer, not a closed sale. Since a change its settings date 14 October and call deliberately optimistic, the same estimator shows $65K to $155K for that very profile, above the offer it was tuned to. It is the only estimator we have recorded whose operator says which real number it was tuned to.

Grepped's referral calculator: a second number from the same firm

Grepped runs a second calculator, on the page for people who introduce companies to it, and that one asks for industry, team size and years in operation. For the mid-sized company, as a software firm, it showed $330K to $1M, against $37.5K to $54K from the firm's own estimator for companies. Grepped's own site settings describe the referral range as "a deliberately generous range a referrer can show an owner".

Industry moves it a long way. The same company in healthcare shows $400K to $1.2M; it shows $800K to $2.4M for a 200-person semiconductor company, $110K to $330K for a 30-person accounting firm, and $34K to $100K for a 20-person marketing agency with 5 years. If someone who introduced you to a buyer showed you a number, ask which calculator produced it. The same firm prints two numbers for the same company: for one small firm, the top of the referral range is 9.1 times the bottom of the estimate for companies.

Ooak: the tools you use, and whether you still operate

Ooak's estimator asks for people, years, whether the company is operating, winding down or closed, and how many of a list of work tools it uses. It is the only public estimator we have recorded in which something describing the records themselves moves the number a lot. An operating company of 51 to 100 people with 4 years of records shows $16k to $22k with three tools and $85k to $121k with six, more than five times as much for the same company.

Whether the company still operates moves it too. For 11 to 50 people, 4 years and six tools, operating shows $51k to $71k and closed $33k to $47k. Years move it less: the larger company with six tools goes from $85k to $121k at 4 years to $99k to $139k at 7 years. Ooak also states the span of the sales it says it has closed, from $20K to several hundred thousand dollars, which is one of the few places a firm puts its estimator beside its own record.

Replay: size bands, upfront part only

Replay publishes no calculator, only bands by head count: $10K to $100K for 20 to 50 employees, $100K to $1M for 50 to 250, and $1,000,000+ for over 250. Its page says years in operation, industry and the systems a company uses move the final price, and that it pays an upfront amount plus a share of every licence, so these bands are the upfront part. A band that runs tenfold from bottom to top tells a holder almost nothing about where in it their company lands.

4. The same company, side by side

The clearest way to see how far apart the estimators are is to run the same company through each of them. The tables below use only recorded outputs. Where a tool takes brackets rather than exact numbers, the closest bracket is named; where we have no recorded output for that size, the cell says so rather than filling it in.

The mid-sized company, in the US

Each row's inputs link to the record that holds them, with the day the tool was run.

The same company is worth, according to these tools, somewhere between the low tens of thousands and a million dollars. The top of the table is a firm paid on a share of each licence and a calculator built to be shown to companies by people who introduce them; the middle is the shared anchor, four times over; the bottom is the two tools that ask about something beyond size. Our valuation for a company like this is judged against the market's records, which section 8 describes.

50 people, 10 years, in the US

At 50 people the anchor estimators still agree with each other and Mode still sits about a quarter below them. Grepped's estimator for companies is roughly a tenth of the anchor. Replay's band covers almost all of them and so separates none.

A small company: 20 to 30 people, 3 to 5 years, in the US

For a small company the spread is starker, because the anchor's fixed part dominates. A 20-person company gets $140,000 before a single person or year is counted, which is why the anchor estimators print six-figure ranges for companies whose records the market has priced far lower. The one real number in this table, the offer Grepped says it received for an eight-person company, sits below every anchor output and close to Replay's band and Grepped's own estimator.

The same company in a different country

Every estimator that asks for region discounts a company outside the US by a fixed factor, and the factors agree across four firms. When we fit the market's estimator outputs together, a company in Europe, the UK or Canada comes out at 0.71 times the price of the same company in the US, and a company elsewhere at 0.27 times. The estimators do not say why; the likely reasons are language, the law that governs the records, and how easily a buyer can verify the rights, but no buyer publishes its reasoning.

5. Buyers that publish no estimator

Most buyers of company records publish no calculator at all, and several publish only a headline figure. On those, there is nothing to reconstruct, and this guide does not guess.

• Scale AI shows an illustrative range per partnership, with no calculator behind it.
• Turing (Lazarus) states a single payout ceiling on its page and no method; for code it gives a typical range per repository that depends on quality and complexity.
• Telegraph Lab states a span for what it pays per licence and has no estimator.
• Defined.ai reports an average revenue figure of its own for its partners, its own claim, not a calculator.
• HUD DataVendor says it pays up to a stated amount per codebase, with no calculator.
• Polyshares asks at intake for how much data each system holds, how long it has been used and how many seats it has, not for files, and has a person read an anonymised sample before it prices the record. Its own page lists a typical deal range per licence and sample amounts for single companies, such as one marketing agency's record.
• SimpleClosure AssetHub offers a free valuation, not a calculator, and says it is on track to a stated total across all its companies by the end of 2026, its own claim.
• Collinear AI, Mercor, Labelbox, Appen, Centific and the rest of our buyer board record no estimator outputs.

Where a buyer publishes no estimator, the first number you see from it is an offer, and the method behind that offer is not public. The buyer board records what each buyer does publish, and the rates page sets out the market's figures per kind of record.

6. Why an estimate is not an offer, and an offer is not a sale

An estimate, an offer and a sale are three different things, and the gaps between them are large. Knowing why they differ is the most useful thing a holder can carry into a negotiation.

An estimate is marketing

Every estimator in this guide is published by a firm that wants you to start a conversation with it. The firm chooses the inputs, the formula and the presentation. An estimator that asks only for head count and years cannot know what your records contain, whether you have the rights to license them, how much a buyer would have to clean, or whether the buyer already holds similar records. Each one says so in its own way: micro1 calls its output indicative, Grepped calls its "a rough guide to start the conversation, not an offer," Mode calls its indicative, Troveo directional, and Pewter says the figure may be lower, higher or unavailable after review.

The estimators also have reasons to print high. A firm that is paid by a share of each licence earns more when holders arrive with high expectations and sign; a firm that buys records outright has a reason to set expectations it can meet. That is visible in the tables above: the firm paid by a share prints the highest numbers, and the firm that published which real offer it calibrated to prints among the lowest. Neither pattern proves intent, but both tell you to treat a high estimate as a reason to ask questions, not as a value.

An offer is narrower than the estimate that preceded it

When a buyer sends an offer, it has usually seen something the estimator did not: a description of your systems, a sample, a list of the kinds of records you hold. An offer can land above the estimate when the records are unusually useful, but far more often it lands lower, because the review finds gaps (short history in some systems, records the company cannot license, personal information that must be removed at cost) that the estimator assumed away. The offer also carries terms the estimator never mentions: whether the licence is exclusive, how long it runs, when payment starts and whether it can be clawed back. Our guide on when AI data buyers pay sets out the payment terms buyer by buyer.

Across our price board, the middle of what buyers offer for one company's records is $316K per company, and the middle half of public offers sits at $174K to $693K per company. Those are buyers' own published offers and price guidance, not estimator outputs. The middle offer sits below even the low end of the anchor estimators' range for the mid-sized company, $379,290 at micro1.

A sale is the only price that happened

The few sales of company records that closed and were made public sit lower again: their median is $28.3K per company, taken over only four prices, so read it as a direction rather than a rate. One of them shows the scale: a small closed startup netted about $25,000 for its code, as reported. There are reasons a sale sits below an offer. A sale price is set after review, after negotiation and after the buyer has weighed alternatives; it reflects the terms actually signed, which are often narrower than the offer that opened the talks; and it is the price at which a buyer was willing to commit money, not the price at which it was willing to start talking. Public sales are also few, skewed toward closed and closing companies whose records were sold in insolvency or wind-down processes, and so they understate what an operating company with clean rights can obtain.

That is the honest shape of the market: estimates on top, offers below them, a handful of sales below the offers, and wide variation at every level. A holder who anchors on the top line and is then offered the middle feels cheated; a holder who knows the shape from the start can judge an offer for what it is.

7. How to use an estimate you have run

An estimate is still useful. It tells you which formula a buyer starts from, what it thinks matters, and what range it is prepared to talk about. Used well, it is a question to ask rather than a number to believe.

Ask what it assumes

For any estimate you have run, or any number someone has sent you, find out what produced it. Was it the shared anchor, read through micro1, Handshake, Nyne or Pewter? Was it Troveo's, built by a firm paid by a share of each licence? Was it Grepped's referral calculator, which its own settings call deliberately generous, or Grepped's estimator for companies, roughly a tenth of it? Was revenue entered? Were the tools counted? Each of these changes the number by a large factor, and none of them describes your records.

Then ask the buyer directly what the estimate assumes about your records: how many years in which systems, which kinds of records, what volume, and whether the figure assumes an exclusive licence. If the buyer cannot say, the estimate tells you nothing about your company.

Ask for the offer in writing

An estimate binds no one. An offer becomes something you can weigh when it is written down with its terms: the amount, what records it covers, whether the licence is exclusive, how long it runs, what uses it allows, when payment starts, what the buyer can reject and on what grounds, and whether money paid can be asked back. Ask for exactly that. A verbal number "in line with the estimator" is the anchor with a discount and no terms attached.

Compare terms, not just amounts

Two offers at the same amount can be worth very different things. An exclusive licence fetches 2.5 to 5 times a non-exclusive one for the same records, so an exclusive offer at the same amount as a non-exclusive one is, by that measure, the lower offer. An evaluation-only licence fetches 0.2 to 0.5 times a full commercial one. An offer whose payment starts only when a lab pays the buyer leaves your money in the hands of a party you never sign with. When you hold an offer, our offer tool places it against our valuation range and against that buyer's own estimator, shows what you would net under our fee, lists the clauses in that buyer's terms to read first and gives the questions to send back.

Run more than one, and know which ones are the same

If you run estimators to get a sense of the range, run at least one from each family: one anchor tool (they agree with each other), Mode, one of the two that ask about your records (Ooak's tools count, Grepped's revenue), and Troveo if you want to see the top. Expect them to disagree by a factor of ten or more, and do not average them: three anchor outputs averaged with one Grepped output is the anchor counted three times.

Get a valuation that reads the records

The estimators place your company by size. A valuation that places it by what it holds starts from the market's records and moves each comparable to describe your company. Ours is at /value: it gives an estimate first, with no upload of records, and asks the questions buyers price on. How it works sets out what happens if you choose to go further.

8. How our valuation weighs the same evidence

Our valuation of company records starts where the estimators stop. It reads every record of what a whole company's records sold for, were offered or were estimated at, 109 of them as it stands, and weighs each by what kind of evidence it is.

A sale counts fully, an offer half, a broker's figure a quarter

The weights are a principle of the model, not a reading: 1, 0.5 and 0.25. A sale that happened counts fully. A buyer's own published offer counts half. A figure from a broker, a reseller or a vendor's price guide counts a quarter. Each comparable also counts less with age and with how far it has to be moved to describe the company being valued. An estimator output is a buyer's or broker's marketing figure, so it never counts as a sale.

This is the opposite of how a holder who has run five estimators tends to weigh them. Five estimator outputs feel like five pieces of evidence; in our model, five estimates from one formula are one source of a slope, and a single closed sale outweighs any of them.

What the estimators are good for: the slopes

The estimators are not useless to a valuation. They are the best public evidence of how buyers move a price as a company changes, because one estimator run at many sizes shows how that firm prices size, years, region and the rest, holding everything else equal. We fit those movements across every estimator's outputs, each estimator counting once and with its own level, so that only the variation inside one estimator counts. That gives:

• Head count: each doubling of a company's people puts the price at 1.47 times what it was.
• Years of records: each doubling of the years puts it at 1.31 times what it was.
• Region: a company in Europe, the UK or Canada at 0.71 times the US price, and elsewhere at 0.27 times.
• Closed or closing: the records of a closed or closing company at 0.65 times those of the same company operating.
• Kinds of records: each doubling of the share of the kinds buyers name puts the price at 5.00 times what it was.

The last two slopes each rest, so far, on a single pair of Ooak's outputs (the tools counted in one, operating or closed in the other), the only estimator that varies either. They are the least settled numbers in the model, and the first we expect to move as more outputs are recorded.

So the estimators set the slopes and the market's sales and offers set the level. That separation is the point: the estimators' level is marketing, their shape is information.

The floor we judge a base case against

For small companies, the valuation's base case is judged against $26,000, the weighted average of what a small company's records sold for or were priced at across the records that price one small company, computed with the same weights for a sale, an offer and a broker's figure ( 1, 0.5 and 0.25). When only the optimistic case reaches it, the valuation says that measuring the records decides, rather than promising a number. The full method is set out with every figure on the rates page and in how much is your company's data worth.

Where buyers publish nothing

No buyer publishes how it turns a review of your records into an offer. The estimators show the arithmetic of the first number; the step from there to a written offer is, on every buyer's site we have read, not public. Our model cannot read that step either. It reads its outcome where an offer or a sale has been made public, and weighs it accordingly.

Questions people ask

How do AI data buyers calculate an offer?

Most start from a size formula. micro1, Handshake and Nyne share one: $140,000 plus $2,333 a person plus $11,667 a year of records, times a regional factor, which is our reconstruction from their outputs rather than a published formula. Others use their own inputs: Grepped weighs revenue, Ooak the tools a company uses and whether it still operates, Mode desk staff, years and region. The step from that first number to a written offer follows a review of your records, and no buyer publishes how it is done.

Why did two estimators give me such different numbers for the same company?

Because they are built by different firms for different purposes. For the same company of about that size and age, Troveo's showed $800K to $1M and Grepped's estimator for companies $37.5K to $54K. Troveo is paid by a share of each licence; Grepped tuned its estimate to an offer it says it received. Neither asked what your records contain.

Is the estimator number what I will be offered?

Rarely. Every estimator in this guide is labelled indicative, directional or a starting point by its own operator. Across our price board the middle of buyers' offers for one company's records is $316K per company, below the low end the anchor estimators print for a US company of about 100 people and 10 years, and the median of the few public sales is $28.3K per company.

Why do micro1, Handshake and Pewter give almost the same number?

Because micro1 and Handshake run the same anchor formula, and Pewter's output for the same inputs lands within about one percent of micro1's. Agreement among them is one formula read three times, not three opinions.

Does the estimator know what my records contain?

Almost never. Only Ooak's asks about the tools a company uses and Grepped's about revenue. The rest price head count, years and region. A company with ten years of linked tickets, code and finance records gets the same range as one with scattered documents.

Should I trust a higher estimate?

Treat it as a reason to ask questions. The highest outputs we recorded come from a firm paid on a share of each licence and from a calculator its own operator calls deliberately generous. The lowest come from the one estimator whose operator published the real offer it was calibrated on.

What does an estimate assume about exclusivity?

None of the estimators says. An exclusive licence fetches 2.5 to 5 times a non-exclusive one, so ask whether a number assumes exclusivity before comparing it with anything.

How does AIData.Broker value records differently?

It weighs every public record by kind, a sale, an offer and a broker's figure counting 1, 0.5 and 0.25, uses the estimators only for how prices move with size, years, region and kinds of records, and judges a small company's base case against $26,000. You can run it at /value without uploading records.

I already have an offer. How do I know if it is fair?

Put it into the offer tool. It shows where the amount sits against our range and against that buyer's own estimator, what you would net under our fee, the clauses to read first and the questions to send back before signing.

What this means for you

• Know which formula you ran. If your number came from micro1, Handshake, Nyne or Pewter, it is the shared anchor of $140,000 plus $2,333 a person plus $11,667 a year, and it did not look at your records.
• Expect a wide spread. For the same mid-sized company the recorded outputs run from $37.5K to $54K to $800K to $1M. Do not average them.
• Read the shape of the market. Estimates sit on top, offers below them at a middle of $316K per company, and the few public sales below that at a median of $28.3K per company.
• Ask what every number assumes, and ask for any offer in writing with its terms: amount, scope, exclusivity, duration, uses, payment trigger, acceptance and clawback.
• Compare offers on terms, not amounts. An exclusive licence fetches 2.5 to 5 times a non-exclusive one, so the same amount is not the same offer.
• Weigh evidence by what it is. A sale counts more than an offer, an offer more than a broker's figure, and an estimate is marketing. That is how our valuation reads it.
• Start from your records, not your head count. Get an estimate that asks what you hold at /value, compare an offer you have at /tools/offer-check, see the market's figures at /rates and /prices, and read how it works before you decide anything.