Skip to content

Deal economics

AI data company bookings, GMV and revenue: three clocks, three charts

By SourceX Editorial · Reviewed by Noah Loul ·

10 sections

Short answer

AI data company bookings, GMV and revenue measure three events: a signed commitment, gross buyer spend through the platform, and revenue earned when the license transfers, counted net when the company acts as agent. Reddit signed $203.0 million of data licensing contract value in January 2024 but expected at least $66.4 million of it as 2024 revenue. Chart each separately, reconcile them quarterly, and keep contingent value off all three.

Value drivers

AI data company bookings

This page covers AI data company bookings, GMV and revenue: three clocks, three charts.Illustrative

Key takeaways

  • Bookings date the signature, GMV dates the transaction and revenue dates the transfer of control; one chart cannot show three clocks.
  • A firm-backlog roll-forward ties them together: opening backlog plus bookings, minus cancellations and deliveries, equals closing backlog.
  • Unexercised options, untested deliveries, unearned usage fees and cancellable terms are contingent value and belong on none of the earned charts.
  • GMV becomes revenue only through the principal-or-agent judgment; as agent, a company reports only the share it keeps.
  • Ask for written definitions, the reconciliation and counterparty concentration before any growth rate.

Why do bookings, GMV and revenue need separate charts at an AI data company?

Because each metric records a different event on a different date, and a single chart forces all three onto one clock. Bookings record a commitment on the day a contract is signed. Gross merchandise value, or GMV, records the gross amount buyers spend through a platform when a transaction happens. Revenue records what the company has earned once control of the licensed data passes to the buyer, and a company acting as an agent counts only the share it keeps.

In data licensing those dates can sit quarters apart. A three-year license is booked when it is signed, invoiced as each dataset is delivered, and recognized as each delivery transfers. A delivery that fails the buyer's acceptance test may never become GMV or revenue at all, although it was booked. Put the three on one line, or stack them in one bar, and committed value starts to look earned.

The decision this page supports is practical: what an investor should ask a data marketplace or intermediary to chart, how those charts must reconcile, and which amounts belong on none of them. The worked numbers are illustrative index points, not figures from any company, SourceX included.

Three clocks: what each chart counts and on which date

Give each chart one unit, one date stamp and a written exclusion list. Two definitions already exist in public sources. Uber defined Gross Bookings in a 2020 SEC filing as the total dollar value of its rides, deliveries and freight shipments, 'in each case without any adjustment for consumer discounts and refunds, Driver and merchant earnings, and Driver incentives.' That is a gross activity measure, reported apart from revenue.

Topic 606 supplies the backlog definition. Paragraph 606-10-50-13, issued in ASU 2014-09, requires disclosure of 'the aggregate amount of the transaction price allocated to the performance obligations that are unsatisfied (or partially unsatisfied) as of the end of the reporting period,' usually called remaining performance obligations. Firm backlog below follows the same idea: committed transaction price not yet delivered.

Cash is the fifth row because a slow-paying buyer can keep recognized revenue out of the bank long after the quarter closes; it complements the other four charts rather than replacing any of them. The specification, ready to hand to a finance team:

Three clocks: what each chart counts and on which date
ChartUnit countedDate stampIncludesExcludes
BookingsContract value signed in the period, gross of the supplier shareSignature dateFirm, non-cancellable consideration in executed agreementsUnexercised options, renewals, letters of intent, cancellable portions
Firm backlogCommitted transaction price not yet delivered or acceptedPeriod endSigned amounts awaiting delivery or acceptanceUsage fees not yet earned, terminated or rejected deals
GMVGross buyer spend that passed through the platformDelivery or invoice dateBoth the supplier share and the platform share of each invoiceCredits, rejected deliveries, deals never invoiced
RevenueAmount earned under Topic 606Transfer of control to the buyerThe gross amount as principal, the retained amount as agentBillings in advance, variable amounts held back by the constraint
CashMoney received from buyersReceipt dateCollections against invoicesOpen receivables and amounts still owed to suppliers

What Reddit's data licensing filings show about contract value and revenue

Reddit's filings are the plainest public example of signed value outrunning same-year revenue. Its IPO registration statement said that in January 2024 it entered into data licensing arrangements 'with an aggregate contract value of $203.0 million and terms ranging from two to three years,' and that it expected 'a minimum of $66.4 million of revenue to be recognized during the year ending December 31, 2024 and the remaining thereafter.'

Drawn as charts, that is a bookings bar of $203.0 million beside a first-year revenue bar of at least $66.4 million, roughly a third of it. Nothing was wrong with the deals; their terms spread recognition across years. Presenting the $203.0 million as annual revenue, or stacking it on top of revenue, would count future years' performance as if it had already happened.

Concentration is the second lesson. Reddit's annual report for fiscal 2025 states that 'substantially all of the contract value associated with our licensing revenue is derived from two of our partners,' and its risk factors discuss the chance that those contracts are not renewed on similar terms. A bookings chart built from two counterparties describes two relationships more than it describes a market, so the counterparty count belongs beside every bookings figure.

Announced demand is a separate category that belongs on no chart. In October 2026 micro1 published 'micro1 announces commitment to spend $1B on enterprise data acquisition over the next 12 months,' describing a plan to acquire and license enterprise operational data for reinforcement learning environments. That is one buyer's stated plan. It becomes a booking for a supplier only when a contract is signed, and it is not evidence that any particular company's records will be bought.

The bookings-to-revenue reconciliation, line by line

A firm-backlog roll-forward is the bridge between the charts: opening backlog, plus new bookings, minus cancellations and failed acceptances, minus deliveries, equals closing backlog. Deliveries then become GMV, GMV becomes revenue through the principal-or-agent judgment, and revenue becomes cash on collection. Every line should tie to a document an investor can sample.

The figures describe one illustrative year for a hypothetical intermediary, in index points where firm signed value equals 100. They deliberately include a rejected candidate: a deal that was signed and delivered, then cancelled after failing acceptance. A reconciliation that drops failures makes conversion look better than it was, which is exactly the error separate charts exist to prevent.

The illustration assumes the intermediary acts as agent and passes 60 percent of each invoice to the supplying company. That share is chosen for arithmetic only; it is not a market rate or a SourceX term. The arithmetic, not the values, is the part to copy:

The bookings-to-revenue reconciliation, line by line
LineIllustrative pointsChart it belongs onDocument that proves it
1. Value announced at signing, including an optional refresh120None: a memo item onlyPress release or board deck wording
2. Less the optional refresh the buyer has not exercised-20None until exercisedOption clause and exercise notice
3. Gross bookings: firm signed value100BookingsExecuted agreements with signature dates
4. Less a signed deal cancelled after failing acceptance-30Reverses bookings; never GMV or revenueAcceptance record and termination notice
5. Deliveries invoiced in the year (GMV)50GMVInvoices matched to delivery receipts
6. Closing firm backlog (100 - 30 - 50)20Firm backlogBacklog schedule tied to each contract
7. Less the supplier share of GMV, agent presentation-30Excluded from revenue as agentSupplier agreements and settlement ledger
8. Net revenue recognized20RevenueRevenue schedule by performance obligation
9. Buyer invoices uncollected at year end30 of the 50Receivables, not cashAccounts receivable aging

One illustrative year, drawn three ways

Plotted by quarter, the same three deals produce three charts with different shapes, which is why they cannot share an axis. Deal A is signed in the first quarter for 60 points and delivers tranches of 20 in the second and fourth quarters, with its last tranche due next year. Deal B is signed in the second quarter for 30, delivered in the third and cancelled in the fourth after failing acceptance; it was to be invoiced on acceptance, so it never reaches GMV. Deal C is signed and delivered in the fourth quarter for 10.

Bookings peak in the first quarter, GMV in the fourth, and net revenue never exceeds 12 points in any quarter. A deck showing only bookings would report 100 for the year, or 120 if it counted the unexercised refresh; the revenue chart shows 20. None of that gap is an accounting error. It is timing, one failed delivery and the agent presentation, and each chart is truthful on its own clock.

Quarterly values for the illustrative portfolio, in index points:

One illustrative year, drawn three ways
QuarterGross bookingsGMVNet revenueFirm backlog at quarter end
Q1600060
Q23020870
Q300070
Q410, with 30 cancelled301220
Year100 gross, 70 after cancellation502020
Three side-by-side bar panels for one illustrative year: gross bookings of 60, 30, 0 and 10 points by quarter with 30 cancelled in the fourth, GMV of 0, 20, 0 and 30, and net revenue of 0, 8, 0 and 12.
The same three deals on three clocks. Illustrative index points, not company data.

Where contingent value hides: options, acceptance tests, usage fees and cancellable terms

Contingent value is any amount that still depends on an event outside the company's control, and it stays off every earned chart until that event happens. Topic 606 names the risk: consideration highly exposed to factors beyond the entity's control, such as the actions of third parties, is one of the factors that make a later revenue reversal more likely, so variable amounts enter the transaction price only to the extent a significant reversal is not expected. A buyer's acceptance decision is that kind of third-party action.

At a glance

Where contingent value hides: options, acceptance tests, usage fees and cancellable terms

Rejected candidates belong in the reconciliation for the same reason. A deal that was signed and then failed acceptance should appear as a booking followed by a dated reversal, not vanish from both periods. Ask how many such reversals occurred, because a company with none in its history either has an unusually smooth delivery record or does not record them.

Four kinds of contingent value recur in data licensing:

  • Unexercised options: refreshes, extra volumes or extensions the buyer may order. Count them as bookings only when exercised.
  • Acceptance-contingent amounts: payments billed or earned only after a delivery passes the buyer's test. Until then they are backlog at most, and a failed test removes them.
  • Usage-based fees: per-query, per-model or revenue-share payments. ASU 2016-20 lets a company leave sales- or usage-based royalties on licensed intellectual property out of its remaining performance obligation disclosure, so a backlog figure can omit them entirely; ask where they appear instead.
  • Cancellable terms: when a buyer can end a license for convenience, treat value beyond the notice period as contingent rather than firm backlog.

The strongest objection: are bookings not the leading indicator investors need?

Bookings do lead revenue and a company should report them; the objection fails only when bookings stand in for revenue or arrive without a reconciliation. A rising bookings chart shows signatures coming in. It does not show that deliveries will pass acceptance, that a margin computed at signing will survive rework, or that buyers will pay on time.

Three things erode a margin computed on bookings before it reaches cash. Rework: a delivery that fails a first acceptance round consumes preparation hours again, and the booked figure never priced them. Overhead: sourcing, rights review and candidates that never ship are paid for by the deals that do. Collection delay: in the illustrative year, 30 of the 50 points invoiced were still unpaid at year end. Each erosion is visible only when the charts are kept apart.

Public-company rules draw the same line. The SEC's 2020 guidance on key performance indicators expects a metric to come with a clear definition and an explanation of how it is calculated, and its staff treats non-GAAP adjustments 'that have the effect of changing the recognition and measurement principles required to be applied in accordance with GAAP' as individually tailored and potentially misleading. A private company does not file under those rules, but a revenue line that is really bookings fails the same test of definition.

Signals that bookings are standing in for revenue:

  • Revenue described as contracted or signed, with no recognition schedule behind it.
  • One chart labeled revenue whose largest bar falls in a quarter with few deliveries.
  • Backlog that never declines, which suggests slipping deliveries or unrecorded cancellations.
  • GMV growing while net revenue stays flat, with no explanation of the supplier share.

What could make this framework, or a data company's charts, mislead

Separate charts remove one error and leave others in place. These are the ways the framework, or the charts a company builds with it, could still mislead an investor:

At a glance

What could make this framework, or a data company's charts, mislead

  • Definition drift: private companies have no required definition of bookings, GMV or backlog, so two companies' charts can differ by policy alone. Ask for the written definition and the date of any change.
  • The principal-or-agent judgment: under Topic 606 a company is a principal if it 'controls the specified good or service before that good or service is transferred to a customer.' The same deals can be reported gross or net, 50 points or 20 in the illustrative year, at identical economics.
  • Concentration: when a few counterparties provide most signed value, as Reddit discloses for its licensing revenue, every chart moves with one renewal decision.
  • Announced budgets: stated plans such as micro1's can change or go to other suppliers, and none of that money is backlog for a company that has not signed.
  • Long recognition tails: multi-year licenses with refresh obligations push revenue into later years, so a strong bookings year can be followed by a revenue year that looks flat with no loss of demand.
  • This page's own model: the illustrative year is invented to show mechanics. Real portfolios have more deals, partial acceptances, price changes and credits.

Questions to ask any AI data company about its charts, SourceX included

Ask for definitions and the reconciliation before any growth rate, because a growth rate on an undefined metric cannot be compared with anything. The questions below apply to every company selling or brokering AI data.

For SourceX, the public facts describe process, not performance. It works with companies licensing selected business records to AI developers, taking each from sourcing and qualification and a rights review through contracting, data preparation and delivery to payment, and the supplying company's signature on an approved scope and terms comes before any data changes hands. None of that sizes any of the three charts: SourceX's company and commercial metrics go to investors directly rather than onto this page, and materials for its Rule 506(c) raise from accredited investors can be requested through its investors page.

Questions for the finance team:

Checklist · 7 items

  1. Which date stamps each metric, and has any definition changed within the periods shown?
  2. Can you provide a quarterly firm-backlog roll-forward that ties bookings to GMV and closing backlog?
  3. Which deals were booked and later cancelled or failed acceptance, and where do they appear?
  4. For each counterparty type, are you principal or agent, and what evidence supports that judgment?
  5. How much signed value depends on options, usage fees or acceptance, and on which chart is it shown?
  6. What share of signed value comes from the two largest counterparties?
  7. What is the gap between revenue recognized and cash collected, quarter by quarter?

0 of 7 ticked. Tick each item as you go.

Methods note: sources, date and what the reconciliation leaves out

The chart specification and roll-forward were built on 2026-10-26 from public sources: Topic 606 as issued in ASU 2014-09 and amended by ASU 2016-08 and ASU 2016-20, read on FASB's site; Reddit's IPO registration statement and fiscal 2025 annual report; a 2020 Uber filing defining Gross Bookings; SEC guidance on key performance indicators and non-GAAP measures; and micro1's October 2026 announcement. Quoted figures carry their source and date in the sentence.

Every number in the reconciliation and the quarterly table is illustrative: three invented deals in index points, an assumed agent presentation that passes 60 percent of each invoice to the supplying company, and invoicing at delivery or acceptance as described. Nothing reflects SourceX's results or any other company's.

Not measured: taxes, refunds and credits, foreign exchange, several performance obligations inside one license, contract modifications, revenue recognized over time rather than at delivery, and the cost lines between revenue and contribution margin. Accounting conclusions depend on each contract's terms; this page is general information, not accounting advice.

Frequently asked questions

What is the difference between bookings and revenue for a data licensing company?

Bookings are the value of contracts signed in a period, dated at signature. Revenue is what the company has earned under Topic 606, dated when control of the licensed data passes to the buyer. A multi-year license is booked once and recognized as it is delivered, so the two can differ by years.

Is GMV the same as revenue for an AI data marketplace?

No. GMV is gross buyer spend through the platform, including the share owed to the companies that supplied the data. If the marketplace acts as agent under Topic 606, its revenue is the amount it keeps; as principal it reports the gross amount, and the supplier share becomes a cost.

What are remaining performance obligations, and why do they matter here?

They are the transaction price allocated to obligations not yet satisfied at period end, disclosed under Topic 606. They are the closest standardized measure of firm backlog and connect bookings to future revenue, although exemptions let companies leave out short contracts and usage-based royalties.

Should a buyer's announced data budget count toward a supplier's pipeline?

Not as bookings or backlog. An announcement such as micro1's stated $1 billion plan describes intent, and only a signed contract with defined consideration creates a booking. Track announcements as market context, kept apart from any company's charts.

Which chart matters most when evaluating an AI data company?

None on its own. Bookings show demand being signed, GMV shows volume delivered, revenue shows value earned and cash shows what arrived. The reconciliation between them, with cancellations and failed acceptances shown, tells an investor whether growth is converting.

Sources

  • Uber defined Gross Bookings in a 2020 SEC filing as the total dollar value of its rides, deliveries and freight shipments without adjustment for discounts, refunds, driver and merchant earnings or driver incentives. Source
  • ASU 2014-09 paragraph 606-10-50-13 requires disclosure of the transaction price allocated to unsatisfied performance obligations at period end, and paragraph 606-10-32-11 limits variable consideration to amounts not expected to reverse significantly. Source
  • Reddit's S-1 reported January 2024 data licensing arrangements with an aggregate contract value of $203.0 million and two-to-three-year terms, with at least $66.4 million expected as 2024 revenue. Source
  • Reddit's fiscal 2025 annual report states that substantially all of the contract value associated with its licensing revenue is derived from two partners. Source
  • micro1 announced in October 2026 a stated plan to spend $1 billion over 12 months acquiring and licensing enterprise operational data for reinforcement learning environments. Source
  • ASU 2016-20 added exemptions from the remaining performance obligation disclosure, including for sales- or usage-based royalties on licensed intellectual property. Source
  • The SEC's 2020 interpretive guidance on key performance indicators in MD&A expects a metric to be accompanied by a clear definition and an explanation of how it is calculated. Source
  • SEC staff guidance treats non-GAAP adjustments that change GAAP recognition and measurement principles as individually tailored and potentially misleading. Source
  • ASU 2016-08 states that an entity is a principal if it controls the specified good or service before it is transferred to a customer, and that an agent's revenue may be the net amount it retains. Source

Request SourceX investor materials

SourceX shares company and commercial metrics with investors directly. Any offering is made only to verified accredited investors.

See if your company qualifies