Insights
Deal economics
193 guides · page 3 of 5
Cheaper synthetic generation won't zero out data pricing: a ceiling test
Cheaper synthetic generation cuts data pricing only by the generation share of a buyer's best substitute. A ceiling rule, a value map and a parity test plan.
Buyer-ready dataset cost: a five-layer model that charges proof and failures
Buyer-ready dataset cost, fully loaded: a five-layer model that charges the evidence pack, specialist review and failed candidates to each accepted dataset.
Long-horizon agent dataset economics: the cost moves to verification
Long-horizon agent dataset economics: why cost per verified episode, not per record, sets value, with a rerunnable cost model, a value chain and the risks.
Small dataset, high value? A lift-per-cost test against a large one
When is a small dataset high value? A study design, cost waterfall and break-even rule that put lift, coverage and all-in cost on the same axes for investors.
The AI data valuation milestones that hold up when a committee checks
AI data valuation milestones that hold up: revenue recognized after acceptance, repeat delivery, rights that survive review, spread buyers and a downside grid.
Proof, leverage or dependence: an AI data buyer concentration risk map
AI data buyer concentration risk, mapped: four exposure measures, three downside cases and the evidence that tells buyer validation apart from dependence.
AI data licensing buyer acquisition cost: how repeat orders flatter CAC
AI data licensing buyer acquisition cost, split three ways: new buyers, repeat orders and the blend, with an illustrative payback model and metrics to request.
Historical dataset scope extension: three checks per added year
Before a historical dataset scope extension, check each added year for surviving records, the rights in force when they were made and unchanged meaning.
Data or labor? How to spot a data startup services business model
A data startup services business model shows in its hours, not its label. Code each fulfillment hour by dataset and buyer, then test the second-buyer ratio.
Dataset documentation defensibility: a moat scorecard with copy tests
When does dataset documentation defensibility hold? Six tests, a placebo and a copy attempt show if docs cut review and integration work a rival cannot match.
AI data contract dependency: what binds after the buyer can leave
AI data contract dependency starts when a seller's restrictions outlast the buyer's commitment. A clause ledger and exit-date test show what a big deal costs.
Dataset refresh contract economics: renewal odds that beat a lump sum
Dataset refresh contract economics, worked through: normalize rights, compound the renewal odds and find the rate a refresh deal must beat to top a lump sum.
Data rights defensibility: five links, and the tests that break them
A five-link test for data rights defensibility: when a license creates repeat demand that rivals cannot replace, with a disconfirmation test for each link.
Data exclusivity economic tradeoff: when a higher price loses value
Data exclusivity economic tradeoff for investors: weigh one exclusive license against the repeat licenses the rights allow, with a break-even rule and map.
Dataset evaluation cash conversion: 4 tests before faster review pays
Dataset evaluation cash conversion depends on contract design. Four tests show when a faster buyer review brings cash forward and when it only moves the wait.
Synthetic preference data validation: when AI-judged pairs need experts
Synthetic preference data validation for investors: where AI judges match people, where expert agreement falls to 64-68%, and a matched-budget test to run.
Dataset qualification economic threshold: when to decline, gate by gate
A dataset qualification economic threshold, priced gate by gate: when declining a candidate dataset beats yes, why sunk cost never counts, and what to ask.
Supplier scale data differentiation: a seven-line dilution scorecard
When does adding suppliers dilute a dataset's moat? A supplier scale data differentiation scorecard for overlap, uneven quality and commodity coverage.
Synthetic versus licensed data economics: the verifier decides
Synthetic versus licensed data economics, task by task: a substitution matrix built on verifiers, real outcomes and rights, with risks investors can test.
Dataset no-deal decision efficiency: test the unfixable blockers first
Dataset no-deal decision efficiency: order tests by cost, stop on blockers no data fix can cure, then close with a reasoned notice and a clean data record.
AI data buying mandate vs minimum commitment: a six-rung demand test
An AI data buying mandate sets a ceiling, not a floor. A six-rung ladder and a concentration matrix show investors which demand buyers must pay for.
Dataset manifest investor diligence: a three-way match for data deals
Dataset manifest investor diligence: match each manifest line to signed scope and to acceptance and invoice records, then test rights, quality and versions.
Annotation versus data licensing economics: labor resets, rights repeat
Annotation versus data licensing economics: why labor revenue resets with each order while licensed rights can sell again, plus a five-shock test for investors.
When does dataset qualification capital allocation beat more leads?
Dataset qualification capital allocation for investors: when an earlier screen adds more accepted supply than buying leads, with a break-even rule and a model.
Large dataset transaction profitability: four costs that grow faster than price
Large dataset transaction profitability, modeled: how bespoke work, delay, buyer concentration and service duties can outgrow price as a data deal scales.
AI data company bookings, GMV and revenue: three clocks, three charts
AI data company bookings, GMV and revenue run on three clocks. A chart spec, a backlog roll-forward and Reddit's filings show how to keep contingent value out.
Contingent data deal value analysis belongs on a probability tree
Contingent data deal value analysis, step by step: chain conditional stage odds, price refund windows, and test shared causes before trusting a headline.
Data operations exception cost: budget the tail, not the average case
Data operations exception cost hides in rare rights, extraction and context cases. A 459-fix public log and an exact model show how to size the reserve.
Data licensing price uncertainty: how a comparable becomes a 5x range
Data licensing price uncertainty, made testable: why stacked adjustments widen comparables, how ranges narrow in a negotiation, and how investors check them.
Dataset volume's diminishing value: 1,800x the lines, 3x the patterns
Dataset volume shows diminishing value: in public system logs, 1,800 times the lines held 3 times the patterns. Value the next block of records, net of cost.
Real data's synthetic seed value: why fewer records can be worth more
Real data's synthetic seed value for investors: a substitution matrix, a seed-only experiment and the license terms that decide who keeps the multiplier.
Revenue per dataset metric: five restatements before you trust a rise
Revenue per dataset metric for investors: five restatements (count, services, term, scope, usable records) that show how much of a reported rise is real.
AI data investment scenario probabilities: one edit, four answers
AI data investment scenario probabilities are judgments. An illustrative model shows why they must stay editable, and how to log each change and its reason.
Data supplier support cost: why the help belongs in gross margin
Data supplier support cost for investors: four kinds of supplier assistance, where each belongs in the accounts, and a cohort model of hours per dataset.
Supplier trust data moat: six tests an export script cannot pass
Test a supplier trust data moat with six evidence lines, from signed permissions to renewals and repeat buyers, each with a metric and a disconfirmation test.
Dataset price per record limitations: 7 fixes before comparing deals
Dataset price per record limitations, fixed: a rights-normalized worksheet that adjusts unit, usable scope, rights, term and services before deals are compared.
Business context dataset pricing: test the lift a buyer cannot rebuild
Business context dataset pricing, tested on public insurer data: the insurer's own records found 41% more buyers than zip-level context. Worksheet inside.
AI data procurement renewal decision: renew, expand or stop?
An AI data procurement renewal decision should rest on task utility: a slice-level scorecard that shows whether to renew, narrow, expand via pilot or stop.
AI training data total acquisition cost, measured in a pilot first
AI training data total acquisition cost: measure preparation, review, storage and refresh drivers on a pilot sample, then scale them with low and high cases.
Customer exclusions in data licensing: does buyer fit survive?
Data licensing customer exclusions change buyer fit. Compare scopes with and without your largest customer to see which coverage claims still hold up.
Data licensing cycle time study: a protocol for timing each deal stage
A data licensing cycle time study protocol: stage timestamps, dependency fields, censoring rules and a pre-registration checklist. Method only, no results.
The cost of delayed payment in data licensing: when bigger is worse
Data licensing delayed payment cost, explained: compare offers by risk-adjusted present value, find the breakeven for later money and protections to ask for.
Data licensing payout reconciliation: from deal price to the amount you receive
Data licensing payout reconciliation step by step: a payout bridge from deal price to cash received, a per-payment tie-out and an itemized expense checklist.
What does it cost to prepare a business dataset? A cost study protocol
A data licensing preparation cost study protocol and cost dictionary: how to measure what it costs to prepare a real business dataset, line by line.
Data licensing renewal option value: what is a refresh option worth?
Data licensing renewal option value from the supplier's side: estimate exercise odds, carrying costs and what the option locks up, with a scenario worksheet.
Internal AI savings vs external data licensing: how to compare the return
Data monetization internal versus external return: compare AI savings with licensing proceeds using one assumptions table, shared costs and switching values.
When to stop buying more of the same data: a stop rule experiment protocol
A dataset acquisition stop rule experiment: tranche arms, controls, an untouched holdout and a preregistered gain-versus-cost rule for when to stop buying data.
Dataset exclusivity opportunity cost: a scenario worksheet for owners
Estimate dataset exclusivity opportunity cost before negotiating a premium: an editable scenario worksheet, a simple formula and three worked structures.