Insights
Deal economics
193 guides · page 2 of 5
AI training dataset buyer value: count the cases that clear the gate
AI training dataset buyer value lives at the buyer's deployment gate: in a public test, a 2.7-point accuracy gain cleared 17.0 points more cases past it.
Partial archive licensing: a supplier's accept, counter or decline memo
The partial archive licensing supplier decision: define the slice in system terms, cost the effort by owner and check what it does to the rest.
Accepted and collected: an AI data startup capital efficiency test
AI data startup capital efficiency, measured as accepted value and cash per full operating dollar, with a marginal test for spending that outruns acceptance.
Dataset total acquisition cost: price the work each offer leaves you
Dataset total acquisition cost puts license, preparation, validation and integration work side by side. Price what each offer leaves you, then compare quotes.
Data preparation economies of scale: a four-layer cost curve to test
Data preparation economies of scale stop at an exception floor. A four-layer model shows where cost per record flattens and the evidence investors should seek.
Archive context reconstruction cost: a per-element restore-or-skip test
Estimate archive context reconstruction cost with a timed pilot, then restore only elements whose buyer-utility gain beats noise and pays for its hours.
One big license, three distorted metrics: data deal annualized revenue
Data deal annualized revenue can double a run-rate when one license lands in one quarter. A labeled model, filings and a checklist show how to report around it.
Reliable agent workflow data cost: why upkeep outgrows the build
Reliable agent workflow data cost is mostly evaluation upkeep: repeated runs, expert review and refreshes. A five-line bill and a model you can test.
Supplier data export cost decision: cheaper routes before a full pull
A supplier data export cost decision guide: compare four export routes, sort hours into no-regret and deal-specific, and spend only what each gate supports.
AI data transaction waterfall: request the period, not the best deal
The AI data transaction waterfall investors should request: every deal opened in a period, from buyer billings to data owner share, contribution and cash held.
Supplier selected dataset payment terms: six triggers before cash
Your dataset was selected. Before counting the money, check the six payment triggers that still sit between selection and cash, and who controls each one.
Stress five risks at once in a dataset licensing downside model
A dataset licensing downside model for investors: no closing, rejected deliveries, refunds, rights challenges and exclusivity, stressed alone and together.
Direct data licensor business model vs data broker: a four-question test
How the direct data licensor business model differs from a data broker: who owns rights, does the work, supplies tooling and keeps recurring value.
AI dataset pricing evidence: what 16,147 data listings cannot tell you
AI dataset pricing evidence comes in five grades. In 16,147 listings, 47.8% of paid ones show no price. A worksheet admits only deals with stated rights.
Dataset valuation vs data company valuation: separate the two numbers
Dataset valuation and data company valuation answer different questions. A rights-normalized bridge separates asset cash from engine, overhead and risk.
Data deal margin vs cash conversion: bridging booked profit to cash
Data deal margin vs cash conversion: a five-line bridge from booked profit to cash, four payment structures compared, and the deal metrics to request.
Data rejection vs commercial terms: log two verdicts, not one
Data rejection or commercial terms? Score every dataset deal on two verdicts with a not-tested state, so funnel analysis shows which fix closes more deals.
Unique data investment value: six tests that separate rare from useful
Unique data investment value depends on task lift, not rarity. A six-test moat scorecard with disconfirmation tests separates novel datasets from useful ones.
Reusable data connector investment pays back in suppliers, not sources
A reusable data connector investment pays back only when one build serves enough suppliers: the payback math, a reuse ledger and a 522-source catalog count.
A curated data subscription business model renews only what goes stale
Curated data subscription business model, tested: refresh duties, what cancelling removes and retention bases that separate renewals from installments.
Hybrid training data economics: a complement-or-substitute test
Hybrid training data economics for investors: test whether synthetic data raises or erodes the value of real records, using cost per held-out point in a grid.
Partial dataset acceptance economics: five pricing terms, one subset
Partial dataset acceptance economics, modeled: when a buyer takes a subset of a fully prepared dataset, which pricing terms protect margin, and what to ask.
Repeat supplier qualification economics: price evidence by shelf life
Repeat supplier qualification economics for AI data: which evidence a second deal can reuse, which must be checked fresh, and a model of the hours saved.
Enterprise AI versus frontier lab data buyers: a re-review ratio test
Enterprise AI versus frontier lab data buyers, compared on contract size, review effort, customization and repeat orders, plus the re-review ratio to request.
Dataset recreation cost moat: the six layers a rival pays to rebuild
Test a dataset recreation cost moat: price six rebuild layers, keep only those tools cannot cheapen and substitutes cannot skip, then try to disprove each one.
Data startup supplier payment risk: five dates that size the cash gap
Data startup supplier payment risk, modeled: five contract dates, a stress grid for late buyers and disputes, and the questions that size the cash gap.
Dataset acceptance criteria diligence: a 10-point discretion score
Dataset acceptance criteria diligence for investors: score five clause elements to see whether a signed data deal is earned revenue or an option for the buyer.
Data business moat scorecard: grade the evidence, not the pitch
A data business moat scorecard for investors: eight lines across substitutability, access, utility and execution, each graded by evidence and a kill test.
Fresh dataset pricing: five proofs before a recency premium counts
Fresh dataset pricing for investors: split the quote into historical baseline, recency increment and refresh cost, then test five proofs before paying more.
Dataset gross margin rework: the cycles a reported margin leaves out
Dataset gross margin rework, priced: a ledger for every repeat cycle, a stage-by-trigger heat map and a corrected margin that charges failed candidates too.
Dataset inventory procurement risk: what to build before acceptance
Dataset inventory procurement risk, tested: which preparation work each buyer stage justifies, what a speculative build strands and the ratio to ask for.
Where a data licensing SaaS valuation comparison breaks, metric by metric
A data licensing SaaS valuation comparison breaks when run-rate, retention and gross margin are borrowed unchanged. What each assumes, and what to measure.
More suppliers, fewer licenses? Testing data supply growth against demand fit
A cohort model of data supply growth and demand fit: how scaling supplier intake before buyer fit strands preparation hours and can cut accepted datasets.
Buyer urgency in dataset pricing: value what outlasts the deadline
Buyer urgency dataset pricing for investors: split a rushed deal into a deadline premium and durable value with a rights-normalized, deadline-off worksheet.
How to measure data licensing buyer retention when nobody subscribes
Data licensing buyer retention needs repeat-purchase cohorts, not renewal rates. A public test shows one cohort reading 17.5% or 48.3% by window alone.
Data inventory or on-demand sourcing? The fit-rate hurdle that decides
Data inventory vs on-demand sourcing in AI data: a fit-rate hurdle built from capital at risk, completion speed and resale count, with diligence questions.
Data supplier acquisition cost: count the suppliers that never signed
Measure data supplier acquisition cost with a cohort waterfall that spreads the effort on failed candidates over the datasets buyers accepted and paid for.
Data supply operational margin drivers: test three on matched cohorts
Data supply operational margin drivers, tested: why qualification accuracy moves margin most, why more applications can cut it, and how mix fakes gains.
Custom data or repeatable product strategy? Start with same-company margins
A custom data or repeatable product strategy sets the margin. Same-company filings show the gap, and a deliverable ledger shows how much custom work can resell.
Data licensing unit economics: trace one dataset from first call to cash
Data licensing unit economics, line by line: charge failed candidates, rework and cash timing to one dataset, with a waterfall and stress test to reuse.
Negative examples training data value: price the proof, not the count
Negative examples training data value rests on measured lift and the hours needed to prove each negative. TruthfulQA's history shows what that upkeep costs.
Dataset rework acceptance cost: count every cycle, not just the yes
Dataset rework acceptance cost covers every failed cycle before the final yes. Use a cycle record, a six-step routing table and four honest outcome labels.
Training tokens vs verified tasks: the economics investors should track
Training tokens vs verified tasks: why the units never convert, a dual ledger of what to track in each, and the demand evidence that counts in AI data deals.
Turn dataset rejection feedback into qualification rules that travel
Dataset rejection qualification feedback, done right: code each rejection, route it with a decision tree, and promote only reasons that recur across buyers.
Dataset valuation framework: three anchors before the first offer
A dataset valuation framework for the first offer: set the seller floor, cap it with replacement cost and use-case value, then test it on executed comparables.
Incomplete dataset valuation: price the repair route, not the gap
Incomplete dataset valuation: size each repair, test the buyer's task before and after, credit only licensable repairs, and see what 427 ZooKeeper commits show.
Dataset price benchmark methodology: four rules before any number
Dataset price benchmark methodology for investors: the sample, definitions, rights cohorts and disclosure rules a benchmark needs before it publishes anything.
SourceX offer verification: 8 price facts sorted by who can check them
SourceX publishes no pricing formula. This ledger sorts eight offer price facts into published, signed and private tiers, with the requests that fill each gap.