A well-calibrated AI estimate gives shareholders a usable range and an honest view of uncertainty. A poorly calibrated one creates false confidence. Skeptical buyers ask how anyone knows a private-equity price is right. The answer depends on what "right" means for a range in an illiquid market, where the same company's shares may clear at different prices in different transactions on the same day. This article examines each signal and what determines range width and estimate reliability.
Why range accuracy is not one number
Earlyasset reports provide low-to-high ranges, not point estimates, and include a stated confidence tier. That choice avoids false precision. The clearing price for a particular lot of private company shares depends on details the model cannot fully observe: the buyer's return requirements, the seller's urgency, competition for that block, the ROFR process outcome, and the legal mechanics of the transfer.
A range should be judged differently from a point prediction. The key questions are whether the cleared transaction price lands inside the stated range and whether the confidence tier ranks uncertainty correctly among estimates. These reflect the methodology's design, not a single publishable percentage. Anyone citing a specific percentage accuracy for AI private equity pricing is almost certainly answering a narrower question than the one buyers think they are asking.
Signal 1: Recent secondary transaction data
If a company completed secondary transactions in the preceding 12 months, Earlyasset uses those cleared prices as its main anchor. This is the strongest structural evidence generally available because it directly shows what arm's-length buyers and sellers agreed shares were worth in a real transaction. The model then focuses on adjusting for share class, buyer type, transaction size, and recency, rather than valuing the shares from indirect proxies.
The reason this signal tends to be more reliable is simple: it requires less inference. Asking what the same company's shares trade at today after adjusting for the last 12 months of changes is less uncertain than asking what they would trade at if secondary participants priced them using only public comparables and round data.
Availability is the weakness of this signal. Some companies lack disclosed or accessible recent secondary activity. In those cases, the model must give greater weight to the next signal tier.
Signal 2: Round multiple extrapolation
Without recent secondary history, the model combines round-over-round multiple progression with sector-adjusted secondary discount rates applied to the last preferred round valuation. This is substantially more indirect than recent secondary data, so the expected range is wider.
The central difficulty is that the secondary discount on a preferred round valuation changes significantly with sector, growth stage, and market conditions. A company at a given valuation differs from an equivalent company when public multiples have compressed since its last round closed rather than remained stable. Sector adjustments help, but the link between preferred round price and secondary common clearing price is not stable in every condition. Estimates relying mainly on this signal therefore carry structurally wider uncertainty.
The useful test is whether a round-multiple range reflects actual uncertainty instead of implying false precision. A tight range based mainly on round multiple extrapolation is more likely to misstate certainty than a range widened to match the quality of the signal.
Signal 3: Exit comparable benchmarking
For a late-stage company with a clear public comparable set, exit comparable analysis applies current public trading multiples from the same sector to estimated private-company financials, then deducts a private market illiquidity discount. It works best when the comparables resemble the private company in business model and revenue scale, their multiples are stable rather than volatile, and the private company's revenue trajectory can be estimated with reasonable confidence from available data.
For earlier-stage companies, comparable benchmarking is intentionally more speculative. Relevant exit comps are future exits at an unknown stage and timeline, rather than current public trading multiples. The discount for that timing uncertainty is wide and partly judgment-dependent, making the range wider than late-stage comparable analysis would. The model reflects this by assigning lower confidence to estimates weighted heavily toward exit comparables for earlier-stage companies.
How cap table complexity affects all three signals
Each signal starts with company-level valuation, which must be converted into a share-class-specific price for the common shares being considered. That conversion creates an error source separate from signal quality: cap table complexity.
With multiple preferred stock series, non-standard participation provisions, or high aggregate liquidation preferences relative to implied enterprise value, accurately pricing common shares requires detailed cap table modeling. Earlyasset includes that analysis when relevant data is available. If it is unavailable, the model identifies the gap as added uncertainty and reflects it in the confidence tier.
A shareholder facing a complex cap table should read the report range as covering reasonable cap table assumptions as well as company valuation variation. A company worth $2 billion may support very different common-share values depending on whether the preferred stack is $200 million or $800 million, and whether the preferred is non-participating or fully participating. The company valuation is identical in both examples, but common share economics differ materially.
What confidence tiers actually communicate
Every Earlyasset report pairs its price range with a confidence tier. High confidence means strong recent secondary transaction anchors, stable public comparable data, and limited cap table uncertainty. Moderate confidence means greater reliance on round multiple extrapolation or comparable analysis, with ranges reflecting those signals' structural limits. Low confidence means indirect signals, older data, or poorly fitting comparables, and the ranges are explicitly wide.
The tier describes the available input data, not a grade for analytical quality. A Low confidence report with an appropriately wide range is a correct output. A High confidence claim built on weak signals is the real quality failure. We are not saying the assignments are infallibly calibrated. We are saying the tier is intended to track the structural quality of available evidence, not conceal its absence.
Using a wide range in decisions
A wide range can still guide a decision, even when a tight number would feel more satisfying. Consider a Low confidence report showing $14 to $34 per share. A tender offer at $12 per share supports the view that the offer is likely below current secondary market clearing, wherever reality falls within the range. At $31 per share, the report does not clearly support non-participation: the price is near the range's upper end, so pursuing secondary alternatives could carry significant downside. Width changes confidence in the comparison, but does not make comparison impossible.
What we do not claim
We are not saying AI pricing produces more accurate valuations than thorough fundamental analysis by analysts with direct access to company financials, management, and competitive intelligence. Knowledgeable analysts with that access will produce better-informed valuations for a specific company than signal aggregation from publicly available data. This does not mean the product has no use. It addresses the lack of institutional secondary research access for shareholders and buyers by providing a structured, market-grounded price range for practical decisions. For most private secondary market participants, the alternative to an AI-derived range is not better analysis, but no structured analysis.
Important: Note: AI outputs are informational tools only, not investment advice or formal appraisals. Private market risks include illiquidity and loss of capital. Earlyasset report ranges are estimates from available market signals, not guarantees of transaction outcomes. The methodology is a general explanation, not measured performance against a specific benchmark. Consult qualified financial and legal advisors before transaction decisions involving significant capital.