Synthetic comparables in PE valuation, when the comp set is generated
A new category of data vendor is offering AI-generated comparable transaction sets to fill thin precedent data. The diligence question is whether a synthetic comp can support a valuation memo at all.
A small but growing set of data vendors have started offering synthetic transaction comparables to fill data gaps in private equity valuation work. The pitch is straightforward. Traditional comparable transaction databases (Pitchbook, Capital IQ, PrivCo, Dealogic, Mergermarket) have thin data in many subsectors, particularly for mid market and lower middle market deals where pricing and terms are often unpublished. A synthetic comp generator uses the available real data to model what a comparable transaction would have looked like, and produces a set of inferred comps with statistical confidence intervals.
The technology is real. The use case is real. The diligence implications are the part that has not caught up.
What synthetic comps actually are
A synthetic comparable is not a real transaction. It is an estimated transaction generated by a model trained on real transactions, intended to represent the kind of pricing and structure that would likely have applied to a hypothetical deal with specified parameters. The model produces a multiple, a structure (cash vs stock vs earnout), and sometimes a structured covenant set, with associated confidence bands.
For a thinly traded segment, this can be more informative than a tiny set of three or four real comps that may not be truly representative. For a deeply traded segment, it adds little.
The vendors offering this generally label their synthetic comps as synthetic. They are not pretending. The risk is not in the labeling. The risk is in how the synthetic comp is consumed by downstream users.
Where the diligence question arises
A valuation memo prepared for a fairness opinion, a board presentation, or a financing structure traditionally rests on identified, real precedent transactions. The opinion writer references specific deals, identifies the parameters that make them comparable, and explains the multiples. When a court or a regulator looks at the memo, the underlying data is verifiable.
A valuation memo that rests partly on synthetic comps is a different artifact. The "deal" being referenced does not exist as a transaction. It exists as a model output, with a confidence interval. The verifiability of the underlying data is a verifiability of the model and its inputs, not of the transactions themselves.
This creates two diligence questions that most corporate practitioners are not yet asking.
The first is whether the use of synthetic comps in a valuation analysis should be disclosed to the audiences that the analysis is presented to. The board getting the fairness opinion may reasonably expect that the precedent transactions are real. The board materials should probably say so when they are not.
The second is the question of professional standards. AICPA and other valuation standards have specific requirements about the methodology and the data used. Synthetic comps are not explicitly addressed in current standards. Whether they qualify as "guideline transactions" under the standards is a question without a settled answer.
How buyer side counsel should treat them
When a seller's process materials reference comparable transactions, counsel should ask whether any of the comps are synthetic. The answer might be no, in which case the question is harmless. If the answer is yes, the follow up questions are about which comps, what fraction of the analysis they support, what the methodology was, and what the confidence bands look like.
If the answer is "we don't know" or "the banker didn't say," that itself is information. It suggests the analysis has not been pressure tested for the data sources.
Why this matters more in middle market deals
In a large public deal, the comparable transaction set is usually robust and the use of synthetic comps would be redundant. In a five hundred million dollar middle market deal in a specialty subsector, synthetic comps may be the only path to a defensible comparable analysis. The exact deals that are most likely to use synthetic comps are the ones where the use of synthetic comps is most consequential.
What this means for AI valuation tools more broadly
The same dynamic applies to AI valuation tools that estimate enterprise value from operating data. The tool is generating an output that looks like a valuation. The question of whether the output can support a memo, a board presentation, or a fairness opinion is the same question as the synthetic comp question. The methodology should be disclosed. The confidence interval should be presented. The downstream user should know what they are looking at.
The corporate bar has not yet developed a consistent practice on this. The first time it shows up in litigation, the standards will start to crystallize. Counsel who think about it now will be better positioned when that happens.
I built a small client side tool that helps think about this. /tools/comp-confidence-checker/ takes a list of comparable transactions and asks structured questions about each one (was the price publicly disclosed, was the deal closed within the past three years, is the industry classification verifiable). The output is a coverage score that flags where the analysis is leaning on inferred rather than observed data. It runs in your browser, saves nothing, and is meant to focus a careful read, not replace one.
Walter Allison is a corporate attorney in Denver. He writes here about M&A, private equity, and venture capital structure.
Follow on LinkedIn ·
Firm bio ·
More posts