Every sale process reaches the point where the commercial story has to stop being a story. The buyer asks for the number behind a slide, then how it was calculated, then for the same number cut a different way. Financial and legal evidence is ready for that, because counsel and the CFO have spent months assembling a data room. Commercial evidence usually is not, because nobody owns it. The requests arrive through management Q&A, land on a marketing director who has never been inside a deal, and get answered three ways across six weeks. What opens up is not a documentation problem. It gets priced.
A commercial evidence pack is the set of numbers a company's commercial function has to be able to produce, and produce again, when a buyer asks during a sale process: revenue by acquisition source, acquisition cost and payback by channel, pipeline and its conversion history, retention by cohort, the owned versus rented mix, and customer concentration, each with a written method and a named owner.
It is not the data room, and confusing the two is the usual reason this work never starts. The data room is built by counsel and finance: contracts, statutory accounts, tax, employment, insurance, the quality of earnings work. The evidence pack answers a different question. The data room proves the business is what it says it is. The evidence pack proves the growth is what it says it is, and that it will still be there under an owner who is not the founder. In the S&P Global 2026 survey, 71 percent of general partners and 53 percent of limited partners said they prioritize operational value creation over financial engineering. When the price is being paid for the growth rather than for the balance sheet, the growth is what gets tested.
What follows is the artifact list, how long each takes to produce from a standing start, and what happens to the price when it cannot be produced. The buyer's side of the same exercise, and the tests a deal team runs, are in marketing due diligence; how a buyer scopes the wider technical and data work is in digital due diligence. This is the seller's side: what the commercial function owes, and when it has to start building it.
Seven items cover almost every commercial request a mid-market process generates. Each is listed with the grain it has to be built at, because grain is where most of them fail, and with an honest production estimate.
| Artifact | Grain | Time from a standing start |
|---|---|---|
| Revenue by acquisition source | Transaction line, 24 to 36 months | Two to three weeks with clean billing data, six when records sit in several systems |
| Acquisition cost and payback by channel | Channel, monthly, spend lagged to the cohort it produced | Two to four weeks once source tagging exists |
| Pipeline and its conversion history | Opportunity record, by stage and segment | One to two weeks to extract, several quarters if the CRM was never used consistently |
| Retention and repeat by cohort | Customer, by month of first purchase | Two to three weeks, from the same extract as the first artifact |
| Owned versus rented mix, with trend | Channel, quarterly, at least eight quarters | Days, once the first two exist |
| Customer concentration | Customer key, top ten and top twenty | One week, after the customer key is settled |
| Method statements and named owners | One page per number | Written as each number is produced, never afterwards |
Revenue by acquisition source is the foundation, and the other numbers are derived from it. It has to be built at transaction grain, not assembled from monthly summaries, because a buyer will re-cut it and a summary cannot be re-cut. The five ways to segment a revenue base are set out in the revenue stream map; treat it as the build instructions for this artifact.
Acquisition cost and payback by channel is where most packs get challenged, and almost always on method rather than on arithmetic. The buyer will ask what sits in the numerator, which margin sits in the denominator, and whether spend was lagged to the cohort it produced. Our guide to blended CAC in a services business covers what to count. The diligence point is narrower: whatever you count, count it the same way every time you are asked.
Pipeline and its conversion history is the one artifact that may not be a production job at all. If the CRM has been used consistently, the extract takes a couple of weeks. If stages were never defined, or half the pipeline has no owner, there is nothing to extract and the work is remediation measured in quarters. That call has to be made honestly and early: it is the difference between a task and a program.
Customer concentration looks like the easiest and is routinely wrong in multi-brand platforms, where one customer appears under two brands with two records and the top ten looks better than it is. Settle the customer key before anyone computes a ratio. Owned versus rented mix is nearly free once the first two artifacts exist, and it is the one number that most directly supports a growth story, because it shows whether demand is being bought again each quarter or accumulating. One five-brand home services platform found a route to a 60 percent reduction in acquisition cost by tripling the share of owned channels, a shift visible in this artifact and almost nowhere else.
A number without its method is not evidence. It is an assertion the buyer has to test, which turns a document you produced into a question you have to answer. Each artifact therefore travels with a single page stating what is included and excluded, the window, the source system, who runs the query, and the known limitations. That last line does more work than sellers expect. A stated limitation is a credible number; a discovered one is a credibility problem that spreads to the numbers around it.
The working rule is that you should never put a number into a diligence process that you cannot produce again in forty-eight hours, the same way, to the same answer. Buyers re-ask. They ask the same question two months later, and they have a second team ask it differently. Two versions of acquisition cost, produced three weeks apart by two people using two definitions, does more damage than one imperfect number produced consistently: the first puts the commercial function in question, the second only a metric.
This is also why named owners belong in the pack. The buyer is judging whether these numbers exist because the business runs on them or because someone assembled them for the sale. A number produced monthly by a named owner reads as an operating metric; the same number with no owner reads as a deal artifact and gets weighted accordingly. Building that reporting spine is covered in building a single source of revenue truth.
Missing evidence is not one thing, and sellers who treat it as one thing prepare in the wrong order. There are four outcomes, and they differ by roughly an order of magnitude.
A closing condition. Discrete, verifiable, quick to fix: an ad account or a domain held in an agency's name, a tracking setup nobody can access. These cost lawyer time and nothing else, and are the cheapest category to be caught by.
A price chip. A gap that can be quantified and argued about: acquisition cost higher than management presented once computed consistently, or concentration worse once the customer key is fixed. The seller can contest the calculation, which is a real negotiation with a bounded outcome.
Deferred consideration. When the gap is about whether growth is durable rather than about a discrete number, price moves into an earn-out or an escrow. The buyer is not refusing to pay for the growth; it is refusing to pay for it now. The seller keeps working for money it thought it had sold.
A silent discount. The worst outcome, and the only one the seller never sees. Revenue the buyer cannot trace does not get argued about; it gets excluded from the underwriting and the model is run without it. In one PE-backed services company, 38 percent of revenue arrived through word of mouth that no system tracked. That is close to two fifths of the business a diligence team could not have credited, and no line item in a negotiation would show it.
The pattern is worth stating plainly, because it sets the order of work. Discrete gaps get priced. Systemic gaps get deferred or discounted. Preparation should therefore start with the artifacts that prove a system exists, not the ones easiest to produce.
The sequence is set by dependency rather than by the calendar. Acquisition source tagging on transaction records is the gate: three of the seven cannot be produced without it, so it comes first, alongside settling the customer key. Then the revenue segmentation, channel economics, cohort retention, and finally the mix trend and concentration, which fall out of work already done. Pipeline runs as a parallel track, because it lives in a different system and its fix is behavioral rather than analytical. Method statements are written as each number is produced; nobody has written them successfully at the end.
The real reason to start early is not the production work. Most of these artifacts take weeks. What takes eighteen months is history. Acquisition source cannot be retro-tagged onto two years of past transactions, cohorts cannot be constructed for customers who were never keyed consistently, and a trend line needs quarters that have already happened. A company that starts twelve months out shows twelve months of history; one that starts three months out produces the same artifacts and shows three months, which reads as preparation rather than as operating discipline. That is the one constraint money cannot compress, and it is why the question is worth answering long before a banker is appointed. The wider eighteen-month plan, including what to fix rather than merely evidence, is in commercial exit readiness.
For a company that does not know which of the seven it can already produce, the starting point is a stocktake rather than a build. A commercial audit runs about eight weeks and establishes where revenue comes from, what it costs to acquire, and which of these numbers the business can stand behind today. It is the same exercise that produced the 38 percent figure above. What that fact base changes is on our results page, and how we work sets out the engagement.
An acquisition cost that flatters the engine. Total marketing budget divided by all new customers, including the ones who were referred, inherited from an acquisition, or walked in. The result is a number no channel decision was ever made on, and the buyer recomputes it on paid customers alone. The difference becomes the conversation, and it is a conversation about the seller's judgment rather than the business.
Cohort retention computed on survivors. Retention measured across the customers still on the books today rather than across everyone who started in the cohort. It produces a flattering curve that collapses the moment the buyer rebuilds it from the transaction file, and it is the most common error in this pack.
A pipeline exported the week it is requested. Stages get tidied, dead opportunities closed, owners reassigned on the way out of the system. Buyers know this, which is why they ask for the same export sixty days later and compare. A pipeline that has been maintained for a year needs no tidying, and one that needs tidying cannot be tidied invisibly.
None is dishonesty. Each is what happens when a number is produced for a process by someone who does not run it. That is the case for the seventh artifact, the one with no analytical content at all: a named owner who produces each number monthly, before anyone outside the company ever asks for it.
A commercial evidence pack is the set of numbers a company's commercial function has to be able to produce, and produce again, when a buyer asks during a sale process: revenue by acquisition source, acquisition cost and payback by channel, pipeline and its conversion history, retention by cohort, the owned versus rented mix, and customer concentration, each with a written method and a named owner.
The production work is weeks, not quarters. Most of the artifacts can be built in two to six weeks each from a standing start, and several share a single data extract. The lead time is about history rather than effort: three of the numbers need a trailing record that only exists if the company was already measuring, which is why the work starts twelve to eighteen months out.
Sell-side due diligence is an advisory engagement, usually financial, in which a firm audits the business before the buyer does and issues a report. The commercial evidence pack is not a report. It is the underlying set of numbers, methods and named owners that the commercial function keeps ready, and that any report or management answer has to reconcile to.
Untracked demand. A discrete gap such as an ad account held in an agency's name gets fixed as a closing condition and costs almost nothing. Revenue that cannot be traced to a source is different. The buyer cannot credit what it cannot see, so the whole forecast is quietly discounted, and nobody tells the seller that it happened.
Partly. Six months is enough to produce every artifact that can be built from existing records, to settle the customer key, and to write the methods down. It is not enough to create trailing history. The honest approach is to produce what the data supports, state the limits in writing, and stop rather than show a number that will not survive a second cut.
Have a revenue problem the board is asking about? Start a conversation.