Arcvue
Sell-side preparation—for the CEO, and the banker running the process
One number you cannot explain costs $109,200. Not being able to explain it costs $1,243,700.
Arcvue Value Builder profiles a GovCon business the way a buyer or a lender will—ten factors, tracked over time—so the questions are answered before they are asked. This page follows Arcvue tracing one unexplained number through nine steps.
Specimen Systems LLC · an example 120-person contractor · all figures synthetic
Adjusted EBITDA of $2,900,000.00 goes into the book, and at that moment nothing is wrong.
Nothing is inherited yet. The banker builds the memorandum from what the company provides, the company provides what its finance team assembles, and the finance team assembles it from several systems in several formats over about three weeks. Everybody involved believes the number.
That is the ordinary case and it is worth being clear that nobody has done anything wrong. The add-backs are defensible, the intent is honest, and the figure is probably close to right. What it is not is sourced—nobody can point at the ledger rows that produce it, because it was assembled rather than derived.
What this page counts. Not whether the number is correct. What the last answer costs you—because in diligence that quantity compounds, and by stop 07 it is eleven times the figure that started it.
Synthetic. An invented firm in an invented process. Follow it down—the doubt is worth 1.28 times the number that caused it, and that ratio is arithmetic rather than a figure of speech.
A twenty-six-year-old analyst opens the waterfall and asks the most ordinary question there is.
The book from stop 01 reaches a buyer. Their analyst does what analysts do: takes the adjusted EBITDA, takes the revenue waterfall, and tries to walk from one to the other.
The question is not adversarial and it is not clever. It is which add-backs are in this number, and can I see them. Every seller expects it. The difference between firms is entirely in how long the answer takes—and whether the answer, when it arrives, agrees with the book.
The ledger supports $2,790,800.00, and nobody in the room can source the difference.
The question from stop 02 gets answered eleven days later, and the answer does not match. Bridged from the ledger—net income plus the interest and depreciation lines—adjusted EBITDA is $2,790,800.00. The book said $2,900,000.00.
Source the same trial balance every page of this site is built on, bridged as the deal page bridges it. Basis $109,200.00 is not a discovered error. It is a figure nobody can point at a row for—which in diligence is functionally the same thing. Synthetic sample.
Nobody on the buy side is worried about a hundred and nine thousand dollars.
They are worried about what else was assembled the same way.
Do the obvious arithmetic on stop 03 and the gap is worth five times itself at the multiple. That is the small half of the answer, and treating it as the whole is how sellers end up surprised.
An analyst who finds one figure nobody can source has learned something that is not about that figure: the numbers in this process are assembled rather than derived. Every other number in the book now belongs to the same class. That is not a mood or a negotiating posture—it is a correct inference, and it changes what the buyer has to do next.
Every experienced deal maker knows the cascade. One inconsistency in diligence leads to a second look, a broader request, a longer timeline, and a conversation about price that would not otherwise have happened. One red flag begets another—not because buyers are unfair, but because they have correctly updated on evidence.
The request list stops being a sample and becomes a census.
Stop 04 changed what the buyer believes, and belief has a mechanical consequence. Diligence works by sampling: test a few, and if they hold, accept the population. A sample only works if the population is homogeneous, and the buyer has just learned it is not.
So the list widens. Every add-back individually. Every contract's revenue recognition rather than the largest five. Indirect rate history for three years rather than a summary. Past performance documentation contract by contract. None of these requests are unreasonable and every one of them is now necessary, and each is answered by the same finance team already assembling documents by hand.
The census from stop 05 is paid for in calendar, and the calendar is what certainty is made of.
A widened request list does not cost money directly. It costs weeks—and a longer process is a materially different process, not the same one at a slower speed.
More time means more quarters to re-baseline against, more chance a contract award slips or a recompete is lost mid-diligence, more opportunity for the buyer's own circumstances to change, and more room for a second bidder to walk because the process looks troubled. Everyone wants the highest price. What a seller actually needs is the price that works for them and the certainty that the deal closes—and every extra week is spent buying the first at the expense of the second.
The doubt is worth more than the number, and the difference is exact.
The process from stop 06 arrives where these processes arrive. The buyer removes the unsupported $109,200.00 from the earnings they will pay for—that is fair and it was always going to happen. Then they mark the multiple down a quarter turn, because a business whose numbers required a census is a different risk from one whose numbers reconciled.
Screen Arcvue deal analysis, applied to a sell-side process. Basis the two losses decompose exactly: $546,000.00 plus $697,700.00 is the whole $1,243,700.00 of value change, with nothing left over. The trust is the larger half and it is the half no spreadsheet models. Synthetic sample. A quarter-turn is illustrative; the point is that whatever it is, it applies to ALL the earnings and not just the disputed part.
That last sentence in the provenance is the mechanism worth carrying away. The unsupported figure is removed from the earnings once. The multiple applies to every dollar of earnings, so a quarter-turn of doubt is charged on the entire business—which is why the small problem is always the expensive one.
The same company, two years earlier, having closed its books a different way.
Everything from stop 02 onward followed from one property of stop 01: the number was assembled rather than derived. Change that one property and none of the rest happens.
A business that has been closing this way arrives at diligence with adjusted EBITDA bridged from the ledger every month, add-backs documented as they are made rather than reconstructed, a revenue waterfall generated from live contract data, and indirect rate history kept monthly because the rates are computed nightly rather than annually. The data room is not built for the process. It is what the last twenty-four closes already produced.
And the analyst's question gets answered the same day. Which does not merely save eleven days—it prevents stop 04 from ever occurring, and stop 04 is where the $697,700.00 came from. The expensive thing was never the missing document. It was the inference the buyer drew while waiting for it.
What you were selling was never the number in the book.
Which is what stop 01 was really about.
Stop 01 put $2,900,000.00 into a memorandum and nothing was wrong. What was wrong—invisibly, and two years earlier—was that the number had no ledger behind it that anybody could reach in an afternoon.
A buyer is not paying for earnings. They are paying for earnings they believe will still be there after they own them, and the price of that belief is set by how the seller's numbers behave under examination. That is a property of how a business has been closing its books for years, and it cannot be acquired in the eight weeks before a process starts.
Which is why Arcvue scores it continuously rather than at kick-off. Value Builder profiles the business on the ten factors a GovCon buyer actually tests, computed from the same ledger and contract records everything else here runs on, and tracked over time—so the profile a buyer will build about you is one you have already been reading for two years.
| Factor | What a buyer is testing |
|---|---|
| Set-aside concentration | How much revenue rests on a status that does not survive the sale |
| Recompete risk | How much of the base has to be won again, and when |
| Prime and subcontract mix | Whether you hold the customer relationship or somebody else does |
| Contract concentration | What one contract ending would do to the rest |
| Customer desirability | Which agencies, and how durable their funding is |
| Service quality | Where the work sits between commodity staffing and mission-critical |
| Revenue trajectory | Direction and consistency, rather than one good year |
| EBITDA margin | The level, and whether it holds |
| Add-back quality | How much of adjusted EBITDA survives the examination at stop 03 |
| Management depth | Whether the business runs without the person selling it |
Screen Arcvue Value Builder, factor profile. Read this one row by row. The ten are scored separately and composed into one profile, so they are not addends and no total rule is drawn. Synthetic sample.
For the banker rather than the seller. Recommend it at kick-off and the change is in your own work: Arcvue refreshes the financials without anybody chasing them, the waterfall does not need rebuilding after every update, and diligence questions are answered by the analyst straight from the data rather than routed through the client and back. The engagements that reach agreement fastest are the ones where nobody had to re-trade, and re-trades start at stop 03.