Pricing and new-logo assumptions fail more often than volume assumptions when management projections are tested against customer and channel interviews. A quality of earnings report covers none of these, by design.
What does a quality of earnings cover?
Whether reported historical earnings are sustainable and correctly stated: normalizing one-offs, checking revenue recognition, establishing a defensible adjusted EBITDA that a buyer can underwrite.
It is rigorous, well-defined work and it is one of the highest-value workstreams in a transaction. None of what follows is a criticism of it.
The output is a number a lender and a buyer can both underwrite, which is why it has become the default first workstream on most processes. It is also why its boundary is worth stating plainly: a quality of earnings answers a question about the past with a rigor nothing in the commercial workstream can match, and it is silent by design about everything that has not happened yet.
Where does a QoE stop?
At the historical base. It establishes what the company has earned and says nothing about whether the growth layered on top of that base will materialise, because it is not scoped to.
Buyers routinely read a clean QoE as validating the model rather than validating its starting point, and that reading is where the exposure sits.
The misreading is understandable. A quality of earnings arrives late, costs a great deal and comes back clean, and the natural conclusion is that the model has been checked. What has been checked is the first row of it. Every row after that rests on assumptions the accountants were never asked to look at, and the confidence the clean opinion generates attaches to all of them regardless.
Which projection assumptions fail most often?
Pricing and new-logo acquisition. Both require behavior that has not happened yet from people outside the company, which is precisely the category of assumption a financial workstream cannot test and a customer workstream can.
Volume assumptions built on an existing base hold most often, because they extrapolate behavior that is already observable.
There is a useful test for which assumptions need separate work: ask whose behavior has to change for the assumption to hold. If the answer is nobody — the existing base renews as it has — the historical record is reasonable evidence. If the answer names a customer, a buyer or a competitor who has not yet done the thing, no amount of financial diligence will reach it, however well the workstream is run.
How do you test a projection?
By testing the assumptions separately, with the people the projection depends on. A pricing assumption is tested with buyers; a new-logo assumption is tested with the segment being targeted.
That is a commercial workstream rather than a financial one, and it uses different evidence and different people.
Scoping that work narrowly is what makes it fit the timetable. Two assumptions, the people who would have to act on them, and a fixed set of questions is a week of fieldwork. Attempting to test the whole model turns it into a full commercial diligence, which is a different budget and a different calendar, and the usual result is that neither gets done properly.
When should commercial work run?
In parallel where the timetable allows. Running it after the QoE means the financial base is confirmed before anyone has tested whether the growth story behind the multiple is real.
Where the timetable does not allow parallel work, the commercial questions are usually the ones worth doing first, because they are the ones that move the price — and the ones a committee asks about.
Where both run, the two should be reconciled explicitly rather than filed side by side. A clean quality of earnings and a commercial finding that the pricing assumption will not hold is not a contradiction — it is the two workstreams doing their jobs — but a memo presenting them separately leaves the committee to reconcile them in the meeting, usually at the point when there is least time to do it well.