Which add-backs should you never accept?

Adjusted EBITDA is the number the deal is priced off, and it is the number over which the seller has the most discretion. On a business trading at 7x, a $400,000 adjustment that should not be there costs the buyer $2.8 million. Most of the value at stake in a lower-middle-market process sits in this schedule, and most of the diligence attention sits somewhere else.

Three tests catch the majority of problems, and after them a list of specific items that should not survive.

The three tests

Does it recur? Any adjustment appearing in two or more of the last three years is not one-time, regardless of the label attached to it. This is the single most productive test available and it is arithmetic. Lay the schedules side by side and look for repetition. "One-time legal settlement" in 2023, 2024 and 2025 describes a company with a litigation cost line, not a company with three unrelated one-time events.

Is it a cost that leaves with the seller, or a cost that leaves because somebody decided it should? Owner compensation above market is a legitimate adjustment — you will not pay the seller's salary after close. The three additional sales roles the seller eliminated four months before going to market are not. The first is structural. The second is a decision the buyer inherits along with whatever happens to revenue as a consequence.

Would you have made this decision? Applied to any adjustment that represents discretionary spend removed rather than a cost that disappears at close. If the seller stopped spending on maintenance, marketing, or R&D and added it back, the question is whether you would run the business that way. Usually you would not, which means the money returns to the P&L in your first year.

The specific items

Recurring "one-time" anything. As above, and worth a separate mention because of how often it survives all the way to a signed purchase agreement.

Pro forma synergies from acquisitions not yet made. Common in roll-ups. The seller adds back the cost savings expected from add-ons in the pipeline. You are being asked to pay today for value you would create tomorrow with your own capital and your own execution.

Full-year effect of a cost reduction implemented in month ten. Legitimate in principle, abused in practice. Ask what actually happened in the two months after the cut. If the reduction was headcount, ask whether the roles were backfilled, whether contractors appeared, and what happened to whatever those people were doing.

Owner compensation adjusted to a market rate nobody has documented. The adjustment is fair; the market rate needs a source. Sellers routinely mark an owner-operator down to a $180,000 general manager salary for a role that in practice was three jobs. Ask what the replacement actually costs, including the second hire that will be needed.

Deferred maintenance and deferred capital expenditure treated as savings. Reduced spend in the two years before a sale is not margin improvement. It is a liability with a delayed invoice, and in asset-heavy businesses it is frequently the largest single item in the schedule with no line of its own.

Customer-specific losses removed as "non-core." A customer relationship that lost money is part of the business's history of pricing discipline. Removing the losers and keeping the winners produces a company that never existed.

COVID-era adjustments, in either direction, in 2026. Still appearing. Any adjustment that requires reasoning about 2020 and 2021 in a schedule presented today deserves a specific explanation of why it bears on forward earnings.

Stock compensation added back in a private company with a real equity program. If the business genuinely uses equity to retain the management team you are buying, that is a cost of retaining them and it does not disappear because it is non-cash.

Rent adjusted to market on a related-party property. Frequently legitimate, since the seller often owns the building through a separate entity at a non-market rate. It needs a broker opinion or comparable leases, and it needs to be checked in both directions — below-market rent understates true cost, above-market rent inflates the add-back.

Any adjustment described as "normalization" without a stated basis. The word is doing work that a number should be doing.

The questions to ask

For each line above a certain threshold — 2% of adjusted EBITDA is a reasonable cut:

1. Which general ledger accounts does this come from, and can we see the detail?
2. In which of the last five years does this item appear?
3. What specific event caused it, and what evidence exists that the event has concluded?
4. If this is a cost that was removed, when was it removed and what has happened since?
5. Who prepared this schedule — the company, the banker, or the quality-of-earnings provider?

The fifth question is the one people forget. An add-back schedule prepared by the seller's banker before a quality-of-earnings engagement began has a different standing from one that has already survived third-party scrutiny, and the two get presented in identical formats.

What the schedule tells you beyond the number

An add-back schedule is also a character reference.

A seller who presents twelve adjustments, each documented, each surviving the recurrence test, has told you something about how the business is run and how the next six weeks will go. A seller who presents thirty-one adjustments totalling 40% of reported EBITDA has told you something else, and the specific thing they have told you is that every other number in the room deserves the same treatment.

That inference is worth more than the adjustment analysis itself. It is the point at which a schedule stops being an arithmetic exercise and becomes a finding.