On October 1, 2026, the most important number in your deal gets calculated by someone you did not hire, working to a scope you did not set, on a timeline you do not control. On acquisitions at or above $3M, the lender commissions the Quality of Earnings report, and must then use that report's earnings figure to size the loan. A buyer-commissioned or sell-side report no longer satisfies the credit file. At the same time the historical coverage floor moves from 1.15x to 1.25x, and projections cannot be used to reach it. The rule applies to applications receiving an SBA loan number on or after October 1. Anything numbered through September 30 stays on SOP 8.0. If you have a file in flight, ask your lender when they expect the loan number, not when they expect to close. Partner buyouts and ESOP transactions are generally excluded from the QoE requirement, though not from the coverage test. The lender's analysis includes a cash proof reconciling bank activity to tax returns and reported income, plus customer concentration. Undocumented discretionary add-backs do not survive it. What a $100,000 Haircut Actually Costs: Take a business doing $1M of seller-adjusted EBITDA, priced at 5.7x, or $5.7M. A 90% senior loan of $5.13M on a ten-year 7(a) at 9.5% costs roughly $797k a year in debt service. Coverage lands at 1.26x, clearing the new floor by a hair. The lender's QoE rejects $100k of undocumented add-backs. Normalized EBITDA becomes $900k and coverage drops to 1.13x. Projections cannot bridge it. Work backwards. At $900k of EBITDA, the most debt service the deal can carry at 1.25x is $720k. Run that back through the same ten-year note and it supports a loan of about $4.64M, roughly $490k less than the $5.13M the bank was ready to write. That $490k has to come from somewhere. Either the buyer writes a bigger cheque, or the price comes down. Assuming the buyer is injecting 10% of the original price, i.e., $570k as equity, a $4.64M loan supports a purchase price of about $5.21M. The price falls by the same $490k the lender withdrew. One hundred thousand dollars of add-backs, roughly half a million of price. Buy the same business at 4x, or $4M, and the loan costs about $559k a year against $1M of EBITDA. That is 1.79x coverage. The QoE could reject $200k of add-backs and the deal would still clear 1.25x. The only difference between the two examples is the purchase multiple. At 4x there is enough headroom that a QoE adjustment changes nothing. At 5.7x the same adjustment moves the loan and the price. At 9.5% on ten-year money with 90% debt, the switch happens around five and a half times EBITDA, and higher rates push it lower, because a costlier loan leaves less room under the same 1.25x. So the question to ask before signing an LOI is not whether the seller's add-backs will survive. It is whether your multiple leaves room if they do not.