Quality of earnings analysis your deal team owns: every add-back tested, ruled and signed before the price is set
A first-cut QoE in three days, every add-back ruled and cited before the price is set.
See one case, screen by screen ↓A quality of earnings analysis, or QoE analysis, tests whether a target's adjusted EBITDA is real: it rebuilds reported earnings from the ledgers, rules on every add-back management claims, finds the adjustments management left out and sets the net working capital peg. Quality of Earnings Analysis is an AI quality of earnings review for private equity deal teams: seven agents read the data room in any language and test every add-back against 100% of its ledger entries. An associate proposes each ruling and the QoE lead signs it, so the deal partner and the deal model see only signed numbers.
AI in asset management: a buyer's guide to agentic AI for investment managers →
Updated
A quality of earnings analysis is how a buyer finds out whether management's adjusted, or normalized, EBITDA is real. Every leveraged buyout prices off that number, and it always starts as management's: reported earnings plus a list of “one-time” add-backs in the CIM's EBITDA bridge. Today the deal team waits three to six weeks for an adviser's databook, built on samples, while the LOI already sits on management's figure.
The risk sits in what a sample misses. A consulting program labeled one-time that has been billed every quarter for three years. Severance that a German subsidiary pays every year. A working capital peg set on the wrong months, or spare-parts inventory carved out as “non-operating”. Ledgers from foreign subsidiaries wait for translation, and the numbers move from the databook to the deal model by copy and paste. At 10x, each $1M of EBITDA that is not real costs about $10M of purchase price.
Estimated split for a mid-market buyout with three entities, by hand and with the solution.
Seven agents rebuild, test, check for recurrence and challenge every add-back; the associate proposes and the QoE lead signs.
Project Lantern: Quillfield Capital Partners IV is buying Halvorsen Industrial Services, an Ohio pump and compressor field-services business with subsidiaries in Germany and Mexico. LTM revenue to June 30, 2026 is $186.4M. Management's CIM shows adjusted EBITDA of $29.19M on 14 add-backs, and the LOI sits at 10.0x: $291.9M. The investment committee meets in 9 days. Here is the quality of earnings analysis, screen by screen, in the working solution.
Rachel's home shows four numbers: 5 of 14 rulings waiting for her signature, Lantern's running adjusted EBITDA of $24.6M, −$45.9M of value at 10.0x against the LOI, and 1 overdue request. Under “Needs you”, six items with one next step each: AB-02 ruled recurring, 4 accepted rulings to bulk-sign, an overdue Mexico trial balance, the seller's dispute on AB-07, the proposed peg and a report draft on another deal.
Report agent, 7:58 a.m.: “Overnight the Recurrence agent ruled AB-02 recurring (Carrowmore, 11 quarters), the Test agent vouched AB-09 to AB-12, and the seller answered on AB-07, which we re-tested.”
The Intake agent classified 3,412 data room files and found 125 of 126 monthly trial balances: Mexico's February 2025 is missing and was requested as RQ-034. It read the German books on DATEV SKR04 and the Mexican books on the SAT code list in the original and mapped 890 of 896 accounts. Six mappings under 85% confidence wait for Daniel Okafor to confirm, and a password-protected payroll file is waiting on the seller's data room administrator for its password (RQ-042).
The Intake agent never maps an account under 85% confidence and never fills a missing month; a person confirms each one.
The Rebuild agent rebuilt monthly EBITDA from January 2023 to June 2026 out of 2,840,000 ledger lines, translating the German and Mexican entities at monthly average rates. Reported LTM EBITDA is $18.19M. FY2025 ties to the audited statements from Kessling & Hart LLP within $12K, a post-closing audit entry, all 125 trial balances agree to the ledger, and the LTM ties exactly to the reported EBITDA on CIM page 47. Any P&L line drills to its ledger entries.
The Rebuild agent never changes a reported number; it shows every variance.
The docket lays out the EBITDA bridge like a court docket: each add-back with its claimed amount, the agents' ruling, the allowed amount, its sources, its status and the running adjusted EBITDA. Below sit five adjustments management did not propose, net −$1,170K: a June bill-and-hold cut-off, an under-accrued 2025 bonus, a missing inventory reserve in Mexico, below-market rent on the founder's building and a prior-year audit fee.
“Unsigned lines count at management's claim.”
AB-02 is $1.8M of “one-time strategic consulting” from Carrowmore Advisory Group, CIM page 47. Daniel asks in plain words, and the Recurrence agent answers with its plan: read account 6420 for all 42 months by vendor, match every Carrowmore invoice to its ledger entry, read the agreement and every statement of work, then the board minutes and next year's budget.
Recurrence agent, 96% confidence: “Carrowmore Advisory Group billed account 6420 in each of the last 11 quarters, $4.575M in all.”
The Test agent agrees all 12 LTM invoices, $1,800,000, to 12 ledger entries with no duplicates and no credit notes. Then the Challenger agent argues for the seller: statement of work #7 ends December 31, 2026, and the CIM says the program completes this year. The Recurrence agent answers that the master services agreement renews automatically each October 1, the board approved the program through FY2027 and the FY2027 budget carries $1.86M. The Challenger concurs.
Challenger agent: “If management produces a termination notice, the buyer's model can show it as a future saving, not as an add-back.”
Daniel proposes Rejected — recurring (RC-01), $0 allowed of $1,800K, and tags Rachel. She reads the evidence and signs with reason code RC-01 and a note. Adjusted EBITDA moves from $24.6M to $22.8M: $18M of value at 10x. Because Daniel proposed the ruling, he could never have signed it.
“Proposed by Daniel Okafor; you sign as checker. Adjusted EBITDA moves from $24.6M to $22.8M. The ruling, its evidence and your reason are logged and pushed to the deal model.”
Monthly working capital for 24 months on the draft purchase agreement's definition shows the seasonality: low in December to February, high in June to August, with closing expected December 15. Management's $29.8M leaves out $3.3M of contract spare parts as “non-operating”; the parts are bought for and billed to customers. The agent sets out three options and recommends the normalized twelve-month average, and lists $6.8M of debt-like items.
“Each $1 of working capital below the peg at closing reduces the price by $1.”
Karen Whitfield, Halvorsen's CFO, writes that 22 roles were eliminated in August 2026 and asks for the full $1.25M on AB-07. The Test agent reads the August 2026 payroll: the roles ended after the LTM, so the saving is real but pro forma. The Challenger concurs, and the recommendation is to keep the ruling and show a $1.25M run-rate saving in the deal model, with a draft reply ready for Daniel.
Price impact runs from the $291.9M LOI to $228.0M at signed adjusted EBITDA of $22.8M, −$63.9M, with a multiple sensitivity from 9.0x to 11.0x. AB-02, AB-07, AB-13, AB-01 and AB-08 move the price most. Until a ruling is signed, management's claim stands in these numbers, so nothing moves the price unseen. The Report agent drafts the QoE report and databook, every number cited, and pushes signed numbers to the deal model as a new version.
The Report agent releases signed numbers only. Nothing leaves the firm without the QoE lead.
The same deal, seen by the five people who carry it: what their diligence looked like, and what it looks like now.
Seven agents rebuild, test, check for recurrence and challenge every add-back; the associate proposes and the QoE lead signs.
Pulls the data room files, classifies them, reads trial balances, ledgers, invoices and contracts in any language, maps foreign charts of accounts such as DATEV SKR04 and the Mexican SAT chart, and checks completeness against the request list.
Rebuilds monthly EBITDA from the trial balances, translates foreign entities at monthly average rates, and ties the result to the audited statements and the CIM.
Vouches every EBITDA add-back against 100% of its ledger entries, invoices, contracts and payroll, computes the allowed amount, and finds the adjustments management did not propose.
Checks whether “one-time” costs repeat across the whole ledger: 42 months of the account, the vendor and similar narratives, then the contracts, the board minutes and next year's budget.
Builds monthly net working capital on the purchase agreement's definition, normalizes it, shows the seasonality, sets out peg options and lists debt-like items for the equity bridge.
Re-tests every ruling independently and argues management's side, drawing on 23 prior reviews; any disagreement goes to the associate and the QoE lead with both arguments shown.
Drafts the QoE databook and report from signed rulings, pushes signed numbers to the deal model as versions, and builds the evidence exports.
Signs each ruling with a reason code. The agents propose; a named person decides.
Ask in plain words, or tell the agents what to do. Every answer points to its source, and nothing changes until a person signs it.
Which add-backs repeat in earlier years?
Three of management's “one-time” add-backs repeat: AB-02 Carrowmore consulting, billed in each of the last 11 quarters (ledger 6420, $4.575M since Oct 2023); AB-08 German severance, paid in 2023, 2024 and 2025; AB-14 trade shows, every spring. I filtered the docket to the rejected and adjusted lines.
What is the price at signed numbers?
At 10.0x the signed basis gives an enterprise value of $228.0M against the $291.9M LOI. Debt-like items take another $6.8M off equity. I opened the price view; drafts never count.
Draft the IC summary of the QoE
Drafted from signed rulings only: reported EBITDA $18.19M, management's $29.19M, the five largest rulings, the $33.1M peg and $6.8M of debt-like items, each cited. Rachel Stein releases it to Marcus Hale.
Re-test the German severance using 36 months
Re-tested AB-08 on a 36-month look-back: severance before 2025 averaged €296K a year, so the normal level stays at about $320K and the one-time excess at $320K (Aufhebungsverträge, in German). No change; the re-run is logged.
13 screens from the working solution, on its sample data. Pick one to see it large.
The thresholds every quality of earnings review is held to are settings. A change goes in as a proposal, takes effect when the head of private equity operations approves it, and is logged.
| Setting | Default | Choose from |
|---|---|---|
| Materiality | $100K of EBITDA | $100K · $250K |
| Recurrence flag | Same vendor, account or narrative in 3 or more periods | Set per firm; each deal inherits it |
| Agent confidence | Under 85% goes to a person; never auto-ruled | Changed by proposal and approval |
| Bulk signing | Accepted rulings under $750K with Challenger concurrence | Changed by proposal and approval |
| Large rulings | Over $1M effect: Challenger concurrence or written override | Changed by proposal and approval |
| Peg default | Normalized twelve-month average | Normalized twelve-month average · close-month seasonal peg |
| Request SLA | 3 business days; overdue escalates to the seller's banker | Changed by proposal and approval |
| Sign-off SLA | 2 business days from proposal | Changed by proposal and approval |
Every figure is labelled: a target the solution is built to, an estimate, a typical published result, or a proven one.
“demo” = seen in the working solution, on its sample deal data · “target” = the design goal, measured in the live solution · “estimated” = our estimate · “typical” = published figures (S&P Global Ratings EBITDA add-back studies, 2018–2025 — add-backs lifted projected EBITDA by 45% on average in the 2015 cohort and were still 26% of projected EBITDA in 2023; adviser and broker publications on QoE timelines of three to six weeks; Cambridge Associates citing PitchBook LCD on buyout multiples; Journal of Accountancy on AICPA CS Section 100; $10M per $1M of EBITDA at 10x is illustrative arithmetic). Quillfield Capital, Project Lantern, Halvorsen Industrial Services and every person, company and figure named on this page are fictional — characters and sample data in the working solution.
A quality of earnings analysis, or QoE analysis, tests whether a target's adjusted EBITDA is real: it rebuilds reported earnings from the ledgers, rules on every add-back management claims, finds the adjustments management left out and sets the net working capital peg. Quality of Earnings Analysis is an AI quality of earnings review for private equity deal teams: seven agents read the data room in any language and test every add-back against 100% of its ledger entries. An associate proposes each ruling and the QoE lead signs it, so the deal partner and the deal model see only signed numbers.
An adviser's QoE report typically takes three to six weeks. The design goal here is a first-cut QoE in three days from the data room opening, so the deal team has its own view of adjusted EBITDA before the price is anchored. The adviser's final report can rely on the same evidence.
The Test agent vouches each add-back against every linked ledger entry, invoice, contract and payroll record, not a sample, and computes the allowed amount. The Recurrence agent reads 42 months of the account, the vendor and the narratives, plus the board minutes and next year's budget, to see whether a “one-time” cost repeats. The Challenger agent then argues management's side before a person decides. Owner normalizations, such as founder pay above market, are ruled against compensation benchmark data under reason code RC-07, market-rate normalization.
The Working Capital agent builds monthly net working capital on the draft purchase agreement's definition, normalizes it and shows the seasonality. It sets out the options, such as a normalized twelve-month average or a close-month seasonal peg, recommends one with its reasoning and lists debt-like items for the equity bridge. It proposes the peg; the QoE lead signs it.
Yes. Trial balances, ledgers and contracts of foreign subsidiaries are read in the original, including German books on DATEV SKR04 and Mexican books on the SAT chart. They are shown side by side with the English, every figure pinned to its page, and foreign entities are translated at monthly average rates in the rebuild.
No. This is a buyer's first-cut quality of earnings analysis and carries no opinion; the report draft says so. In the US a QoE is usually performed as a consulting engagement under AICPA CS Section 100, and the adviser's final report keeps its role.
Yes. The agents never sign, never change a reported number and never contact the seller. The associate proposes, the QoE lead signs with a reason code, the proposer can never sign their own ruling, and rulings over $1M of EBITDA effect need the Challenger's concurrence or a written override.
The Agentic Solution Engine builds and deploys it from your requirements and documents: your QoE methodology, reason codes, thresholds and the data room of a past deal. It goes live once every quality gate has passed. We will walk you through it on one of your own past deals first.
Every KPI cited to its page, every covenant tested the day the pack lands.
1-min demoPrivate Equity Deal ScreeningEvery CIM tested against the fund’s own limits in minutes, with every page cited.
1-min demoVirtual Data Room Due DiligenceEvery document read and every red flag cited to its page, before you sign.
1-min demoWe'll run Quality of Earnings Analysis on the data room of one of your past deals.