What Is a Quality of Earnings Report? Caterpillar's Siwei

A quality of earnings report is an independent accountant's test, commissioned by a buyer, of whether a business's reported profits are real, likely to repeat, and actually collected in cash. It is not an audit. An audit asks whether last year's books followed the accounting rules; this report asks whether the profit a buyer is about to pay a multiple of will still be there the morning after closing.

Caterpillar learned the difference expensively. In 2012 it completed the purchase of ERA Mining Machinery, the Hong Kong-listed parent² of a Chinese maker of underground mining equipment known as Siwei, for roughly $650 million.¹ Months later, during integration, a physical count of inventory failed to match the books. The review that followed found costs allocated so that profit looked larger than the business could support, and sales recorded early — at times with nothing behind them. In early 2013 Caterpillar took a non-cash charge of about $580 million against the goodwill on the deal:³ most of what it had paid, gone inside roughly a year.⁴

Caterpillar was a disciplined industrial buyer with a large deal team, outside advisers and a listed, audited target. It still bought a set of numbers rather than a business. For founders redeploying exit proceeds into a second operating company, the lesson is less about accounting than about timing. The questions below are what a serious quality of earnings process should answer — and what it sounds like when the answer arrives too late to matter.

Has anyone counted what is actually on the floor — and on whose behalf?

A good answer names the date of the count, who stood in the building, which sites were sampled, and the gap between the ledger and the warehouse. It says plainly what that gap does to the price: a write-down now, or an open item still in play. It also asks whether ordinary operating costs were parked in inventory, which flatters margins until someone opens the door.

A weak answer treats inventory as settled because auditors signed the accounts, or schedules the count for after closing, as part of integration. Auditors do observe stock, but they sample for the seller's financial statements, not for the cheque a buyer is about to write. That order of operations is precisely what produced the Siwei charge. A missing pallet found while the price is open is a negotiation; the same pallet found afterwards is a write-off.

How old are the unpaid customer bills, and who owes the largest ones?

Siwei's sales had grown impressively, and much of that growth was financed by Siwei itself — generous terms extended to mine operators who were slow to pay. By the time the bid was launched, unpaid customer bills had stretched well beyond a year of normal terms, with the large majority already overdue. Receivables growing faster than sales is one of the oldest warning lights in accounting, because it means profit is being booked that nobody has collected.

A good answer provides an ageing schedule — what is under 30 days, 60, 90, past 180 — names the customers behind the biggest balances, and reconciles reported revenue to money that actually reached the bank. Calling a handful of those customers directly, the habit founders already use when they want the truth about a business, usually settles whether a balance is a debt or a story.

A weak answer quotes a single average for "days outstanding" and calls long terms industry practice. Averages hide piles nobody intends to pay. As the shelf has argued, booking a sale and collecting the cash are entirely different events — and only one of them belongs in a purchase price.

Which sales would fail if someone asked for a delivery record?

A good answer sets out the exact moment the business allows a sale to count as a sale, then tests the borderline cases: goods billed before they shipped, deposits recorded as revenue, long jobs booked as a percentage of completion, invoices with no proof of delivery. It then restates the year with the unsupported sales stripped out and shows the profit that survives.

A weak answer notes that the auditors accepted the revenue policy and concludes the policy is therefore fine. An audit opinion is a check on the system, not a verification that each material sale happened. Price is almost always a multiple of a profit figure; sales booked early inflate the very year that sets the multiple. Removing them is not pessimism — it is refusing to pay for work that was never done.

Which of these profits walk out the door with the seller?

Most sellers present an adjusted profit with one-off costs added back. Some adjustments are fair. A good report lists every one with a document behind it — a genuine one-time gain, rent paid to a family entity below market, a salary that will have to be replaced at a real rate, a contract that expires inside the year — accepts some, rejects others, and produces a normalised figure the accountants will put their name to. It is the same discipline the shelf applied to adjusted EBITDA.

A weak answer hands over the seller's own schedule of add-backs and labels the total "customary." Add-backs proposed by the person being paid are a claim, not a finding. Founders buying a smaller company tend to know the owner-perk category well, having lived it, and to underweight the other category: profit that was simply never there. Siwei's problem was the second kind.

What looks like an acquired asset but is really a future bill?

Working capital is the cash tied up in running the business day to day — inventory, unpaid customer bills, and the company's own unpaid suppliers. A good report sets a normal level of it, so a buyer does not pay extra for a balance sheet briefly flushed before sale, and then lists the debt-like items that are not called loans: customer deposits, accrued bonuses, unresolved tax exposures, maintenance the seller stopped doing on equipment that will need it.

A weak answer stops at "the company has no bank debt" and treats every receivable and every machine as pure value acquired. Earnings that look excellent on paper can still hand the next owner a stack of bills those earnings had already assumed away.

Who hired the accountants, and what were they not allowed to see?

A good answer names an independent firm engaged by the buyer, starting early enough that findings can still move terms, with a scope section honest to the point of discomfort: which subsidiaries were tested and which were not, whether bank statements came from the bank or from the seller, whether major customers were contacted, what the seller declined to provide. A refusal is itself a finding and should be written down as one.

A weak answer leans on a report the seller commissioned, or on the banker having "run the numbers," or promises the work can be finished in the fortnight before signing because the seller is in a hurry. Speed is often a negotiating tactic. A report of this kind issues no formal opinion and cannot guarantee it will catch concealed, coordinated misconduct — Caterpillar described exactly that. What it can do is move the count, the ageing and a sample of sales to the side of the table that can still change the price. Owners should read the scope section first; it tells them what the rest is worth.

If the books prove wrong, how much of the price is still unsent?

A good answer separates cash paid at closing from cash withheld — an escrow or holdback tied to specific figures such as receivables, inventory and the closing balance sheet — and names who controls release, what evidence triggers it, and how long the window stays open. That is different from an earnout, which is a bet on results not yet seen; a holdback is a reserve against books that were already wrong.

A weak answer treats a general warranty that the accounts are "true and fair" as though it were money. A promise is a lawsuit, and a cross-border lawsuit is not a refund. Caterpillar had deferred part of the consideration and later negotiated most of that remainder away⁵ — useful structure, but the cash already sent did not come back, and the write-off was never money returning.

---

The way to use this list is not to hand it to an adviser and wait. It is to hold whatever report arrives against these seven questions and notice which come back with a date, a document and a number that can still move the price — and which come back with an audit opinion. A reply that ends at the opinion is the reply already in hand when Siwei closed. The episode is a record of one purchase, not a view on Caterpillar; the instinct it tests is one most second-time owners already have.