A turnaround too good to be true? Sunbeam's 1997 accounts

• Profit against cash. Earnings from continuing operations of 123.1 million against operating cash flow of −8.2 million: in the best year in the company's history, the business used cash.
• Where the cash went. Receivables rose 84.6 million and inventories 100.8 million (by the cash flow statement). Receivables grew 38.5% and inventories 57.9%, against sales growth of 18.7%.
• Receivable and stock days (year-end balances): receivables 92 days of sales against 79; inventory 112 days of cost of sales against 66 (about 73 if 1996 cost of sales excludes the special charges, using the 17.9% gross margin).
• The restructuring accrual fell from 63.8 to 10.9 million, with 43.4 million used during the year: 1997 costs were being met from provisions made in 1996, and about 9.5 million of the fall isn't explained by the cash used (a release or a reclassification: a question for management).
• The 1996 base. A 337.6 million charge in the year before a new management's first full year, followed by a jump in gross margin from 17.9% (excluding charges) to 28.3%: the pattern of a big bath, which can flatter the following year if any of the charge was excessive.
• Bill-and-hold and early buys, disclosed but described as limited; with receivables and stock both growing much faster than sales, the question is how much 1997 revenue was pulled from 1998.

How it's built. Start where the profit and the cash part company, then find which lines carry the gap, in days, not just money. Read the revenue policy and risk factors for anything that changes when sales are booked. Look at the base year: a large charge in t−1 changes everything that follows.

Workings: DSRI = (295,550 ÷ 1,168,182) ÷ (213,438 ÷ 984,236) = 1.17; accruals (Sloan's balance-sheet method, with the cash flow statement's change in taxes payable because the balance isn't shown) = (33,842 − 40,852) − (−73,484 + 253 − 52,844) − 38,577 = 80,488; TATA = 80,488 ÷ 1,120,284 = 0.072.

On the reported figures, M = −2.04: below Beneish's −1.78 cut-off. The screen as specified doesn't flag Sunbeam. Two inputs are worth questioning, and an analyst might reasonably adjust them; the table shows each adjustment on its own and together, so you can judge them yourself:

• The balance-sheet accruals are understated because 1996 current assets include 102.8 million of businesses held for sale, which left the balance sheet in 1997 (the measurement problem described by Hribar and Collins, Journal of Accounting Research 40(1), 2002). Using cash flow accruals instead: (123,128 + 8,249) ÷ 1,120,284 = 0.117.
• The 1996 special charges make 1997's margins look far better than the trend (GMI of 0.30 says margins more than tripled). Using the company's own figures excluding the charges is the fairer comparison.

How it's built. Compute the score as specified, then ask whether each input measures what it's meant to. Here the big bath in the base year and the divestiture both push the score down. The adjusted figures are one analyst's reasonable view, not a result: other analysts might adjust differently, and the model was estimated on unadjusted data. The lesson isn't "adjust until it flags"; it's that the parts (DSRI 1.17, SGI 1.19, cash flow far below profit) were telling the story whatever the total said.

Sunbeam: record profits not yet backed by cash · Source: Form 10-K405 for 1997 (filed 6 March 1998)

The 1997 turnaround is not yet visible in cash, and parts of it may have been borrowed from 1996 and from 1998. Earnings from continuing operations were 123 million, but operations used 8 million of cash, because receivables grew 39% and inventories 58% against sales growth of 19%. Receivables now equal 92 days of sales (79 a year ago) and stock 112 days of cost of sales (66).

Three questions for management: (1) How much of 1997's sales came from early-buy, bill-and-hold and extended-credit programmes, by quarter, and what does that imply for first-quarter 1998 orders? (2) How much of 1997's margin improvement came from costs charged against 1996 restructuring and other provisions (43 million of the restructuring accrual was used in the year), and are any provisions being released into profit? (3) What is the ageing of receivables, and how much is from the early-buy programme's extended terms?

The Beneish M-score is −2.04 on the reported figures, below the −1.78 cut-off. In my view two inputs understate the risk (the 1996 special charges and a divestiture in the base year); adjusting for both gives −1.67, just above the cut-off. Either way the score is borderline, and its main drivers (receivables, growth, accruals) match the signals above.

What would change our view: operating cash flow turning strongly positive in the first half of 1998 without a fall in sales; receivable and stock days returning to 1996 levels; or management showing that 1997 provisions and incentives were small. The opposite would mean 1997 profits overstated the business's real earning power.

These are the SEC's findings and allegations; Sunbeam and the individuals named settled without admitting or denying them.

• The order against Sunbeam (Securities Act Release 7976; AAER 1393, 15 May 2001) found that management created at least 35 million of improper restructuring and "cookie jar" reserves at year-end 1996, which were released into income in 1997; that it used guaranteed sales, improper bill-and-hold sales and other practices; and that at least 62 million of the reported 189 million of 1997 income from continuing operations before taxes "came from accounting fraud". It also found undisclosed channel stuffing, and that the 10-K's description of bill-and-hold sales was misleading: fourth-quarter 1997 bill-and-hold sales added 29 million of sales and 4.5 million of income, about 10% of that quarter's sales.
• Restatement. In November 1998 Sunbeam restated its results from the fourth quarter of 1996 to the first quarter of 1998; restated 1997 income was about 93 million, roughly half the amount first reported. The share price fell from about 52 dollars in March 1998 to about 7. Sunbeam filed for Chapter 11 on 5 February 2001.
• The individuals. The SEC's complaint (Litigation Release 17001, 15 May 2001) charged the former CEO, CFO, controller, two vice-presidents and the Arthur Andersen engagement partner. The former CEO and CFO consented, without admitting or denying the allegations, to injunctions, permanent bars from serving as officers or directors of public companies, and civil penalties of 500,000 and 200,000 dollars (Litigation Release 17710). Settlements with the other defendants followed; the last, in January 2003, included an order against the audit partner finding improper professional conduct (Litigation Release 17952).

How it's built. Keep what was knowable then separate from what was found later, and attribute every later finding to its source and its settlement terms. Write "the SEC found" or "alleged", never "committed fraud" as your own statement.

Some of the same signals, but not the same evidence. Like Sunbeam in 1997, Super Micro reports rapid growth with profit far ahead of operating cash, and receivables and stock rising faster than sales; unlike Sunbeam, its sales grew 78% and its 10-K describes "supply constraints for certain components, including memory and storage, as well as GPU and CPU availability", which makes building stock understandable, its current deferred revenue rose 1.2 billion, and its 2026 10-K reports no finding of improper revenue or reserves. Its history does call for more scrutiny: in 2020 it reached a consent settlement with the SEC over its financial accounting and disclosures for fiscal 2014–2017, paying a 17.5 million dollar penalty (Form 8-K, 25 August 2020), and in 2024 its previous auditor resigned citing governance and internal-control concerns. That history is a reason to check harder, not evidence about 2026. A high M-score is a reason to keep checking the parts (receivable and stock days, write-downs, deferred revenue), not a conclusion; resemblance to a known case isn't evidence. What would change my mind: evidence in the notes or later filings of revenue booked before control passed, of returns or incentives growing, or of cash failing to follow profit once the stock ships.

What a reviewer would flag: treating pattern-matching to a famous fraud as proof; ignoring the explanations the company gives; or, the opposite, ignoring the signals because the company explains them.