Earnings management · Fraudtracker

Earnings management — reverse-engineering forward projections

Companies guide forward revenue and earnings; the gap between what they project and what they actually deliver — combined with how they bridge it (non-GAAP addbacks, RPO bookings, stock-based comp, deferred revenue) — exposes whether the numbers are managed. This page reconstructs five reverse-engineering signatures from public 10-Q / 10-K disclosures and flags the names that show them most clearly.

Five signatures of forward-looking management

Guidance lowballing. Management guides 2–4% below internal forecast every quarter to manufacture "beats". Pattern: actual − guided midpoint is positive 90%+ of the time, never negative.
RPO inflation. Remaining Performance Obligations grow faster than revenue, particularly through non-cancellable multi-year contracts with anchor customers. RPO becomes a "future revenue" balance the market capitalizes today.
Non-GAAP / GAAP divergence. Non-GAAP "adjusted" EPS consistently exceeds GAAP EPS by 30%+ via stock-based compensation, restructuring "one-time" charges that recur every quarter, and amortization of acquired intangibles.
Deferred-revenue mechanics. Aggressive use of unbilled receivables and contract liabilities to smooth revenue across quarters. The asset side (unbilled AR) growing while revenue plateaus is the same signature as channel stuffing.
Period-extension / fiscal-calendar games. Lengthening fiscal quarters (e.g. 53-week years), changing segment definitions, restating comparable numbers — operationally rare, statistically suspicious when frequent.

Palantir (PLTR)

3 of 5 signatures

Trades at ~70× sales. Non-GAAP EPS > GAAP EPS by 35–60% per quarter on stock-based-comp addbacks.

Guidance midpoint vs. actual revenue ($M)

GAAP vs. non-GAAP EPS ($)

Reverse-engineering findings

Guidance lowballing: Beat midpoint guidance by 1–4% in 11 of last 12 quarters. Probability of that pattern under honest guidance ≈ 0.3%.
SBC at 20–28% of revenue: Stock-based comp is the single largest non-GAAP addback. Treating SBC as "non-cash" while issuing equity diluting existing holders is a paper distinction.
RPO concentration: Top 20 customers represent ~50% of US Commercial RPO. If any anchor (Anduril, Wendy's-style government deals) revisits, RPO drops materially.
What this doesn't show: The standard Beneish/Sloan ratios screen clean — accruals are low, CFO > NI. This is a multiple-and-quality-of-earnings issue, not an accruals issue.

Oracle (ORCL)

RPO inflation flagged

RPO jumped from $99B → $455B in two quarters on the OpenAI deal. The single largest forward booking-disclosure event in software-industry history.

Remaining Performance Obligations ($B)

RPO / trailing revenue (years of forward bookings)

Reverse-engineering findings

RPO/revenue ratio went from ~2× to ~9×. No software peer carries close to this ratio. Either Oracle disclosed a deal that historically would have been excluded as "not yet performance obligation" — or future revenue recognition is being booked aggressively early.
Counterparty risk on the RPO is concentrated. If OpenAI fails to fund the build, the $300B revaporates — and the GAAP RPO disclosure remains until performance is found impossible. That gap could last 4–8 quarters.
Capex tripled to $25B+/year. Cash going out doesn't show up in RPO at all; revenue conversion lags by 2–4 years. The cash-flow shortfall is real now.
Auditor scope: EY would normally require performance-obligation reasonable-assurance disclosure. The annual 10-K language on this RPO is worth re-reading carefully.

Salesforce (CRM)

Non-GAAP gap flagged

Acquisition-driven revenue smoothing; non-GAAP operating margin reported at ~32% vs. GAAP at ~17%.

GAAP vs. non-GAAP operating margin (%)

Acquisition revenue contribution (vs. organic)

Reverse-engineering findings

15-point gap between GAAP and non-GAAP margins. SBC alone is ~$3B/y. Amortization of acquired intangibles (Slack, Tableau, MuleSoft, Demandware) adds another ~$2B/y. Recurring "non-recurring" charges.
Slack acquisition revenue obscures organic deceleration. When acquisition tailwinds fade, true organic growth is in single digits.
Activist pressure changed the disclosure tempo. Following Elliott / Starboard / Inclusive 2023 campaigns, margin trajectory improved — but via SBC suppression and headcount cuts, not pricing power.

Snowflake (SNOW)

Guidance lowballing + SBC

RPO-focused disclosure narrative; 38–42% SBC as % of revenue.

Guided vs. actual product revenue ($M)

Stock-based comp as % of revenue

Reverse-engineering findings

Beat guidance by 0.5–3% in 14 consecutive quarters. Pattern is mechanical.
SBC ~40% of revenue. Highest among scaled (>$3B revenue) software. The non-GAAP path to "profitability" is purely the SBC addback.
Net Revenue Retention deceleration: NRR went from 178% (2022) to ~125% (2025). Management still presents the metric prominently; a leading indicator of saturation.

C3.ai (AI)

All 5 signatures

Persistent operating losses; revenue model changes 3 times in 4 years; large insider sales coincident with re-rating events.

Quarterly guidance vs. actual ($M)

GAAP loss vs. non-GAAP "income"

Reverse-engineering findings

Revenue recognition model changed in FY23 from subscription to consumption. The transition lets management defer/accelerate revenue across quarters with substantial latitude.
Federal/Defense revenue concentration: Department of Defense represents ~30% of revenue, contract structures opaque, lump-sum recognition common.
Non-GAAP swings between $0.20 EPS positive and $0.30 EPS negative quarter-to-quarter. SBC is a large addback; "other items" line moves materially.
Insider selling: CEO Tom Siebel has consistently sold into rallies. Not necessarily fraudulent but the pattern of guidance changes preceding sales is worth examining.

Tesla (TSLA)

Deferred-revenue + reg credits

FSD deferred revenue grows quarterly without proportionate FSD usage; regulatory credits booked as revenue when earned.

FSD deferred revenue balance ($B)

Regulatory credit revenue (~95% margin)

Reverse-engineering findings

FSD deferred revenue ≈ $4B. Recognition criteria depend on "feature delivery" — management discretion is wide. Each FSD price cut has triggered re-recognition of large deferred chunks.
Regulatory credits are essentially gross profit. When auto margins compress, credit revenue masks the trend. In Q1 2025 ~30% of net income came from regulatory credits.
Service-and-other revenue inclusive of supercharging: Mixing fast-growing (charging) and slow-growing (service) segments under one line obscures the underlying mix shift.

Summary: who's most exposed to a guidance-credibility reset

Company	Guidance pattern	RPO inflation	Non-GAAP gap	Other signature	Composite
C3.ai	Volatile	—	Large	Recognition changes	Severe
Palantir	Mechanical beats	Concentrated	Large (SBC)	Multiple priced	High
Snowflake	Mechanical beats	Yes	Very large (SBC)	NRR decel	High
Oracle	Stable	Extreme	Modest	Counterparty conc.	High
Salesforce	Stable	—	Large	Acquisition smoothing	Medium
Tesla	Volatile	—	Modest	Reg credits + FSD	Medium

None of the patterns above is illegal in itself. Each is an accounting choice the company is allowed to make under GAAP. The point is that they require active interpretation: a casual reader takes "non-GAAP earnings" or "RPO of $455B" as proxies for "real" — they aren't, and the market multiple is built on the optimistic interpretation.