NorrisAI Index · Earnings Quality

NorrisAI Accrual Screen

A deterministic, quarterly screen of US mid-cap companies for divergence between reported earnings and operating cash flow. Built entirely from SEC XBRL filings — no analyst judgement, no AI, no model. It is a starting point for research, not a verdict on any company.

23 of 591
US mid-caps ($2B–$10B market cap) — 3.9% — tripped 2 or more of the four cash/earnings-divergence screens.
Snapshot: 2026-09-02  ·  Universe frozen at pull time · not updated between snapshots

Coverage: 591 of 659 companies in the $2B–$10B band had a usable latest 10-K for all required XBRL concepts (net income, operating cash flow, revenue, total assets, receivables). 68 were dropped for incomplete tags and are excluded from the count above. Across the full 3884-name candidate pool the tag-completeness rate was 82.5%.

What this measures

When a company's reported profit consistently runs ahead of the cash its operations actually generate, the gap has to be explained — sometimes by ordinary timing, sometimes by aggressive accounting choices. This screen flags where that gap is unusually wide across four independent tests. It does not detect fraud and it is not an accusation. Many flagged companies have perfectly good explanations; many unflagged companies have real problems this screen cannot see.

The four screens

ScreenFormulaFlags whenTripped (591)
Sloan accrual ratio(Net income − Operating cash flow) / Total assetsratio > +0.1011
Cash conversionOperating cash flow / Net incomeratio < 0.50  (only when net income > 0)25
Share dilutionYoY change in shares outstandingincrease > 8%73
Receivables vs revenuereceivables growth − revenue growthgap > 25 pts  and receivables growing79

A company is counted in the headline only if it trips 2 or more screens and at least 3 of the four screens could be evaluated from its filings.

Companies that tripped 2+ screens — 2026 Q3

TickerCompanyAccrualOCF/NIDilutionRec−Rev gapFlags
AEHRAEHR TEST SYSTEMS+0.127+0.053+9.0%+145.3%4/4
AVAVAeroVironment Inc+0.008+0.256+10.8%+220.0%3/4
AAOIAPPLIED OPTOELECTRONICS, INC.-0.041+51.4%+111.6%2/3
BKDBrookdale Senior Living Inc.-0.059+18.8%+30.1%2/3
CELHCelsius Holdings, Inc.+0.017+0.623+9.3%+39.1%2/4
COGTCogent Biosciences, Inc.-0.041+42.6%+84.9%2/3
DNLIDenali Therapeutics Inc.+0.186+9.2%-204.5%2/3
DNOWDNOW Inc.+0.016+0.752+76.3%+151.9%2/4
GBTGGlobal Business Travel Group, Inc.-0.033+9.5%+28.5%2/3
GPORGULFPORT ENERGY CORP+0.247+0.492+3.7%-16.2%2/4
JBTMJBT MAREL Corp+0.040+0.432+0.2%+80.1%2/4
KTOSKRATOS DEFENSE & SECURITY SOLUTIONS, INC.-0.030+11.4%+54.3%2/3
LEUCENTRUS ENERGY CORP+0.031+0.108+8.0%2/3
ONDSOndas Inc.-0.004+341.8%+382.2%2/3
PRMPerimeter Solutions, Inc.+0.025+0.003-0.1%+25.5%2/4
SPBSpectrum Brands Holdings, Inc.+0.654-0.227-14.8%-39.9%2/4
TARSTarsus Pharmaceuticals, Inc.-0.033+10.9%+114.3%2/3
TXNMTXNM ENERGY INC-0.044+35.241+17.5%+30.2%2/4
UUUUENERGY FUELS INC+0.082-0.154+14.9%-209.8%2/4
VALValaris Ltd+0.113+0.309-2.5%-28.2%2/4
VIAVVIAVI SOLUTIONS INC.-0.053+10.5%+44.1%2/3
VSECVSE CORP+0.030-0.558+35.9%-5.3%2/4
VVVVALVOLINE INC+0.547-0.029-0.9%+5.6%2/4

Full universe of 591 companies, all four ratio values, flag counts and the SEC accession number used for each — download the CSV (latest.csv always points to the newest snapshot).

Worked example: the full breakdown of Valvoline (VVV) — one of the names flagged this quarter, run through all 15 research frameworks. Its cash/earnings gap traces to a 2023 divestiture, not an ongoing earnings-quality problem — a case study in how to read a screen output.

Methodology

Universe

Starting from the SEC's company_tickers.json (2026 Q3 pull), we keep US operating companies that file a Form 10-K, and exclude: foreign private issuers (entities filing only 20-F / 40-F / 6-K), and the financial sector by SIC code (6000–6799 — banks, brokers, insurers, real estate, REITs, and blank-check/SPAC shells at 6770). Market capitalisation is the 2026-09-02 closing price (Finnhub) × the latest dei:EntityCommonStockSharesOutstanding reported in XBRL, snapshotted on the same date, and the band is $2B–$10B. The universe list and snapshot date are frozen at pull time and do not change until the next quarterly run.

Period

Every figure is the latest available full fiscal year reported on a Form 10-K — not a trailing-twelve-month or 10-Q blend. Year-over-year screens use the prior fiscal year from the same or an earlier 10-K.

XBRL tag map

SEC bulk companyfacts.zip. Each concept tries the primary tag, then the fallbacks in order:

ConceptPrimary tagFallbacks
Net incomeus-gaap:NetIncomeLossProfitLoss; NetIncomeLossAvailableToCommonStockholdersBasic
Operating cash flowus-gaap:NetCashProvidedByUsedInOperatingActivities…ContinuingOperations
Revenueus-gaap:RevenuesRevenueFromContractWithCustomer(Excl/Incl)AssessedTax; SalesRevenueNet/GoodsNet/ServicesNet
Total assetsus-gaap:Assets
Receivablesus-gaap:AccountsReceivableNetCurrentReceivablesNetCurrent; AccountsAndOtherReceivablesNetCurrent; AccountsReceivableNet
Shares outstandingdei:EntityCommonStockSharesOutstandingus-gaap:CommonStockSharesOutstanding

A company with no usable value for any required concept after all fallbacks is dropped from the universe and counted in the coverage figure above — never silently excluded.

Thresholds

The four cutoffs (accrual > +0.10, OCF/NI < 0.50, dilution > 8%, receivables−revenue gap > 25 pts) are starting defaults, reviewed against the actual 2026 Q3 universe distribution so that the flag rate is neither trivially small nor a coin-flip. They are held fixed between snapshots.

Data sources

SEC EDGAR XBRL (company_tickers.json, bulk submissions.zip, bulk companyfacts.zip) for all financial data and classification; Finnhub for the closing price used in the market-cap filter only. No part of this pipeline calls an LLM.

Cite as

NorrisAI Accrual Screen, 2026 Q3. norrisai.us/index/earnings-quality/

Want a deeper, single-name analysis?

This index is pure ratio arithmetic on one fiscal year. The framework and scorecard go deeper on a single company — multi-year accrual trends, segment detail, and the qualitative reasons behind a wide gap.

Disclaimer: The NorrisAI Accrual Screen is an educational, quantitative screen — not investment advice, not a recommendation to buy or sell any security, and not a determination that any company has engaged in improper accounting or fraud. It reflects a single fiscal year of standardised XBRL data and simple ratios, which can be distorted by acquisitions, one-time items, accounting-standard changes, or tagging errors in the source filings. NorrisAI is not a registered investment adviser, broker-dealer, or fiduciary. Do your own research and consult a licensed professional before making investment decisions.
Educational screen only. Not investment advice. — NorrisAI AlphaLens