How the InvestIQ Quant Engine turns a company’s published data into one score from 0 to 100, what that score measures, and where it stops. The engine is deterministic code: the same data always gives the same score, and no AI model computes any part of it.
1What the score measures
The score summarises how a stock’s latest data stands on five groups of factors that academic and practitioner research has linked to future returns: earnings momentum, price momentum, informed buying, financial quality and valuation. A high score means the data is strong on those factors at the time of calculation. It does not mean the price will rise, and it says nothing about whether the stock suits you.
Each factor is first turned into a sub-score from 0 to 100 with fixed thresholds, 50 being neutral; the sub-scores are then combined with fixed weights. Every weight and threshold is written in the engine’s code, and none is set by hand for a particular stock.
2The five factor groups
| Group | Weight | What it measures |
|---|---|---|
| Earnings | 28% | Earnings and revenue surprises over the last four quarters (44% of the group), revisions to analysts’ earnings estimates (39%) and the trend in analyst ratings (17%). The group is damped ahead of a results release: ×0.95 within 30 days, ×0.85 within 14, ×0.70 within 4 and ×0.50 on the day itself until the results are out. |
| Price momentum | 22% | Returns over 3 and 6 months and over 12 months excluding the latest month (30%); strength relative to the S&P 500 (25%); trend confirmation from the 50- and 200-day moving averages and the ADX (25%); the RSI, which scores best between 55 and 70 — firm without being overheated (20%). |
| Insider and congressional trades | 20% | Open-market purchases by company insiders (60%): share grants and option exercises count as neutral, sales weigh a quarter of purchases and never push the score below neutral on their own, and the signal is halved for companies above $200bn and strengthened for those below $10bn. Trades disclosed by members of the US Congress (25%). The remaining 15% is neutral. Outside the US, where these filings do not exist, the whole group is neutral (50). |
| Quality | 18% | Altman Z-Score, a measure of financial distress (35%); Beneish M-Score, a screen for earnings manipulation (25%); capital efficiency — return on invested capital and free-cash-flow yield (40%). Banks, insurers and real-estate trusts use adapted measures (return on equity, margins, stability or leverage), because the standard ratios misread their balance sheets. |
| Valuation | 12% | Upside to the discounted-cash-flow value published by our data provider (40%), replaced by the PEG ratio for fast growers (revenue growth above 30% or a P/E above 50); P/E relative to the sector (35%); EV/EBITDA (25%). |
The weighted sum is the raw score. The quality adjustment below turns it into the final score.
3The quality adjustment (Piotroski)
The raw score is multiplied by a factor that depends on the company’s Piotroski F-Score, a 0–9 checklist of profitability, balance-sheet and efficiency tests (Piotroski, 2000). It works as a gate on the whole score, not as a sixth factor:
| F-Score | Multiplier |
|---|---|
| 0–1 | ×0.85 |
| 2 | ×0.88 |
| 3 | ×0.92 |
| 4 | ×0.96 |
| 5–6 | ×1.00 |
| 7 | ×1.04 |
| 8–9 | ×1.08 |
The multiplier is 1.00 when the F-Score is unavailable, and for banks, insurers and real-estate trusts, for which the test was not designed. A low F-Score does not penalise a company whose revenue grew by more than 35% in a year: heavy reinvestment mechanically weakens the ratios the test reads. The final score is kept between 0 and 100.
4A rank within the scored universe
A score on its own says little; its rank says more. InvestIQ keeps a universe of about 600 stocks — the S&P 500, the CAC 40, Nasdaq-100 companies outside the S&P 500 and a selection of thematic stocks, some listed in London and Frankfurt. A scheduled job rescores it in rotation every night, so that no score in it is more than a week old. A stock you open in Stock Analysis is scored at that moment, with current data, and compared with the same universe.
The app shows where a score falls in that universe, in five bands: the top 10%, the next 15%, the middle half, the next 15% and the bottom 10%. To change band, a score must cross the boundary by 2 points, so that small day-to-day moves do not make a stock flip back and forth. If fewer than 100 stocks have a current score, fixed thresholds (75, 65, 40 and 25) are used instead.
A rank compares a stock with the others on a given date. Some stocks are always in the bottom 10%: that position says their data is weaker than the rest of the universe, not that their price will fall.
5Missing or partial data
When a source fails or has no value for a company, the factor concerned takes the neutral value 50 and the analysis records a flag, shown in the app. An analysis never fails because one source is missing, but a score built on many neutral factors carries less information: check the flags.
Coverage is fullest for US companies. For European companies, insider and congressional filings do not exist, so that group is neutral, and some financial statements are less detailed.
6What the AI does — and does not do
The score, the sub-scores, the ratios and the prices come from the engine and from the data, never from a language model. AI models read these computed results and write the explanations you see: research modules, Copilot replies, Deep Research reports and War Room debates. How they are instructed and checked is set out in the editorial policy; where the data comes from, in data sources.
7Limits
- The data looks backwards: published results, past prices, disclosed trades. Nothing in the score anticipates news that has not happened yet.
- Weights and thresholds are fixed choices informed by published research. We publish no back-tested performance for the score.
- Data can contain errors or arrive late, and the score inherits them.
- A score ignores your objectives, horizon, tax situation and tolerance for risk. See the disclaimer.
8Changes and references
When the method changes, this page changes with it: the date and the engine version at the top are updated.
The factors rest on published research, among others:
- Piotroski, J. (2000). “Value Investing: The Use of Historical Financial Statement Information to Separate Winners from Losers.” Journal of Accounting Research, 38.
- Altman, E. (1968). “Financial Ratios, Discriminant Analysis and the Prediction of Corporate Bankruptcy.” Journal of Finance, 23(4).
- Beneish, M. (1999). “The Detection of Earnings Manipulation.” Financial Analysts Journal, 55(5).
- Jegadeesh, N. and Titman, S. (1993). “Returns to Buying Winners and Selling Losers.” Journal of Finance, 48(1).
- Bernard, V. and Thomas, J. (1989). “Post-Earnings-Announcement Drift: Delayed Price Response or Risk Premium?” Journal of Accounting Research, 27.
On our blog: the Piotroski score, the Altman Z-Score, momentum, RSI and moving averages, insider trades, DCF valuation.