A conviction score sums up, in a single number from 0 to 100, how far several families of data about a stock point the same way: results, price momentum, insider trades, quality of the accounts, valuation. At InvestIQ it is computed by a deterministic engine, with no artificial intelligence involved, and read as a rank among about 600 large stocks. It describes data on a given date; it is neither a price forecast nor a recommendation.
Why condense many data points into one number?
Analysing a stock by hand means cross-checking dozens of indicators: earnings surprises, analyst revisions, relative performance, insider purchases, Piotroski, Altman, P/E, discounted cash flows. One by one they tell partial, sometimes contradictory stories. A score weights them by rules set in advance and combines them into one reading.
The downside is real: a single number suggests a precision it does not have. A score of 73 and a score of 71 are not really different. A good score is always read together with what makes it up.
How is the InvestIQ score calculated?
The engine combines five groups, each scored from 0 to 100, with fixed weights:
| Group | Weight | What it looks at |
|---|---|---|
| Results | 28% | Earnings and revenue surprises, estimate revisions, the trend in analyst ratings; damped just before a results release |
| Price momentum | 22% | Returns over 3, 6 and 12 months (excluding the latest month), strength relative to the S&P 500, moving averages, RSI |
| Insiders and US Congress | 20% | Open-market purchases by executives, trades disclosed by members of Congress; neutral outside the US |
| Quality | 18% | Altman Z-Score, Beneish M-Score, capital efficiency (ROIC and free cash flow yield) |
| Valuation | 12% | Discounted-cash-flow value (or PEG for fast growers), P/E relative to the sector, EV/EBITDA |
The weighted sum is then multiplied by a factor tied to the Piotroski F-Score, from ×0.85 to ×1.08: quality therefore counts twice, as a group and as a filter. When a data point is missing, the factor concerned takes the neutral value 50 and a flag is recorded: the analysis never stops, but it then carries less information. All the rules are in the methodology.
A worked example: a fictitious company
Take Company H, a fictitious business, with these sub-scores:
| Group | Sub-score | Weight | Contribution |
|---|---|---|---|
| Results | 80 | 28% | 22.4 |
| Price momentum | 70 | 22% | 15.4 |
| Insiders and Congress | 50 | 20% | 10.0 |
| Quality | 75 | 18% | 13.5 |
| Valuation | 55 | 12% | 6.6 |
| Raw score | 67.9 | ||
| Piotroski factor (F-Score of 8) | ×1.08 | ||
| Final score | 73 |
The breakdown says more than the total. Results and momentum carry the score; the insider group sits at 50, neutral, for lack of executive purchases; valuation is middling. A 73 built this way does not mean the same as a 73 carried by valuation and quality.
How to read the score: a rank, not a verdict
A score only means something compared with others. InvestIQ rescores a universe of about 600 stocks in rotation every night (S&P 500, CAC 40, Nasdaq-100 names, a thematic selection) and places each score within it, in five bands: the top 10%, the next 15%, the middle half, the next 15%, the bottom 10%.
Suppose, for the example, that the top-10% line sits at 76 and the top-25% line at 70. At 73, Company H falls in the next 15%. To stop a stock flipping between bands on every small move, a boundary must be crossed by 2 points: if Company H had been in the top 10% the month before, it would only leave once it fell below 74.
The app shows the score itself ("Conviction 73") and attaches no instruction to buy or sell to it.
What a conviction score is not
- Not a forecast. A high score means the data is strong on the chosen factors, not that the price will rise.
- Not a recommendation. It is the same for every user and ignores each person's situation, horizon and tolerance for risk.
- Not a price target. It says nothing about a "fair" price.
- Not a sell signal when it is low. Some stocks are always in the bottom 10%: it means their data is weaker than the others' on that date.
The limits
The data looks backwards: published results, past prices, disclosed trades. The weights and thresholds are choices informed by published research, and InvestIQ publishes no back-tested performance for the score. Coverage is fullest for US companies; for European ones the insider group is neutral. And provider data can contain errors, which the score inherits.
How to use it
The score is a starting point for questions: which groups carry it, which hold it back, which missing-data flags come with it, what the opposing arguments say. In InvestIQ the AI explains these figures in plain language but computes none of them: the score always comes from the engine (see the editorial policy). To move from a reading to a decision, see buy, sell or hold: how to decide and how to analyse a stock.
In short
A conviction score condenses varied data into one readable number. InvestIQ's is deterministic, documented and read as a rank within a universe of stocks. Useful for sorting and for knowing where to look, it replaces neither an understanding of the business nor the investor's own decision.
Educational guidance only, not investment advice.
Tools and data for this topic
Frequently asked questions
A score is read first as a rank: at InvestIQ it is placed within a universe of about 600 stocks, in bands (top 10%, next 15%, middle half…). A high score means strong data on the chosen factors, not a predicted rise.
No. It is calculated the same way for everyone and ignores each person's situation, horizon and tolerance for risk. It describes data on a given date.
No, it summarises it. It helps to know where to look (which groups carry or hold back the score), but does not replace understanding the company, its sector and its risks.
At InvestIQ it forms a group (18% of the score) and also acts as a filter: the Piotroski F-Score multiplies the score by ×0.85 to ×1.08. Fragile accounts therefore weigh on the whole score without wiping it out.
