How Stock Sonar is scored

Every axis is a number from 0 to 100, and it is a percentile: 72 on Profitability means the company's returns are higher than 72% of the companies it is compared with. Higher is better on every axis, including Risk, where a high score means little risk showing. A score is a ranking of facts already filed. It is not a forecast, a price target or a recommendation, and a company can score well and still fall.

Its sector, from the SIC code the company itself files under at the SEC, grouped into the sectors investors use (carmakers apart from aircraft, drug makers apart from chemicals, REITs apart from shell vehicles). A sector comparison needs at least 30 peers with a value for that exact input; where a sector is thinner, the comparison is against every scored company instead, and the input says which one was used and how many peers it had. Two companies from different sectors are each ranked inside their own group, so comparing their shapes compares standing, not size.

Profitability: return on equity, on assets, on invested capital and on capital employed, from the last four quarters filed. Financial health: debt to equity (lower is better) and operating cash flow to debt. Growth: revenue against the same period a year earlier, and backlog (remaining performance obligations) where the company reports it. Insider conviction: net open-market buying over 180 days, with trades made under a pre-scheduled 10b5-1 plan left out where the filing flags them, and the number of distinct insiders buying. Institutional and government: funds adding in the latest 13F quarter, net Congress buyers over a year, federal contract awards over 180 days, and lobbying spend over four quarters. Risk and compliance: short interest as a percent of the free float, product recalls over three years (FDA, NHTSA, CPSC), EPA enforcement cases over five years, and the share of the Risk Factors section in the newest 10-K that is new text against the year before, all inverted. Innovation: patents, not loaded yet.

Each input is ranked on its own, then the axis is the plain average of the inputs the company has. Nothing is weighted: we have no evidence that one return measure deserves more weight than another, and a weight we chose would be an opinion wearing a number. Profitability needs at least two of its four inputs; every other axis needs one. Open any axis on a company page to see every input, its raw value, its percentile and the peer group used.

A company that does not report backlog is not the worst at backlog; it is unmeasured, and the axis says so instead of carrying a number. The radar is drawn only for companies scored on five or more axes, through the axes they have. Below that the page shows bars only, because a triangle of three scores looks like a chart and reads like a guess. About 3,400 of the 8,100 companies have a radar; most of the rest are funds, shells and recent listings with few filings.

A software company with zero recalls is not safer than a carmaker with two. It makes nothing that can be recalled. So a company with no events is scored on recalls or EPA cases only in a sector where at least one peer in ten has any; elsewhere the input is left out. A company that has events is always scored on them. Counts are raw, not scaled by revenue, so the largest makers in a sector carry more recalls than the small ones partly because they ship more; read the input next to the company's size. Recalls are attributed to the maker or importer, never the retailer that sold the product, and only where the name matches a listed company or a subsidiary we have checked by hand. Foreign makers listed only outside the US are not in the universe.

Scores are rebuilt once a night from the tables behind the rest of the site, so an insider purchase filed this afternoon moves the Insider axis tomorrow morning. The inputs move at their own pace: fundamentals with each 10-Q or 10-K, fund holdings quarterly and 45 days late by law, short interest twice a month, recalls and EPA cases weekly, lobbying quarterly.

In the Sonar screener each handle is a floor: drag Profitability to 70 and the list keeps companies scoring 70 or more. A floor is met only by a company scored on that axis. When few companies fit, the closest misses are listed with the axis that falls short in red. Compare lays two or three shapes over each other with every input side by side.

Not advice, not a rating of whether to buy, and not a model of the future. It tells you where a company stands today among its peers on facts that were filed with a regulator, and it shows every one of those facts so you can disagree with the summary.