Stock screener app output looks complete, because it always looks complete. A list of tickers appears, and every company absent from that list reads as a company that failed the test. Some of them did fail. Others were never tested at all, because the app held no number for one of the rules, and a rule with no number behind it is not the same thing as a rule a company missed.
That distinction decides more of the output than most people expect. On the eight value screens covered below, 11,162 individual rule checks ran against a company for which the underlying figure was absent. Not one of them passed. Every one was recorded as a company that did not clear the rule.
A stock screener app is a hosted tool that applies a fixed set of numeric rules to a universe of companies and returns the ones that satisfy every rule. The screens described here run in a browser on any device, with nothing to install.
The per-rule figures in this article were computed by GuruScreens from the Pro audit data for the run dated 1 August 2026, which evaluated 3,787 US companies. Nothing here is investment advice, a recommendation, or a forecast. A full disclosure sits at the foot of this article.
Stock screener app coverage across eight value screens
Each screen below runs a fixed set of rules against the same 3,787 companies. "Fully testable" counts the companies that carried a value for every rule on that screen. "Rule checks with no value" counts the individual comparisons that ran with nothing behind them.
| Screen | Rules | Fully testable | Blank on at least one rule | Rule checks with no value | Clears every rule | Published list |
|---|---|---|---|---|---|---|
| Acquirers Multiple | 2 | 3,714 | 73 | 143 | 81 | 75 |
| Buffett Quality | 5 | 3,113 | 674 | 1,217 | 17 | 16 |
| Defensive Investor | 5 | 1,990 | 1,797 | 2,606 | 3 | 3 |
| Enterprising Investor | 5 | 1,990 | 1,797 | 2,606 | 108 | 102 |
| Graham Net-Net (NCAV) | 1 | 3,193 | 594 | 594 | 58 | 43 |
| Magic Formula | 2 | 3,262 | 525 | 586 | 9 | 8 |
| Neff Value | 4 | 1,827 | 1,960 | 2,489 | 10 | 10 |
| Quality at Reasonable Price | 6 | 3,363 | 424 | 921 | 6 | 5 |
The rule-check column sums to 11,162, and none of those checks passed.
Two columns in that table disagree on purpose. "Clears every rule" counts companies that pass each printed threshold in the audit record. "Published list" counts the names the screen actually shipped that month, which is the count the free screen pages report. On six of the eight screens the published list is the smaller number. GuruScreens does not document the reason for that gap, so this article reports both counts and states no cause for the difference. Analysis below uses the audit population, and prints the published figure alongside it.
What does a stock screener app do with a number it does not have?
A screening rule is a comparison. Return on equity at or above 15 percent needs a return on equity figure. Current ratio at or above 1.5 needs current assets and current liabilities. When the value is present the comparison returns true or false, and the result is meaningful either way.
When the value is absent, there is no third answer available. The comparison returns false, the company drops out, and the output records the same thing it records for a company that reported the number and missed it. Across the eight screens, 11,162 rule checks ran with no value behind them, and zero were marked as passing. That is the behavior GuruScreens has, and it is the sensible default, because passing a company on evidence you do not hold would be worse. It is also invisible in the finished list.
The practical effect is that the output reports two different things under one label:
- A verified rejection. The company reported the figure, the figure missed the threshold, and the exclusion is a real result.
- An unverified rejection. The app had no figure, so the rule could not be evaluated, and the exclusion is an artifact of coverage.
A list of tickers cannot tell you which is which. A per-rule audit can, if the tool prints the computed value next to the threshold rather than a checkmark.
Which denominator should a stock screener app report?
The tested universe is the right denominator when you are comparing how selective two strategies are. Screen pages usually print a headline of the form "8 stocks pass out of 3,787 evaluated". The second figure is the number of companies the app looked at. It is not the number of companies the app could rule on.
| Screen | Audit passers | Published list | Rate on the 3,787 base | Testable base | Rate on the testable base |
|---|---|---|---|---|---|
| Acquirers Multiple | 81 | 75 | 2.14% | 3,714 | 2.18% |
| Buffett Quality | 17 | 16 | 0.45% | 3,113 | 0.55% |
| Defensive Investor | 3 | 3 | 0.08% | 1,990 | 0.15% |
| Enterprising Investor | 108 | 102 | 2.85% | 1,990 | 5.43% |
| Graham Net-Net (NCAV) | 58 | 43 | 1.53% | 3,193 | 1.82% |
| Magic Formula | 9 | 8 | 0.24% | 3,262 | 0.28% |
| Neff Value | 10 | 10 | 0.26% | 1,827 | 0.55% |
| Quality at Reasonable Price | 6 | 5 | 0.16% | 3,363 | 0.18% |
Both rate columns count audit passers. On the published counts the rates are lower on the six screens where the two figures differ, and Enterprising Investor reads 2.69 percent rather than 2.85 percent on the 3,787 base.
On Acquirers Multiple the two rates barely differ, because 73 companies out of 3,787 carry a blank. On Enterprising Investor the pass rate almost doubles, from 2.85 percent to 5.43 percent, because 1,797 companies carried a blank on at least one rule. Same screen, same month, same engine. The only thing that changed is whether the denominator counts the companies the app could fully test or every company it loaded.
Neither rate is the correct one for every purpose. The larger base tells you how selective the screen is against the whole listed universe. The testable base tells you how selective the rules are, which is the question you are actually asking when you compare one strategy to another. An app that publishes only the first number is answering a question you probably did not ask.
Which screening metrics run out of data first?
Coverage is not evenly spread. The metrics below are the inputs behind the eight screens, ranked by how often the value was absent in the 1 August 2026 run.
| Metric | Companies with no value | Share of 3,787 |
|---|---|---|
| Dividend yield | 1,887 | 49.8% |
| Long-term debt to net current assets | 1,648 | 43.5% |
| NCAV to market cap | 594 | 15.7% |
| Current ratio | 545 | 14.4% |
| Return on invested capital | 513 | 13.5% |
| EPS growth, 5 years | 456 | 12.0% |
| EPS growth, long-run comparison | 413 | 10.9% |
| Revenue growth, 1 year | 286 | 7.6% |
| Net profit margin | 272 | 7.2% |
| Debt to equity | 207 | 5.5% |
| Return on equity | 75 | 2.0% |
| Earnings yield | 73 | 1.9% |
| P/E ratio | 73 | 1.9% |
| Market cap | 70 | 1.8% |
| Years of positive dividends | 0 | 0.0% |
| Years of positive earnings | 0 | 0.0% |
The split is not simply income statement against balance sheet. Price ratios and single-line income figures stay near complete, and so does debt to equity at 5.5 percent. What thins out is anything that needs net current assets derived from a classified balance sheet, or a multi-year history reconciled across restatements. Dividend yield sits at the top of the table for a different reason, covered in the next section. That ordering is worth carrying into any tool you evaluate, because it predicts which of your filters will quietly do the most damage.
Not every blank is a hidden passer
Before counting what the blanks cost, one correction matters more than the headline.
The GuruScreens Neff Value screen requires a dividend yield of at least 2 percent. Dividend yield is blank for 1,887 companies, the worst coverage of any metric in the set. If every blank were a hidden opportunity, that would be devastating.
It is not. Cross-checking those 1,887 companies against the dividend history recorded on the Defensive Investor screen, which carries a value for all 3,787, shows that 1,883 of them have zero years of dividend payments on file. For a non-payer, a blank yield behaves exactly like a zero, and a zero fails a 2 percent floor. The blank produces the right answer for the wrong reason.
This is the test that separates a coverage problem from a coverage complaint. Ask what a blank on a given metric usually means:
- A blank that encodes a real value. No dividend paid, so no yield to report. The exclusion is correct.
- A blank that encodes an unmet requirement. The company does not report the line item the calculation needs, or the calculation is undefined for its business model. The exclusion is a coverage artifact.
A stock screener app that shows the computed value lets you make that call. One that shows a list does not.
How many companies get excluded without a failing rule?
The sharpest version of the question is this: how many companies are excluded with no failing rule against them, purely because an input was missing? That group splits into two very different populations, and the split matters more than the total.
The counts below already apply the dividend adjustment from the previous section, which is why Neff Value reads 6 rather than its raw 156. No other screen changes, because none of the other seven depends on a metric where absence carries an economic meaning.
| Screen | Published list | Excluded only by a blank | Cleared at least one rule | No rule evaluated at all |
|---|---|---|---|---|
| Graham Net-Net (NCAV) | 43 | 594 | 0 | 594 |
| Enterprising Investor | 102 | 199 | 199 | 0 |
| Magic Formula | 8 | 166 | 105 | 61 |
| Buffett Quality | 16 | 80 | 17 | 63 |
| Acquirers Multiple | 75 | 72 | 2 | 70 |
| Quality at Reasonable Price | 5 | 58 | 0 | 58 |
| Defensive Investor | 3 | 8 | 8 | 0 |
| Neff Value | 10 | 6 | 6 | 0 |
The fourth column holds the companies that actually got partway to a verdict. They cleared at least one threshold on a real number and then hit a blank. The fifth column holds companies for which nothing at all was evaluated, so their absence carries no information about them either way. Quality at Reasonable Price and Graham Net-Net contribute nothing to the fourth column, and Acquirers Multiple contributes 2.
Graham Net-Net is the structural extreme. The screen has exactly one rule, so a company either has a net current asset value figure and gets a verdict, or has no figure and is dropped without one. There is no partial evidence to hold. The published list runs to 43 names while 594 companies were never tested. A further 15 companies clear the single printed rule in the audit record and do not appear on the published list, and GuruScreens does not document why.
Enterprising Investor is the row that matters. Every one of its 199 cleared at least one threshold on a real number, so the app publishes 102 names and holds partial evidence on 199 more that it never resolves.
The Magic Formula case: one missing metric, 104 companies
The Magic Formula screen runs two rules, an earnings yield above 10 percent and a return on invested capital above 20 percent. Return on invested capital is blank for 513 companies, and 166 companies are excluded from this screen only by a blank.
Those 166 split cleanly. 104 carry a value for earnings yield, clear that rule, and are missing only the return on invested capital figure. 61 are missing both values, so nothing about them was tested. The remaining 1 clears the return on invested capital rule and is missing the earnings yield figure. The 104 are the interesting group, because the app got halfway to a verdict and then stopped.
Those 104 are not a random sample. Checking every ticker against SEC company records:
- 95 carry an SEC Standard Industrial Classification code in the 6000 to 6799 finance range: state and national commercial banks, savings institutions, real estate investment trusts, insurers, personal credit institutions and mortgage bankers.
- 9 carry no SIC code in their EDGAR record, namely Barings BDC, Fidus Investment, Hercules Capital, Horizon Technology Finance, SLR Investment, TriplePoint Venture Growth BDC, Trinity Capital, MSC Income Fund and Chicago Atlantic BDC. Each of the nine has registered securities on Form N-2, the registration statement used by closed-end investment companies and business development companies, and all nine have also filed Form N-54A, the notification of election to be regulated as a business development company.
Not one of the 104 is an industrial, a retailer, or a technology company. That is not a coincidence. Return on invested capital divides operating profit by invested capital, and invested capital is not a comparable quantity across banks, insurers, mortgage trusts and lending vehicles, whose liabilities are the raw material of the business rather than a way of financing it.
Here is the part worth sitting with. Joel Greenblatt's published version of the Magic Formula excludes financial companies and utilities from the ranking, and the blanks here reproduce the financial company half of that exclusion. The mechanism does not match, because an exclusion that happens through a missing value is invisible, unversioned and impossible to audit, while an exclusion written into the rules can be read, questioned and changed. If the data source started supplying a return on invested capital figure for banks next month, 104 companies would quietly become eligible for a screen whose method never wanted them there.
That is the general shape of the risk. A screen that relies on missing data to enforce a boundary is one data update away from behaving differently, and nothing in the interface will tell you it changed.
The Enterprising Investor case: a rule banks cannot answer
The Enterprising Investor screen is the GuruScreens interpretation of the approach Benjamin Graham described for the more active investor. GuruScreens selected every threshold on it, including a current ratio of at least 1.5 and long-term debt held to no more than 110 percent of net current assets. Both rules read a classified balance sheet, which splits assets and liabilities into current and non-current.
Banks do not present a classified balance sheet. Their assets are not ordered by liquidity in that way, so a current ratio is undefined rather than merely unavailable. The data reflects that. Of the 199 companies excluded from this screen only by a blank, 189 carry an SEC Standard Industrial Classification code in the 6000 to 6799 finance range, and the largest single group is state commercial banks at 81 names, followed by national commercial banks at 46.
Graham described his approach for industrial companies and treated financial institutions separately. So again, a boundary the original approach intended is being enforced by silence rather than by a rule you can read.
The remaining 10 are ordinary operating businesses. The tenth is FirstCash Holdings, which lacks both balance-sheet figures. The other nine carry a current ratio and lack only the long-term-debt-to-net-current-assets figure: Argan, Badger Meter, Build-A-Bear Workshop, PC Connection, Exponent, IRADIMED, Monolithic Power Systems, NVE Corporation and Weyco Group. On each of the nine the screen records a computed value for its other four rules and no failure among them, which is what places them in this group. Whether any of the nine belongs on the list is a question the current output cannot answer, and that is precisely the point. They are named here as examples of missing inputs, not as candidates for anything, and the values the screen holds for them are the screen's own, not figures verified against the filings here.
How to audit a screener before you rely on it
None of this is unique to one product. Five checks separate a screener you can audit from one you have to trust, and you can run all five yourself in a few minutes.
| Check | What to look for | What a pass looks like |
|---|---|---|
| The threshold | The rule as the engine evaluates it | The metric and number appear before you pay |
| The computed value | The company figure beside the threshold | You see 14.08 percent against a 15 percent bar |
| The blank | A difference between missed and unavailable | The audit view separates the two states |
| The denominator | Which universe the pass rate is quoted against | The tool publishes both bases, or names one |
| The history | A prior month loaded as it stood then | The date selector returns the stored run |
This article does not score competing apps against those checks, because a review cannot audit what an interface does not show, and a scorecard built on guesswork would be worth less than the checks themselves. Each check takes a minute or two to run on a live product, so run them on the shortlist you already have.
One result is worth stating plainly here. GuruScreens does not pass check three. Its audit record marks an absent value as not passing and carries no separate marker to distinguish it from a value that missed. That is a real limitation of the product this article is published on, and it is stated here rather than left for you to discover.
Which kind of stock screener app fits your process?
Most of the disagreement about which app is best comes from readers wanting different things from the same word. Three different products are hiding inside it.
| If you want | The kind of app that fits | Where that sits in this family |
|---|---|---|
| Published criteria you can read, audit and compare over time | A fixed-rule screener with an evidence trail | The GuruScreens screens index |
| Filters described in plain language rather than built by hand | An AI-assisted screener | The MarketXLS AI stock screener |
| Complete ownership of the logic, and output that feeds your own model | A screen you build yourself in a workbook | The Excel walkthrough linked below |
A fixed-rule screener answers "what does this published method select right now, and why". A custom-filter or AI screener answers "what does my idea select right now". They are different questions, and the coverage problem in this article applies to all three, because every one of them resolves a comparison against data that is sometimes absent.
For a deeper look at how the per-rule record is structured on a fixed-rule screener, see the companion article on the pass and fail audit trail.
What the free tier shows, and where Pro begins
Every screen on GuruScreens is browsable without an account. The free tier opens all eight screens with their descriptions, every rule in the exact form the engine evaluates it, the published pass count, the first five names on each passing list, and the stored run history in the date selector. The screens index is the place to start, and it costs nothing to confirm the thresholds in this article for yourself.
The Pro plan at 99 dollars a year opens the full passing lists, the per-company audit that prints each computed value beside its threshold, the longer per-metric history, the complete backtest output, threshold customization so you can move a bar and see what happens to the list, and CSV and PDF export. The per-rule figures in this article came from that audit layer. Full details sit on the pricing page.
Prefer to build this in Excel?
If you would rather own the logic in a workbook than open a hosted tool, the MarketXLS team published a walkthrough for building a screen from scratch: how to build a custom stock screener in Excel. The coverage problem described here follows you into a spreadsheet, so the same discipline applies. Decide in advance what a blank cell means in each column, and make it visible rather than letting it fall through a comparison as a silent false.
Stock screener app FAQ
What should a stock screener app show besides the list of passing stocks?
It should show the rule in the form the engine evaluates, the company's computed value beside the threshold, and a marker for values that were not available. Without the third item, a coverage gap and a genuine failure look identical in the output.
Two apps run the same guru screen and return different lists. Which one is wrong?
Possibly neither. Two tools resolve the same rule against different data, and the differences compound: fiscal periods, handling of trailing twelve month figures, treatment of companies that do not report a line item, and the universe loaded in the first place. On the eight screens measured here, between 73 and 1,960 companies out of 3,787 lacked a value for at least one rule. That alone is enough to produce two defensible lists from one published method.
Does a missing value count as a pass or a fail?
On these eight screens it counts as a fail. Of 11,162 rule checks that ran with no value behind them, zero were recorded as passing. That is the sensible default, because passing a company on data you do not have would be worse. The problem is not the default, it is that the result is reported without a label.
Is a low pass count evidence that a screen is strict?
Not on its own. A screen can produce a short list because its thresholds are demanding, or because its inputs are thin. Enterprising Investor passes 2.85 percent of the loaded universe and 5.43 percent of the universe it could actually test. The first number reflects both effects mixed together, and only the second describes the rules.
Do I need to install anything to use a stock screener app?
GuruScreens runs in a browser on any device, so there is no download and no add-in to maintain. The screens rebuild monthly from SEC filing data and market data, and each rebuild is stored, so the work happens between visits rather than while you wait.
Can a stock screener app tell me what to buy?
No, and none of the figures in this article should be read that way. A screen sorts a universe by rules someone wrote down. It carries no view on valuation at today's price, business quality, or whether a method suits your situation. Everything named here is named as a data example.
The bottom line
The most useful thing a stock screener app can tell you is not which companies passed. It is which companies it could not fully evaluate, and why. On these eight screens that group runs from 73 companies to 1,960, it changes the pass rate on one screen from 2.85 percent to 5.43 percent, and on two screens it is quietly enforcing boundaries the original methods actually intended, through a mechanism no one wrote down and no one can audit.
You can read every rule and every threshold in this article without an account, on the GuruScreens screens index. The free screen pages report the published pass count, which is the Published list column above. The audit passer counts, the per-company values, the full lists and the backtest output come from the Pro plan described on the pricing page.
Disclosure
GuruScreens is an independent product of MarketXLS Limited. GuruScreens is not affiliated with, sponsored by, or endorsed by Warren Buffett, Benjamin Graham's estate, Joel Greenblatt, John Neff, Tobias Carlisle, or any firm or entity associated with them. All trademarks belong to their respective owners. Each screen is a GuruScreens interpretation of a publicly described investing method, not a rule set written or approved by the investor whose name it carries. Thresholds are chosen by GuruScreens.
All figures in this article describe the screening run dated 1 August 2026 and change with every monthly rebuild. Screening data is derived from company filings submitted to the SEC and from market data supplied by Fiscal.ai and QuoteMedia, and it may contain errors, omissions or timing differences relative to the source documents. Annual report figures are audited by the filer's auditor, quarterly figures are reviewed rather than audited, and neither is verified by GuruScreens. Company names and tickers appear solely as examples of data coverage, including the five shown in the screenshot at the top of this article, which are screen output rather than suggested purchases. Nothing in this article is investment advice, a recommendation to buy or sell any security, an offer, or a forecast. Past screening output is not a guide to future results.