A stock screener is a tool that filters a list of companies by rules you set. You might ask for companies with a market cap above a certain size, a P/E below a certain level, and profit that has grown in each of the last three years. The screener returns the companies that pass every rule. It turns a list of thousands of shares into a shortlist you can research.
A screener is a useful first step, but it is only as good as the rules you give it. Used carelessly, it produces a list that looks precise and means very little. This guide explains how to build a screen, what each kind of filter tells you, the traps that produce false results, and how to test a screen before you rely on it.
What a screener does
A screener reads data about each company and checks it against your conditions. The data usually includes price, market cap, trading volume, valuation ratios, profit, growth, debt and dividends. Each condition you add narrows the list. The result is a set of companies that meet all of your conditions at the time you run the screen.
A screener does not judge whether a company is good. It only checks the numbers you ask it to check. The judgement is in choosing the numbers and the thresholds, and in what you do with the list afterwards.
Step 1: say what you are looking for
Start with a question, not a filter. Examples include: which profitable companies in this market trade at a low valuation? Which mid-sized companies have grown profit steadily? Which shares have enough trading volume to buy and sell easily?
A clear question tells you which filters matter. A question about valuation needs valuation filters. A question about liquidity needs volume filters. Adding filters that do not serve the question makes the list smaller without making it more useful.
Step 2: choose a few filters
Most screens need only three to six filters. Here are the common ones, grouped by what they tell you.
Size. Market cap or average trading volume. These set the minimum liquidity and the kind of company you are looking at.
Valuation. P/E, price-to-book, or dividend yield. These compare the price with the business.
Quality. Return on equity, debt-to-equity, or the consistency of profit. These describe the business itself.
Growth. Growth in sales or profit over several years. Use multi-year growth, not one good year.
Momentum or price. The share's price relative to its moving average, or its change over the last year. These describe what the market has been doing.
Each filter has a threshold, and the threshold is a judgement. A P/E limit of 15 suits some sectors and rules out others. Set thresholds by looking at the spread of the data in the market you are screening, not by copying a number from an article.
Step 3: check the data behind the filters
A screener is only as accurate as its data. Before you trust a result, check the things that most often go wrong.
- Stale figures. Profit and debt figures come from the latest reports, and a company may have just reported something new that the screen does not yet show.
- Mixed periods. A P/E from trailing profit and a growth rate from forecasts can describe different time periods.
- Currency. Companies reporting in different currencies can give misleading size figures unless the screener converts them correctly.
- Share classes and float. A market cap that counts all shares can overstate the size of a company where most shares are held by a parent.
- Missing values. A company with no P/E, because it made a loss, can drop out of a screen or be treated as zero, depending on the tool.
Check that the screen does what you think it does. Run it on a company you know well and confirm the numbers match its published figures.
Step 4: read the results, not just the count
A screen that returns 40 companies is not a list of 40 buys. It is a list of 40 companies to research. Open each one and check the reason it passed. Did it pass because profit is strong, or because a one-off gain inflated it? Did the low valuation come from a rising profit or from a falling one?
Good screens often return companies with an explanation that does not hold up. A company with a very low P/E may have a profit that is about to fall. A company with strong growth may have growth that depends on one large contract. The screener does not check for these things, and you must.
The traps
Look-ahead bias. If you test a screen using data that was not known at the time, the results look better than they would have been. Always use the figures as they were published when the decision would have been made.
Survivorship bias. A screen tested only on companies that still exist misses the ones that failed. Failed companies often look good on past screens and bad in reality, so their absence flatters the result.
Overfitting. If you keep adjusting thresholds until the screen returns the companies you already like, the result is a description of your past bias, not a method. Decide the rules first, then test them.
Too many filters. Each filter removes companies. A screen with twelve filters may return two companies, and those two may have been selected by accident rather than by judgement.
Ignoring liquidity. A screen that finds cheap, profitable companies with almost no trading can return shares you cannot sell at a fair price. Add a volume filter.
How to test a screen
Testing a screen means checking whether its rules would have produced useful results in the past, using only the information available at the time. A simple test has four steps.
- Write the rules down, with the exact thresholds, before you look at the results.
- Run the screen on past data as it stood on a set date, and record the companies it returned.
- Hold those companies for a fixed period, such as one year, and record their returns against a broad index for the same period.
- Repeat on several dates, and include the costs of buying and selling.
A screen that beats the index consistently across different dates, after costs, has some evidence behind it. A screen that beats the index in one period only is most likely a description of that period. Be honest about the sample size: a few dates and a few dozen companies prove little.
A worked example
Suppose you screen for companies with a market cap above 500 million, a P/E below 12, profit that grew in each of the last three years, and a debt-to-equity ratio below 1. The screen returns 14 companies.
You open each one. Four have a low P/E because one large profit came from the sale of a property, not from the core business. Two have profit growth that reverses in the latest quarter. Two have debt that is near the limit and rising. The remaining six pass a closer look. The screen has not chosen your portfolio. It has saved you time by narrowing the list to 14, then your research narrowed it to six.
Checklist for building a screen
- Write the question the screen should answer.
- Choose three to six filters that serve that question.
- Set thresholds from the spread of the data in the market you are screening.
- Check the data for stale figures, mixed periods, currency and missing values.
- Read each result and check the reason it passed.
- Test the rules on past data, without looking ahead, and include costs.
Key takeaways
A stock screener filters companies by rules you set, and it turns a large list into a shortlist. Build it from a clear question, a few filters that serve it, and thresholds set from the data. Check the data behind the filters, read each result rather than counting it, and test the rules on past data without looking ahead or survivorship. A screen is a way to choose what to research, not a list of companies to buy.
This guide is for general information, not financial advice. See the disclaimer.