Summarize with AI
Good SaaS features section design shows three cropped screens from the real product, each with a one-line caption that names a job the buyer would pay for. The usual row of icon cards with labels like Real-time sync and Smart forecasting gets skipped, because those words fit every product in your category. If you cannot crop a real screen for a feature, cut that card instead of writing a vaguer one.
This block matters more than its size suggests. It is usually the first thing a visitor sees after deciding to scroll, so it reaches the people who were already curious. You know exactly what your product does on a Tuesday afternoon. The features section is where a stranger either sees that too, or reads three nouns and leaves.
Real-time sync, Smart forecasting, Multi-location ready
Picture an AI tool for restaurant groups that run four to twenty kitchens. This is an example, not a client. Every afternoon it reads each location's sales history item by item, tomorrow's reservations and the weather forecast. Then it writes a prep list for every station, so the grill cook knows to braise 38 portions of short rib instead of guessing 50 and throwing a dozen away.
Under the hero sits the features section. Its heading reads Everything your kitchens need. Below are three white cards with thin line icons. A circular arrow over Real-time sync, with Your data, always up to date. A bar chart over Smart forecasting, with Predictions you can trust. A map pin over Multi-location ready, with Built to grow with your group. Each card ends in a Learn more link, and all three links open the same features page.
Nothing on those cards is false. That is the problem. A scheduling app, an inventory app and a payroll app could all ship the same three cards tomorrow. The words prep, portion, station and waste do not appear anywhere in the section. A chef who runs three kitchens reads it and learns that the product has data and uses it.
Why the card row is the block people skip
The usual advice says the cards are fine and the copy is weak, so rewrite each line as a benefit. We think that misses it. An icon, a two word label and one line under it is the shape of a template, and visitors have seen thousands of them. They read the shape before they read the words, and the shape says nothing is here. Better copy in the same box still gets the same glance.
Watch what the founder does on a demo call. Within the first minute he shares his screen and opens tomorrow's prep list for one kitchen. The prospect asks which items it covers and whether it knows about the patio. That screen does the selling, and it is nowhere on the homepage. The three cards came with the site template, and he filled them in on the afternoon the site went live.
He has rewritten the hero headline twice since then. He has not touched the cards, because he reads them knowing what sits behind each word. When he sees Smart forecasting, he pictures the model he spent eighteen months tuning. A visitor sees two words.
The rewrite: three crops, each with one job
Change the section heading first. Everything your kitchens need becomes Tomorrow's prep list, written before lunch service ends. Then swap each card for a crop of a real screen, shown large enough to read, with one caption under it.
Crop one is the grill station's list for the Austin kitchen on a Tuesday. Short rib, 38 portions. Chicken thigh, 52. Charred broccolini, 4.5 kg. Next to the short rib sits a small note: rain after 6pm, patio closed, down 6. The caption reads: Writes each station's prep list for tomorrow, item by item, by 3pm.
Crop two is the closing log from a Friday. One line reads Short rib 86'd at 8:40pm. Beside it, next Friday's list already shows Short rib 44, with a tag that says plus 6, sold out last Friday. The caption reads: Adds to next week's count when you run out, and trims it when the bin fills up.
Crop three is the group view for an operations lead. Five kitchens in rows, with a column called Prepped, then binned, last week. The South Lamar row is highlighted at 14 kg. The caption reads: Shows which kitchen over-prepped last week, and by how much.
Real-time sync is not gone. It becomes one plain line under the three crops: updates from your point-of-sale system every fifteen minutes. That is a fact a buyer can check, and it no longer needs an icon to hold it up.
Name the job, not the attribute
Fast, smart, secure and real-time are attributes. They describe a quality the product has. A job is something the product does for one person on one day, and it has an object the buyer recognises: a prep list, a closing log, a kitchen. Attributes ask the visitor to imagine your product. Jobs show it to them.
Two tests catch most attribute captions. First, the neighbour test: put the caption under a competitor's screenshot. If it still fits, it is not yours yet. Second, the noun test: the caption should contain at least one word your buyer says out loud at work, like portion or station, and no word only your team uses, like signal or pipeline.
The crop has to prove the caption. If the caption says by 3pm, the screen should show a time stamp. If it says by how much, the screen should show a number. Our guide to product screenshots on SaaS websites covers how tight to crop and how to keep the screens current once the product changes.
You can see the same idea on the Sawdust website we designed for a music licensing catalog. Instead of a card that promises a big library, the catalog section shows one real track card, Slow Burn, with its cover, a preview window from 00:15 to 00:50, 88 BPM and the tags Cinematic and Slow. A brand can listen before contacting anyone. The rights are shown the same way, as six named layers, rather than a badge that says fully cleared.
How many features before it turns into a list
Three is right for most homepages. Pick them by asking which screens you open first on demo calls, in that order. A fourth earns a place only when it answers the objection that stops a deal, such as what happens when the forecast is wrong. Past four, a visitor stops reading each one and starts scanning, and the section becomes the list you were trying to escape.
Layout follows from that. Do not force three crops into equal boxes in a grid. Give the one that sells hardest the most width, then stack the other two beside or below it. Alternating rows, crop on one side and caption on the other, also work well because each screen gets room to be read. On a phone, put the crop first and the caption under it, and make sure the numbers stay legible at that width.
Everything else belongs on a features page, in a comparison table, or in the docs. Integrations, permissions, export formats and admin settings are real reasons to buy, but they are second visit questions. The visitor who needs them will look for them.
When to cut the section instead
Sometimes the right features section is no features section. Cut it when the product is still on a waitlist and there is no screen to show yet. One honest line about what you are building beats three cards about a product nobody can see. Cut it too when your hero screenshot and your how it works block already show the same moments, because a third pass over the same screens reads as padding.
Security is the card founders most often keep out of fear. Enterprise-grade security in a card convinces no one who cares about it. Move it to a single line near the main button, such as encrypted at rest, with a link to a security page that lists the actual controls. The buyers who need it will click, and everyone else keeps moving.
What a skipped block costs a funded team
Every visitor who reaches the features section was paid for, in ads, in content or in hours of the founder's own posting. The ones who scroll past the hero are the curious ones. If the next block tells them nothing, that spend is gone, and more traffic only makes the loss larger at the same rate.
For a sales-led tool the cost shows up on the calendar. Prospects book a demo to find out what the page should have told them, and the founder spends the first ten minutes of each call on it. Many of those calls end there. At the stage where the seed bar sits around 300 to 500 thousand in ARR for SaaS companies, and the all-sector median gap from seed to Series A is around 616 days, those hours come straight out of the runway. For the whole page, our SaaS landing page guide covers the sections around this one.
Rebuild your features section this week
- Open your homepage on a phone and scroll to the features section. Write down the three labels exactly as they appear.
- Run the neighbour test on each one. Could it sit on a competitor's page unchanged? Mark every one that could.
- List the screens you open on demo calls, in order. The first three are your new features section.
- Crop each screen to the single panel that proves the job. Use real data, and check every number reads at phone width.
- Write one caption per crop that starts with a verb and names an object your buyer says at work.
- Rename the section heading to the job, not the product. Move security and integrations to one line each, with links.
- If you cannot find three screens worth showing, cut the section and let the page get shorter.
Some teams get to step three and find the screens themselves need work first. When that is your case and you want outside help, start with this comparison of landing page studios for B2B AI SaaS teams, and ask each one to show a features block they shipped with real product crops in it.
Studio Maydit designs and builds websites for AI founders, in Framer, Webflow or custom code, usually as a fixed scope of three to four weeks. A features section like the one above is mostly a cropping and captioning job once you know which screens to show, and you can do much of it yourself this week. If you are stuck on which three screens tell a stranger what your product is, bring your homepage and the screen you open first on demo calls, and book a 30 minute call.
