Priya runs a wedding flower studio. On Thursday night, a friend who works in marketing glances at her website over dinner and says, kindly, “It's nice. But I wouldn't book from it.”
Priya asks why. The friend shrugs. “I don't know. It just doesn't feel finished.”
That is the least useful feedback in the world, and the most common. It is not wrong. It is just not a list. Priya cannot fix a feeling. She can fix “your phone number is only on the contact page” and “the button says Submit”.
Priya and her studio are imagined. The research, tools and laws below are real.
This chapter turns the previous 23 into that list. It is a self-scoring audit you can run on any landing page in about ten minutes: yours, a client's, a competitor's. Every item links back to the chapter that explains the why.
Why a checklist works
Checking a design against a short list of principles has a name and a history. In 1990, Jakob Nielsen and Rolf Molich published “Heuristic evaluation of user interfaces” at the CHI conference. The method is plain: a few people inspect an interface on their own, compare it with a set of recognized usability principles (“heuristics”), and list where it breaks them.
Their finding was honest about the limits. Individual evaluators were not very good at it: in their experiments each person found only around 20% to 51% of the known problems. But when they pooled the lists of several evaluators, even three to five people, the combined list caught far more. Later guidance from Nielsen Norman Group still recommends three to five evaluators working independently.
Two lessons follow for a small business. A checklist is cheap and catches a lot. And you alone will miss things, so ask one or two other people to run the same list and merge the results.
Inspect a design against a short list of established principles instead of testing it with users. Fast and inexpensive. One evaluator finds a minority of problems; the pooled findings of three to five find many more. It finds violations of known principles well, and surprises about real users poorly, so it complements user testing rather than replacing it.
Ten heuristics, one page
Nielsen refined the principles in 1994 into the ten usability heuristics that are still taught everywhere. NN/g updated the explanations and examples in 2020 and last reviewed the article in January 2024, but notes the heuristics themselves have stayed essentially unchanged since 1994. They were written for software, so here is what each means on a landing page, in our words.
- Visibility of system statusButtons react when tapped; forms say they are sending and that they sent.
- Match between system and the real worldYour customer's words, not your industry's jargon.
- User control and freedomEasy to go back, close the pop-up, change an answer.
- Consistency and standardsLinks look like links; the logo goes home; the cart is top right.
- Error preventionClear formats and sensible defaults before mistakes, not scolding after.
- Recognition rather than recallShow the options and prices; do not make people remember them from three sections ago.
- Flexibility and efficiency of useShortcuts for the ready: a “Book now” near the top for people who already decided.
- Aesthetic and minimalist designEvery element competes for attention; remove what does not help the decision.
- Help users recognize, diagnose, and recover from errorsError messages in plain words, next to the field, saying how to fix it.
- Help and documentationAn FAQ that answers the real objections, close to the button.
The scorecard below is built on the same idea, but tuned for one job: a page that has to earn trust and a decision from a stranger.
Five users, and the fine print
Checklists find violations of principles. They do not find what you never thought of. For that you need to watch real people use the page, and the most famous number in UX says you only need five.
It comes from Nielsen and Thomas Landauer's 1993 paper “A mathematical model of the finding of usability problems”. Nielsen popularized it in 2000 in “Why You Only Need to Test with 5 Users”. The model says the share of problems found by n users is 1 − (1 − L)n, where L is the share a single user reveals. With a typical L of 31%, five users find about 85%.
Now the fine print, which matters more than the headline.
It is an average, and averages hide luck. Laura Faulkner tested 60 users and drew random groups from them (Behavior Research Methods, 2003). Some groups of five found 99% of the problems; others found only 55%. With 10 users, the worst group still found 80%; with 20, 95%.
L is not a constant. 31% was a typical value in the projects Nielsen drew on. On a complex site with varied users, one person reveals less, and five find less.
Nielsen's own advice was iterative. His 2000 article recommends three rounds of five users with fixes in between, rather than one study of fifteen, and separate users for clearly different audiences.
And beware its folk cousin. Joel Spolsky's well-loved “Joel Test” (2000) recommends “hallway usability testing”: grab the next person who walks past and watch them try your thing. Great advice. His line that five such people teach you “95% of what there is to learn” is a rule of thumb, not a measurement.
The practical version for a small business: run the checklist yourself, then watch three to five people from your audience try to do the main thing on your page. A quick variant is the five-second test, described by Christine Perfetti in a 2007 UIE article: show a page for five seconds, hide it, and ask what people remember and what the page was for. It suits pages with one clear purpose, which is exactly what a landing page should be. We cover a tighter version in the three-second test.
The ten minutes, minute by minute
Here is the running order. Start the speed test in a second tab at minute zero so it is ready when you get there.
Two rules make it work. First, do it on a phone, because most visitors will. Second, tick only what is clearly true. “Sort of” is a no. The score is for you; flattering it helps nobody.
The scorecard
Thirty-two yes-or-no statements in six groups. Tick what is true for the page you are auditing. The score and the gaps update as you go, and the print button gives you a clean sheet to share or keep. The checkboxes work with a keyboard (Tab to move, Space to tick) and the list still works as a plain checklist without JavaScript.
0 / 32
Count your ticks: 28 or more is a strong page; 20 to 27 is solid with clear gaps; 12 to 19 is leaking trust or decisions; under 12, a rebuild may be cheaper than patching.
A note on the numbers: every item counts the same, which is a simplification. A missing price may cost you more than a slow hero image on a page with little traffic, and a legal gap can matter more than both. The score is a conversation starter for your own priorities, not a validated instrument.
Free tools that do the measuring
Half the checklist is looking. The other half is better done by tools. All of these are free.
| Tool | What it tells you | Items |
|---|---|---|
| PageSpeed InsightsGoogle | Real-user Core Web Vitals from Chrome users (if your page has enough traffic) plus a Lighthouse lab report with fixes | 22–25 |
| LighthouseBuilt into Chrome DevTools | Lab performance, accessibility and SEO checks on one simulated load; names your LCP element | 22–26 |
| WAVEWebAIM | Visual overlay of accessibility errors: contrast, missing alt text, missing labels, empty buttons | 26, 28 |
| axe DevToolsDeque, free browser extension | Automated WCAG checks with explanations of each failure | 26–28 |
| Contrast CheckerWebAIM | The exact WCAG contrast ratio of any two colors | 26 |
| Search ConsoleGoogle | Core Web Vitals across your whole site, grouped by page type, over 28 days | 22–24 |
One warning applies to all of them: automated tools find only part of the problems. WebAIM says so plainly in its own Million report: a page with no detected errors is not thereby accessible. Tools are your floor, not your ceiling.
Legal quick-checks
This is not a legal audit. It is the short list of things a visitor, a competitor or a regulator can spot from the outside in a minute, and that a surprising number of small sites miss. EU rules first, since that is where most of our readers trade, then the UK and US equivalents.
Each check below names the main rule behind it. National laws add detail, and some apply only to businesses selling to consumers. Use this to spot gaps, then confirm with an adviser.
General information, not legal advice. Check the rules in your own country and sector.
| Check | Main rule | What to look for |
|---|---|---|
| Company details (imprint) | e-Commerce Directive 2000/31/EC, Art. 5(1)Germany: DDG §5. Hungary: Ekertv. (2001. évi CVIII. tv.) §4 | Name, geographic address, email, trade register number and VAT number where applicable, “easily, directly and permanently accessible” |
| Privacy notice | GDPR (EU) 2016/679, Art. 13 | Who you are, what you collect through the form and why, how long you keep it, people's rights. Linked near every form |
| Cookie consent with real choice | ePrivacy Directive 2002/58/EC, Art. 5(3); GDPR Arts. 4(11) and 7(3) | No non-essential cookies before consent; consent must be freely given, and “as easy to withdraw as to give”. A “Reject all” as visible as “Accept all” is the safe design |
| Prices and discounts | Price Indication Directive 98/6/EC, Arts. 2 and 6a; e-Commerce Directive Art. 5(2) | Consumer prices include VAT and other taxes; say whether delivery is included; any announced price reduction shows the lowest price of at least the previous 30 days |
| Honest urgency | Unfair Commercial Practices Directive 2005/29/EC, Annex I point 7UK: DMCC Act 2024, Sch. 20 para. 7 | No false claims that an offer is available only for a limited time. A countdown that resets on reload is exactly this |
| Reviews | UCPD Art. 7(6) and Annex I points 23b–23cUK: DMCC Act 2024, Sch. 20 para. 13. US: FTC rule, 16 CFR Part 465, in effect since 21 Oct 2024 | Say whether and how you check that reviews come from real customers; never write, buy or cherry-pick fake ones |
| Accessibility statement | European Accessibility Act, Directive (EU) 2019/882, Art. 13(2) and Annex V | If your consumer service is in scope (and you are not a microenterprise), information on how it meets the accessibility requirements, published in an accessible format |
The UK provisions of the DMCC Act's unfair-practices regime came into force on 6 April 2025. The US FTC reviews rule bans, among other things, fake and AI-generated reviews, buying reviews with a required sentiment, and undisclosed reviews by insiders. For the design side of urgency and defaults, see Defaults, nudges and honest urgency; for reviews, Social proof patterns.
After the audit: fix in order
A score of 19 is not a verdict. It is a to-do list in disguise. The trick is the order.
Legal gaps first, because they are cheap to fix and expensive to ignore: an imprint, a privacy link, a fair cookie banner take an afternoon. Then the first screen, because every visitor sees it and most see nothing else. Then decisions: button labels, form fields, a price. Then speed and accessibility, which often share fixes (smaller images, fewer scripts, real buttons). Trust and structure usually need new content, photos or testimonials, so they take longest; start collecting now.
Then run the audit again in a month, ideally with someone else holding the phone. Remember Nielsen and Molich: your second evaluator will find things you could not.
Run the scorecard tonight
On your phone, on your own page, honestly. Print or save the result so you can compare later.
Ask two other people to run it
Separately, without seeing your answers. Merge the three lists; the overlap is your priority and the differences are what you could not see.
Fix every legal gap this week
Company details, privacy notice, a cookie banner with real “Reject all”, VAT-inclusive prices, honest discount and review statements.
Rewrite the first screen next
Headline, one button with a specific label, a real photo. Then check it with the three-second test.
Watch three people use the page
Give them the real task (“book a consultation for a June wedding”) and stay quiet. Write down every hesitation.
Re-audit in 30 days
Field speed data takes 28 days to refresh anyway. Compare the two printouts.
Priya ran the scorecard on Friday morning and got 17. Her phone number was on one page out of five, her button said “Submit”, there was no price anywhere and her hero photo was a stock bouquet she had never made. None of it needed a redesign. By Tuesday she was at 26, and her friend's verdict had changed to something far more useful: “I'd book.”
What the research supports
Heuristic evaluation is an established, inexpensive inspection method with known limits: single evaluators miss many problems, and pooling a few evaluators helps (Nielsen & Molich, 1990). Small-sample user testing finds many problems on average, but results vary widely between groups (Nielsen & Landauer, 1993; Faulkner, 2003). This scorecard is an editorial synthesis of the Playbook's chapters, not a validated scale; its equal weights and score bands are judgment calls. The legal checks summarize real rules but are not complete or tailored advice.
Sources
- Nielsen, J. & Molich, R. (1990). Heuristic evaluation of user interfaces. Proceedings of CHI '90, 249–256. Abstract consulted for individual evaluator detection rates (about 20–51%) and the benefit of aggregating three to five evaluators.
- Nielsen, J. (1994, updated 2020, reviewed 2024). 10 Usability Heuristics for User Interface Design. Nielsen Norman Group. Heuristic names, history and update notes.
- Nielsen Norman Group (2023). How to Conduct a Heuristic Evaluation. Recommendation of three to five independent evaluators.
- Nielsen, J. & Landauer, T. K. (1993). A mathematical model of the finding of usability problems. Proceedings of INTERCHI '93. Cited for the model; details taken from Nielsen (2000).
- Nielsen, J. (2000). Why You Only Need to Test with 5 Users. Nielsen Norman Group. Formula, L = 31%, iterative testing advice.
- Faulkner, L. (2003). Beyond the five-user assumption: Benefits of increased sample sizes in usability testing. Behavior Research Methods, Instruments, & Computers, 35(3), 379–383. Abstract consulted for the 55–99% range and the 10- and 20-user minimums.
- Spolsky, J. (2000). The Joel Test: 12 Steps to Better Code. Joel on Software. Hallway usability testing.
- Perfetti, C. (2007). 5-Second Tests: Measuring Your Site's Content Pages. UIE / Center Centre. Method and suitable pages.
- WebAIM (2026). The WebAIM Million. Limits of automated testing.
- Directive 2000/31/EC (e-Commerce), Art. 5, EUR-Lex; Germany, DDG §5; Hungary, 2001. évi CVIII. törvény §4.
- Regulation (EU) 2016/679 (GDPR), Arts. 4(11), 7(3), 13, EUR-Lex; Directive 2002/58/EC as amended, Art. 5(3), consolidated text.
- Directive 98/6/EC (Price Indication) as amended by Directive (EU) 2019/2161, Arts. 2 and 6a, consolidated text.
- Directive 2005/29/EC (Unfair Commercial Practices) as amended, Art. 7(6), Annex I points 7, 23b, 23c, consolidated text.
- UK Digital Markets, Competition and Consumers Act 2024, Schedule 20 paras. 7 and 13; commencement 6 April 2025 per s. 226 commencement note (S.I. 2025/272).
- Federal Trade Commission (2024). Final rule banning fake reviews and testimonials (14 August 2024) and Rule Q&A (effective 21 October 2024).
- Directive (EU) 2019/882 (European Accessibility Act), Art. 13 and Annex V, EUR-Lex.
Sources checked 18 October 2026. The scorecard is an editorial tool, not legal or professional advice. Priya and her studio are fictional; the curve in Figure 1 was calculated from the published formula.