Why is my landing page not converting?
A landing page can lose the visit at several points: the wrong person arrives, the offer doesn't answer their question, the next step is hard to find, or the interaction fails. Find which one is happening before changing the page. A new headline won't repair a broken form, and a working form won't make an unsuitable offer useful.
First, check that conversion measurement works. Then follow a comparable group of visits from arrival to completion. Where do they stop progressing? That's a better place to spend your attention than the part of the page you already wanted to redesign.
What would a successful visit accomplish?
Be specific about success. A click on Book a consultation isn't a consultation request. A submitted form isn't necessarily a qualified inquiry. Someone downloading a guide doesn't tell you whether they used it. You need a meaningful outcome, along with the smaller steps that help you diagnose how people get there.
If the real outcome happens later in the sales process, keep that distinction visible. The website metric can help you understand progress without pretending every form submission is a new customer.
Do not mistake engagement for intent to buy. In Google Analytics, an engaged session can qualify through duration, a key event, or multiple views. Those criteria do not establish purchase intent. See Google's engagement definition.
Run through the journey with non-sensitive test data. Does the completion event fire at the right moment? Does refreshing count it again? Once it works, put the denominator beside the rate: completed inquiries per eligible landing-page visit, for example. That makes later comparisons much less ambiguous.
Did the visitor get what brought them here?
Consider an illustrative bicycle shop advertising a same-day puncture repair. The ad sends visitors to a general workshop page headed “Keep moving, beautifully.” The page features restorations, custom builds, and several service categories. Puncture repair appears near the bottom.
The page may be attractive and work perfectly. But the visitor has a puncture, and now they're browsing restorations. You can polish that page for a long time without resolving the mismatch. The first useful improvement may simply be showing the service the ad promised.
Compare the search term or campaign promise with what appears on arrival. Can someone confirm the service, location, eligibility, availability, or price context they need? Does the next action help with the task that brought them there?
Keep traffic sources separate while checking this. The emergency-repair visitor and the vintage-restoration reader have different immediate jobs. If you combine them, a change in the traffic mix can look like a change in the page's effectiveness.
Where does progress stop?
Think through the journey: arrive, understand enough to proceed, find the action, attempt it, and finish it. Then distinguish what you can measure from what you need to investigate. A section impression doesn't prove someone understood the offer, even if you give the event that name.
Use the observable steps to narrow the investigation:
| Evidence | What to investigate | Premature conclusion to avoid |
|---|---|---|
| Few eligible visits reach the action's section | Layout, information order, and traffic fit | Everyone needs a shorter page |
| Visits reach the section but few attempt the action | Offer, commitment, clarity, and competing paths | The button needs a brighter color |
| Attempts are common but completions are rare | Form requirements, errors, and response behavior | The audience is low quality |
| Completions exist but inquiries are unsuitable | Targeting, qualification, and the promise | The page needs more conversions at any cost |
Work on the earliest problem the evidence supports. If the primary action doesn't load on a common mobile layout, the confirmation screen can wait. Visitors need a way to reach it first.
What does the visitor need to know right now?
At the repair shop, “Book now” could mean committing to a paid appointment or asking whether a slot is available. Those are different decisions. If the form doesn't explain which one you're making, a short label has left a fairly large question unanswered.
“Check today's repair availability” could be clearer, provided that's what happens next. Explain what information the shop needs and when it will respond. If a person must approve the request, don't make the button sound like instant confirmation.
Look for the same gaps elsewhere: what's included, what happens next, whether the service fits this situation, and what submitting the form commits someone to. Put each answer near the decision it supports. Having it somewhere in a long FAQ isn't much help if someone doesn't know to go looking.
Give people something they can evaluate: a named review with permission, a concrete service description, or a clear example. Calling the service exceptional doesn't explain what it includes or whether it's suitable.
Try the path a visitor is trying to finish
Open recordings of attempts that didn't finish, and compare them with successful visits on the same page version and device class. Does the keyboard cover a field? Is an error easy to find? Is an overlay catching taps? Does the next step acknowledge the action? Those are specific things you can test.
Reproduce the suspected problem. A delayed response may be a network-dependent operation, a busy main thread, or missing progress feedback. These need different fixes. Interaction to Next Paint measures responsiveness to a subsequent paint; it does not measure an entire asynchronous business operation. The distinction is explained in web.dev's INP guide.
The repair shop's button might respond at once while the availability lookup takes longer. That still leaves a question: does the page make the wait understandable? Try the full interaction, including failure and retry, instead of treating a performance score as the whole answer.
Complete the path with a keyboard, too. If you can't reach the next action or find the error, you have a problem worth fixing. A reassuring average conversion rate doesn't make that path usable.
Choose the change you can explain
Ask which change has the strongest evidence, addresses the most consequential problem, and can be checked clearly. A submission error you can reproduce usually deserves attention before a headline you're unsure about. You don't need to invent a score with two decimal places to make that judgment.
For the repair shop, try a puncture-repair destination that matches the campaign and explains the availability request. The hypothesis is straightforward: visitors arriving for that service should be able to identify it and attempt the next step without searching through unrelated offerings. You can test that task before waiting for conversion results.
Keep a brief change record:
Audience and source: Local puncture-repair campaign Problem: Campaign promise is difficult to find on arrival. Change: Relevant service explanation and availability action Primary outcome: Completed requests per eligible campaign visit Guardrail: Inquiry suitability and duplicate submissions Page version: Record before and after releases separately. Other changes: Campaign, price, availability, and tracking
Choose the success metric before looking at the result. Otherwise, it's very easy to find a number that makes the change look successful.
Decide what “better” means before the results arrive
Verify that the new page and its measurement work, then compare outcomes for the intended audience. Include counts and context. Did the traffic source change? Did more appointments become available? Was demand different? Put those changes beside the result so the next person doesn't have to guess.
A before-and-after increase can support further investigation, but it does not isolate the effect of your edit. Use a properly designed experiment when you need a causal estimate and have the conditions to support one. With limited traffic, combine careful measurement with task-based testing and fixes to demonstrated defects.
You might end up with fewer inquiries that are a better fit. You might get the same conversion rate with fewer support questions. Both could be worthwhile. Agree on the customer and business outcome before the chart arrives, or the easiest number to increase will quietly become the goal.