Why Is My Website Not Converting? The Traffic Businesses Pay For and Then Discard

The traffic businesses pay for and then discard

There is a permanent asymmetry in the cost of digital marketing, and almost nobody budgets around it. A visitor is bought every time: every click carries a price, that price rises as competitors bid, and the cost repeats next month. The page that receives the visitor is built once, and it improves the result for every visitor who arrives afterwards.

Budgets are allocated the other way round. The media line grows each year because it is easy to justify. The page keeps the treatment it was given at launch, and is revisited only when results disappoint.

Traffic is a recurring cost and the page is a fixed one. Businesses spend as though the reverse were true.

The size of that asymmetry can be calculated from figures a business already holds. Divide monthly media spend by monthly sessions to get the price of one visit. Multiply that by the visits the diagnosis below shows had no commercial intent, then by twelve. Compare the result with the one-off cost of the page work being proposed.

The reason this persists is structural. Nobody involved is being dishonest. The party best placed to notice that the traffic is wrong is the party paid to deliver it, since that supplier sees the search terms, the placements, and the behaviour after the click. A conclusion of "the visitors we are sending you cannot buy this" is an accurate diagnosis and also an argument for ending the arrangement that funds it. The available alternative, more volume, is comfortable and often partly true.

So the standard response multiplies the original loss: buy more of the traffic that was already failing. Only one omission is required. Nobody with an independent position ever formally asks whether the traffic is right.

What conversion actually means

A conversion is a defined action a visitor takes that has commercial value: an enquiry submitted, a call placed, a purchase completed. The conversion rate is the share of visitors who take it. Both definitions are looser in practice than they look. Most sites measure a mixture of actions of very different value and report them as one number, so a rise in newsletter sign-ups can conceal a fall in enquiries.

Published benchmarks are of little use as a target: reported medians sit in the low single digits for whole sites, purpose-built landing pages report considerably higher, and the spread between sectors runs several-fold. The useful comparison is internal: the same defined action, by traffic source, against the same figure last quarter and against the value of the customers it produced. That measurement also holds up against the attribution problem, where several platforms each claim the same sale.

One definition matters for the rest of this guide. A page is not failing because its rate is low in absolute terms. It is failing when the value of what the traffic produces sits below the cost of buying that traffic.

The diagnostic order

Almost every published guide on this subject is a list of page-level fixes: headline, button colour, testimonials. Those fixes are real, and they are the cheapest possible causes, tested first because they are the easiest work to sell. The correct order is the opposite: test the expensive hypotheses first, because if one of them is true, everything below it is wasted effort.

1. Source and intent. Who is arriving, and what they were looking for. If the intent is wrong, every later test measures the wrong population.
2. The offer. What the page asks the visitor to do, and whether that is a reasonable next step. A page can be immaculate and still ask for too much.
3. The page. Whether it confirms within a few seconds that the visitor is in the right place, and whether it presents one clear action instead of several competing ones.
4. The form or checkout. What the final step costs the visitor in effort, hesitation, and information they would prefer to withhold.

The four stages are also ranked by cost. Fixing intent changes what is bought, which affects the media budget. Fixing the offer changes the commercial proposition. Fixing the page and the form is comparatively cheap. Working upwards from the cheap end produces months of activity with no movement. Run all four stages as a diagnosis, then act on the highest-placed fault found.

Traffic intent is the first test

Intent is what the visitor wanted at the instant they clicked. It is not their demographic profile, and it is the largest single determinant of whether a page can convert them. Someone searching for a definition or a free alternative has a real intent that no commercial page satisfies. The mechanics are set out in the guide to why the majority of search traffic fails. What matters for a diagnosis is that intent is checkable.

• The actual search terms. Read the queries that produced the visits, not the keywords that were bought. The two routinely differ, and the gap is where the budget is wasted. That gap belongs to Google Ads settings, but the report belongs in this diagnosis too.
• The stage of the question. Sort the top queries into learning, comparing, and buying. If most sit in learning, the page is being asked to convert people who are months from a decision.
• Geography and eligibility. Traffic from markets the business does not serve, or from organisations below its minimum engagement size, should be removed from the denominator.
• Placement quality. For display and social campaigns, read where the impressions actually ran. Very short visits at high volume from one source indicate something upstream is wrong: an accidental click, automated traffic, or a page that failed to load.

If these checks show that most paid visitors had no commercial intent, the problem has been found where the money is, and the correct action is to change what is bought. This is the finding that rarely arrives unprompted.

Message mismatch between the advert and the page

Message match is the degree to which the page a visitor lands on confirms the promise that made them click. It is the most checkable failure in this subject, and anyone can audit it in an afternoon with no tools.

The method is mechanical. Open each live advert and write down its headline claim. Open the destination page and read only what is visible before scrolling. If the same claim, or an unmistakable restatement of it, is absent from that area, the match has failed.

The failures cluster in recognisable forms. An advert names a product and the link goes to the homepage, so the visitor must search for what they were just promised. An advert names a price that appears nowhere on the page. An advert names a service in the customer's vocabulary while the page uses internal brand names.

A visitor who cannot confirm within seconds that the page keeps the promise of the advert returns to the results and clicks a competitor, and the click has already been paid for.

Closing this gap has a second benefit. Landing page relevance is a documented input into how platforms rate an advert, and those ratings influence what each click costs, so correcting a mismatch improves the conversion rate and the price of the traffic at once.

The size of the step being asked for

The second stage of the diagnosis often ends the investigation. The traffic is right, the message matches, and the visitors still leave, because the only available action is far larger than anything they were prepared to do on a first visit. A page selling a considered purchase that offers nothing except a sales conversation is asking a stranger for time, exposure to a sales process, and a decision they have not yet made.

The repair is an intermediate step: something of genuine value that costs the visitor less than the main action and identifies them as interested. A good intermediate step is one a serious buyer would want even if they never proceeded.

• A specific answer in place of a meeting. A costed indication or a written response to a described situation, delivered without a call as a condition.
• A defined small engagement. A short paid piece of work at a fixed price and a fixed output, which lets a buyer assess a supplier at low risk.
• A comparison instrument. A calculator or questionnaire that produces a result the visitor keeps, and reveals their requirements.
• A lower-commitment channel. A message or callback request for buyers who will not book a calendar slot with a stranger.

The choice of step is a commercial decision. It is not a design decision. That is why it sits above the page in the diagnostic order, and why it is so often skipped: changing it requires authority over what the business sells. Where the step cannot be reduced, the honest conclusion is that the traffic should be bought at a different stage.

Competing calls to action and the cost of every field

Once intent and offer are confirmed, the page is worth examining. Two faults account for most of the remaining loss, and both share a mechanism: the visitor is offered choices when one action is required.

The first is competing calls to action. A page arrives with a main action and then accumulates alternatives: a newsletter box, a chat prompt, a related-products rail, and a consent notice covering part of the screen. Each was added for a defensible reason and each divides attention. A visitor presented with several possible next steps frequently takes none. The comparison worth making is against the same page with fewer competing elements, never against a published benchmark.

The second is the form. Every field is a decision point at which some proportion of visitors stop, and since the visit was purchased, each stop has a price. Fields asking for a telephone number or a budget band are the most expensive, because they carry an expectation about what happens next as well as the effort of answering.

Every field has to be justified by the use made of the answer: if nobody acts on it within the week, the field is being paid for and not used. Fields that exist to save qualification time later are paid for by the advertising budget. Fewer fields are commonly associated with higher completion, but the relationship is not reliable, because fields that signal seriousness can raise the quality of what arrives while lowering the count.

Speed and mobile treatment are direct conversion costs

Load time and mobile treatment are usually treated as technical maintenance and delegated accordingly. They belong in the conversion diagnosis instead, because they act on purchased visitors before any of the page's arguments have been read.

The industry has settled on published measurements: how long the main content takes to appear, how quickly the page responds to a tap, and how much the page moves while it loads. The thresholds are set against what real visitors perceive, and they are assessed at the 75th percentile of real visits on the devices visitors use. Mobile accounts for something above half of global web visits, so a page that performs acceptably on office hardware can still fail.

The treatments that cost the most are familiar. Content that shifts as images arrive causes mistaken taps. Targets sized for a mouse produce failed taps. A consent notice, a chat window, and a promotional panel together cover most of a phone screen. Tracking and chat scripts add delay, and their combined cost is rarely measured against the conversions they protect.

A failing measurement is worth converting into money before anyone argues about priority. Take the paid sessions arriving on the affected template, take the proportion arriving on mobile, and express the failure as a monthly cost of purchased visits lost before the page presented anything. Measure on a mid-range device, using field data and not a laboratory score.

What testing cannot settle

The standard advice at this point is to run a split test. For most businesses that advice cannot be followed, and the guides giving it rarely say so.

The volume requirement is higher than most sites can meet. The number of visitors a comparison needs depends on the existing conversion rate, the size of the difference to be detected, and the confidence required, and it rises steeply as the rate falls. Extending the duration does not rescue it, because a test left running for months spans changes in season, campaign, and pricing, so the variants are no longer compared under the same conditions.

Below a certain volume, split testing does not produce weaker evidence. It produces the appearance of evidence, which is worse than none.

The method that works at low volume is research followed by single sequential changes, and it is more demanding than testing.

1. Ask the people who did convert. A short question at the point of enquiry, asking what nearly stopped them, produces usable answers long before the volume a split test would require, because the answers are read individually.
2. Watch real sessions, and ask whoever answers the telephone. The questions callers repeat are the questions the page failed to answer.
3. Change one substantial thing at a time. Hold each change for a full business cycle and compare against the equivalent prior period.
4. Prefer large changes to small ones. Only differences visible above the noise are worth making, which rules out most cosmetic advice given elsewhere.
5. Record what was changed and when. Without a dated log the comparison cannot be made later.

This approach yields directional confidence. It does not yield statistical proof, and saying so plainly is the honest position. Evidence built from real customers is worth more than an underpowered test that reports a number.

Why localised pages convert worse in secondary markets

Businesses expanding into additional markets meet a failure with a consistent shape. Conversion drops in the new market, the traffic figures look healthy, and nothing in the analytics identifies a cause.

The sequence is usually this. An English page is built to serve a specific English search intent, and it is then translated. The translation is accurate, and it is a faithful rendering of a page built to answer a question people in the new market do not ask.

• The search intent was not re-researched. Keyword research is done in the source language and translated, producing terms that are linguistically correct and commercially wrong.
• The offer was not adapted. An intermediate step that works in one market can be unusual in another, where the expected next move is a written quotation instead of a booked calendar slot.
• Proof does not transfer between markets. Publication names and accreditations that carry credibility in one market are unrecognised in another.
• Practical details go untranslated. Currency, tax treatment, and payment methods are often left in source-market form, and each introduces doubt at the moment of commitment.
• Formality and register are wrong. Machine translation carries meaning reliably and register unreliably, and copy reading as too casual reduces trust without producing a measurable event.

None of these generate an analytics signal. The diagnosis requires reading the localised pages against local search behaviour and having them reviewed by someone who buys in that market. That work sits at the front of any expansion, alongside the market research and planning that decides which markets are worth entering, and it is the discipline behind how we approach international markets.

Doing it yourself versus specialist help

A substantial part of this work needs no supplier, and an owner who does it before commissioning anything will spend the following budget far better. In an afternoon it is possible to audit message match across every live advert, read the search terms report, remove form fields nobody uses, count the competing actions, and test the site on a mid-range phone. Most businesses that complete that list find one fault large enough to explain the results.

Specialist help earns its fee at three points. The first is research, because a business is rarely able to see its own page as a stranger does. The second is interface and page work at scale, where structure, hierarchy, and the mechanics of forms and checkouts are a professional discipline, covered by user experience and interface design and by landing page optimisation. The third is measurement, where establishing what a conversion is worth and reporting it against source underpins every other decision, which is the work of analytics and performance reporting.

One filter is worth applying to any supplier proposal. Ask which of the four diagnostic stages it addresses, and ask what evidence would lead the supplier to conclude that the traffic they supply is the problem. A supplier with an answer to the second question is worth engaging.

Key takeaways

• Visitors are bought repeatedly and the page is built once, so the cheaper of the two gets the least attention.
• The supplier best placed to identify wrong traffic has the least reason to do so, and more volume is the comfortable response.
• Diagnose in order: source and intent, then the offer, then the page, then the form.
• Failures that survive that order are usually failures of the offer, because the only available step is too large for a first visit.
• Below a certain volume, split testing cannot reach significance, and research with single sequential changes replaces it.
• Localised pages fail because the intent was translated and never researched, and the offer was never adapted.

The pattern behind every version of this problem is the same. The expensive question is asked last, and the months in between are spent improving something that was never responsible. Reversing that order costs nothing.

For an independent read of where a site loses the traffic it pays for, book a strategy call with Reachford. The answer sometimes is that the page is fine and the traffic should be bought differently, and that answer is included.