Why isn’t my website getting customers?
The question has one shape and five answers, and they happen in a fixed order. Only the earliest broken one is worth spending money on — every fix applied above it produces a better website and the same result. Each of the five checks below is free and can be run this week.
From the inside, all five failures look the same: silence.
A phone that does not ring produces no information about why. That is the whole difficulty. A business that nobody can find, a business people find and do not trust, a business people trust and have no reason to prefer, a business people choose and cannot reach, and a business where all four work but nothing accumulates — from the owner’s chair, every one of those is the same quiet week.
So the question gets answered by whatever is most visible instead, which is the design. The site looks tired, therefore the site is the problem, therefore the fix is a new site. That reasoning is not wrong so much as untested, and it is expensive to test last.
The five failures are the five stages of the Growth System, which is where the method itself is set out — what the stages are, why they run in that order, and why only the earliest broken one is worth spending on. This page does the other half: given the five, how do you work out which one is yours, from outside, this week, without paying anybody.
Fixing a later stage while an earlier one is broken does not produce a smaller result. It produces no measurable result at all.
Why the order is the whole argument.
A better booking flow cannot help a business that is not being shown to anyone. A sharper offer cannot help a business whose published web address answers for somebody else. Each stage depends on the ones before it, so a stage below a break cannot pay— not “pays less”, cannot pay. The money spent there is not badly spent; it is spent early, which is worse, because it also consumes the attention that would have found the real thing.
This is also why a list of forty faults ranked by severity is not a diagnosis. Every site has forty faults. The question a diagnosis answers is which single one is currently costing you, and that question has exactly one answer at a time.
The five, in order. Stop at the first one that fails.
Each finding below is a real fault found in real work and already published on the method page, where the businesses are anonymised because none of them gave permission to be named. They are here because a stage described in the abstract is easy to pass and hard to recognise; a stage described by what breaking looks like is neither.
- 01 · Discoverable
Can the right people find you?
What failing here looks like. You are found for your own name and for nothing else. The business exists online for people who already know it exists.
What it looked like in real work. A restaurant appeared in zero of nine category searches — in either of the two languages its customers use. It ranked for its own name and nothing else.
Check it yourself, free: Open Search Console and read the query list, not the position column. Count how many distinct queries are not your business name. If that count is near zero, this is your stage and nothing further down the list can be your constraint yet.
- 02 · Credible
Do they trust what they find?
What failing here looks like. They found you, then checked you somewhere you cannot see, and what they found did not agree with itself.
What it looked like in real work. A company's only published web address now belongs to a different business. Every prospect who followed the link met someone else — and the top review result for its name was a one-star record belonging to a firm it has no connection to.
Check it yourself, free: Search your own business name in a private window and read the whole first page, including the parts that are not you. Then follow every link on your own Google Business Profile and directory listings and confirm where each one actually lands.
- 03 · Preferred
Do they have a reason to choose you?
What failing here looks like. You are one of several credible options and the page gives a stranger nothing to decide on in ten seconds.
What it looked like in real work. A business with 4.9 stars from 1,197 people published no price anywhere. When customers asked Google what it cost, Google answered — using figures published by a competitor.
Check it yourself, free: Read your own homepage as somebody who has never heard of you and has three other tabs open. What do you sell, what does it cost, and why you — answerable in ten seconds, or not.
- 04 · Converting
Can they act without friction?
What failing here looks like. Someone decided in your favour and then could not complete the act. This failure leaves no trace at all unless you go looking for it.
What it looked like in real work. A business whose entire trade runs through the telephone had its number published on none of the directories a customer would find first. It was on the profile it controlled, and nowhere else.
Check it yourself, free: Submit your own contact form and confirm the message arrives. Call your own published number. Do both from a phone, on mobile data, not from the machine you built the site on.
- 05 · Compounding
Does it get better over time?
What failing here looks like. The first four are working and the business is still standing still, because nothing accumulates from one month to the next.
What it looked like in real work. A restaurant with a working site, a live ordering path and real customers had no analytics installed at all. Ten scripts on the page; not one of them measurement. Nobody could tell whether anything worked.
Check it yourself, free: Ask what changed on the site last month and what happened as a result. If the second half of that question has no answer, this stage is broken regardless of how the other four look.
Run them forwards and stop. The instinct is to run all five and total the score. That produces a list, and a list is the thing this article exists to argue against. The first failure is the answer; the remaining stages are not yet measurable and any judgement about them is a guess dressed as a finding.
A worked example, including the part where the constraint moved.
Evergreen Spa in Rawai, Phuket, is named here with permission, and every number below is already published at /work/evergreen-spa where it can be checked. A good business with strong reviews and returning customers, and the constraint was the earliest one there is: nobody could discover it. Not a conversion problem. A findability problem, and the site was the instrument.
One month after the new site went live it appeared in 23 distinct searches, having appeared in none, and was shown 552 times. Stage one moved.
And then the same month produced 8 visits. The report led with that rather than with the 552, because the interesting part is what it proves: the constraint had moved to a different stage, and staying pleased about the 552 would have been the comfortable way to miss it.
What the new constraint turned out to be is worth quoting exactly, because it is the failure people least expect. The wellness page was shown 191 times and clicked zero times, because Google was matching it to people searching wellness resort in rawai phuket — people looking for somewhere to sleep. Evergreen is a day spa. That single search accounted for 256 of the 552appearances. Separately, three of the site’s own pages competed for rawai spa, an exact-offer search shown 51 times a month, and Google resolved the competition by ranking none of them well: all three sat between position 42 and 55.
Neither of those is a design fault, and the site was four weeks old.The remedy for the first is words on a page — say “day spa”, say there are no rooms — and accept fewer impressions as the correct outcome. The remedy for the second is to give each page one job and point the internal links at whichever should win. The other two faults recorded that month were desktop performance and enquiries that could not be counted because tracking had never been installed. They are named so the two above are not mistaken for the whole list.
Traffic arriving is not the same as the right traffic arriving, and the difference is invisible in a total.
What the same five questions say about this website.
It would be convenient to end there. The honest version is that Aroc is currently answering this question about itself, and the answer is stage one.
A baseline was frozen on 3 August 2026, before the current site went live. In the 28-day window it covers, the site was shown 157 times across 13 distinct queries and received 2 clicks, at an average position of 62.01 — page seven. All thirteen queries were the previous business’s location terms or the brand name. Not one query the intended buyer would type registered a single impression.
The first post-launch reading, on 7 August 2026, was 146 impressions, 10 distinct queries and 0 clicks — and the same finding, unchanged: no query from the intent map appeared at all. The series that measures whether this is working still has no first value. It starts from nothing rather than from a number.
By the test in this article that is unambiguous. Aroc’s own constraint is Discoverable, and every stage after it is currently unmeasurable here for the same reason it would be unmeasurable on your site. So the work is the earliest stage, in public, weekly, with the numbers published whether or not they move.
Why publish a zero. An article arguing that businesses should diagnose honestly, written by a business withholding its own diagnosis, is an advertisement. The numbers above are two readings apart and prove nothing yet; the reason they are here is that they were recorded before the outcome was known, which is the only condition under which a number is worth anything later.
What to do once you know which stage it is.
- Write down the number that is currently wrong,before changing anything. Distinct queries, impressions, enquiries counted — whichever the failing stage produces. Without it, “did it work?” has no answer and you will settle it by impression.
- Change one thing. Changing five at once means the one that worked cannot be identified, and next time you will repeat all five.
- Give it the time that stage actually takes. A broken listing or contact route can move enquiries the same day. A content or title change takes weeks to be recrawled and re-evaluated. New content maturing into rankings takes three to six months. These are not the same wait and treating them as one is how people abandon a fix that was working.
- Then ask the question again from the top.The constraint moves — Evergreen above is the worked example — and a business that keeps fixing last month’s constraint is doing careful work at the wrong stage.
If the five checks come back obvious, act on them and keep the money. That is not a throwaway line: three of the five cost nothing but an afternoon, and a business that can diagnose itself should.
Aroc Assess is the paid version of the same exercise: every stage examined with evidence, the earliest broken one named as your constraint, what it is costing you, and what to do first — including, where honest, that major work is not justified yet. It is built entirely from what is visible from outside, with no account access, and that limit is written into the document rather than discovered later. If the answer turns out to be a rebuild, website redesign services is where that decision is set out in full. If it turns out to be the slower, compounding kind of fix, Grow is where it lives instead.
Sources for every figure above
- The five stage findings — zero of nine category searches · a domain answering for another business · 4.9 stars from 1,197 people with no published price · a phone number on none of the directories · ten scripts and no analytics — Aroc assessments run between June and August 2026, quoted from /growth-system where each already appears. The businesses are unnamed because none gave recorded permission, which ARO-CS-001 requires before a client or prospect appears on this site.
- 23 distinct searches · 552 appearances · 8 visits · 191 impressions and zero clicks · 256 of 552 · three pages at position 42–55 · “rawai spa” shown 51 times a month — Google Search Console, 28 June to 27 July 2026, re-verified 3 August 2026. Published with recorded client permission at /work/evergreen-spa, where each figure already appears.
- Aroc’s own baseline — 157 impressions · 13 distinct queries · 2 clicks · position 62.01 — Google Search Console, domain property
sc-domain:arocwebdesigns.com, captured 3 August 2026 before the cutover. The window is 6 July to 1 August, not the 28 days requested:the daily rows the export returned stop at 1 August because of Search Console’s reporting lag, so the period is 27 days. The 157 is a correct sum of what the file contains. The position figure is the impression-weighted site-wide average; the average across the thirteen queries is 69.61, and the two are not interchangeable. - The first post-launch reading — 146 impressions · 10 distinct queries · 0 clicks · no mapped query — same property, captured 7 August 2026, period 11 July to 7 August. This window overlaps the baseline by 22 days and its daily rows stop at 5 August, so it contains only two post-launch days. It is published as a level, not as a change: the movement between the two readings is inside the noise of a 146-impression sample and no conclusion is drawn from it here. The finding that is not noise is the one that is identical in both files — that no intent-mapped query appeared in either.
- “Three to six months” for new content to mature— Aroc’s own planning assumption, recorded in ARO-C-001 §0 before this article was written. It is a working assumption drawn from experience, not a measured figure from this site, and it is labelled as such here rather than presented as data.
Nothing above is estimated.Where a number could not be verified it is absent rather than approximated, and the two places where this article’s own evidence is weaker than its argument would like — the overlapping measurement windows, and the maturation range being an assumption — are stated in the source list rather than left for a reader to discover.