Website analysis: four unrelated jobs share one name

You typed your own address into a checker and got a score out of 100, a list of images to compress and a warning about a missing description tag. You fixed some of it. Nothing happened to the number of enquiries, which was the reason you ran the check in the first place.

The tool was not wrong. The trouble is that four unrelated jobs share one name, and only one of them answers the question you had.

Four jobs that share one name

Is the site technically sound

This is what most free checkers measure: load time, rendering on a phone, broken links, whether a crawler can reach every page. PageSpeed Insights and Lighthouse do it at no cost, and the output can be handed straight to whoever built the site.

What it cannot tell you is whether the people reading your pages are people who could hire you. A site can score well and be read all year by students, job seekers and competitors.

Does anyone find it

Search Console shows the queries you already appear for. Keyword tools estimate what strangers type.

The limit here is sharper for a firm your size than the tools admit. If you do structural work on cold storage warehouses, the phrase that describes what you sell may be typed a handful of times a month in your whole country, and most tools will report it as no data. That is not proof that nobody needs the work. It is proof that search is a thin channel for narrow expertise, and that a report full of zeroes says more about the tool than about your market.

What visitors do once they arrive

Analytics answers this: how many arrive, where from, which pages they read, where they leave.

With a few hundred visits a month you cannot test anything. Every figure is an anecdote with a decimal point. It is still worth ten minutes a quarter, mostly to see which service page people actually read and which one you wrote for your own satisfaction.

What the site says about who pays you, and why

This one has no score and no category, and it is the only one that answers the question you really have: who should I be talking to.

The answer is already in your pages, written in the words you used when you were not thinking about marketing. The sectors named in your project descriptions. The job titles of the people who agreed to be quoted. The size of job your examples imply. The problem you say you solve in the second paragraph of the services page, rather than the slogan in the first.

That reading is the first step of any approach to somebody who has never heard of you, and it is the step most owners skip, because it feels like something they already know by heart.

Choosing who to write to is real work, and it is work a machine can do in your place: read the pages, work out who pays the company and why, then find the decision makers who resemble those customers. That is the step Klientee performs. Klientee is my product, and you should know that before you read another line of this.

How a commercial reading of a site actually works

You can do this yourself on a Friday afternoon with a document and no tools. Six steps, in this order.

1. Gather only the pages that make claims: home, services, project or case pages, about, and any job advert you have published. Leave the blog out of it for now. Blog posts describe what you wanted to rank for, not what you sell.
2. Separate what you sell from who buys it. Almost every site describes the first in detail and leaves the second implied. Write the two in separate columns and notice how uneven they are.
3. Attach an extract to every statement you make about your buyer. One sentence, copied from your own site, under each claim. If nothing on the site supports it, the claim is a guess and has to be written down as a guess. This single rule is what separates a reading from a brainstorm, and it is the rule that makes the result arguable by someone else.
4. Write the buyer in one sentence: role, company size, sector, and the moment they start looking for someone like you. The moment is the part people forget, and it is the part that decides whether an approach lands.
5. Hold that sentence against your last ten invoices. Not your favourite ten. The last ten. Where the site and the invoices disagree, the invoices are right and the site is out of date.
6. Only then look outward, at companies and people who resemble what you just described.

Steps one to five take an afternoon. Step six is the one that never gets done, because it does not finish. It is a standing job.

What goes wrong, from reading a lot of small service sites

Four patterns come up again and again.

The method is described, the client is not. Three paragraphs on your process, a diagram, your certifications, and not one sentence about who was in trouble and why they called. A reader who is not already convinced has nothing to recognise themselves in.

The real buyer is hiding in the case pages. The services page says you work with organisations of all sizes. The three project pages all describe a factory between fifty and three hundred people, in two sectors. The project pages are telling the truth.

The site is written for people who already know you. This is the most common one, and it is a direct consequence of living on referrals. When every visitor arrives pre introduced, you stop explaining, because the person who sent them did the explaining for you. The page slowly becomes a confirmation rather than an argument.

The generality trap. If the whole site reduces to helping businesses grow with tailored solutions, then no reading, human or automated, will produce anything specific. The free proof we run comes back thin on those sites, and the honest answer is that the fix is on the site, not in the tool. One page naming a sector, a problem and a size of job changes the outcome more than any analysis.

If the reading is meant to end in a conversation

Most people arrive at this question eventually, and the market answers it with three different shapes of product sold as if they were one thing.

Data and enrichment. Clay, Apollo, ZoomInfo, Lusha. They hand you rows. You still decide who deserves a message, write it, warm the mailboxes and send. On 29 September 2026 Clay published tiers from $113 to $1,913 a month, with credits at $0.01 and $0.05. On the ZoomInfo and Lusha pricing pages that day I could not read a public price. This shape fits someone who already knows exactly who to contact and lacks addresses. It is the wrong purchase if your open question is still who.

Sending. Instantly at $47, $97 and $358 a month. Smartlead at $32, $59 and $78, with add ons priced separately at $29, $39 and $59. Saleshandy at $34, $76 and $149 a month on annual billing, higher month to month. They send sequences; you bring the list and the words. On the lemlist and Woodpecker pricing pages I could not read a public price on 29 September 2026. Right for someone with a list and copy they trust, wrong for someone who has neither.

End to end. Read a site or a brief, choose the targets, write, send. AiSDR starts at $250 a month and goes to $2,500, with a fully managed option at a further $2,500 a month. Regie.ai showed tiers from $0 to $150. Explee prices in credits, from $5 per thousand. On the Artisan and 11x pricing pages I could not read a public price that day.

Klientee is in that third group. It costs $149, $349 or $749 a month per site, published, with no credits and no per seat multiplication, and the first reading runs before any account exists: you type your address and read what it understood about your business, each claim shown next to the extract from your own pages it came from. It fits a B2B service firm with no salesperson and a site with at least a few pages of specifics. It is the wrong purchase in four cases: your order book is full and your referrals are steady; your site is one page of generalities; you sell to consumers; or you already have a list of twenty named accounts, in which case you are paying for a step you have finished.

Those prices were read from each vendor's own public pricing page on 29 September 2026 using a headless browser. Some pages block automated reading or keep the figure behind a demo request, which is why several lines above say I could not read a price, rather than claiming none is published. If a figure has moved or I read it wrong, write to me and I will correct it with the date.

The reason any of this matters

A referral does three things for you at once. It explains what you are worth, to somebody who resembles the person doing the referring, at the moment that person needs the work. That is an enormous amount of invisible labour, and it is why the channel feels effortless.

The part you never control is the third one. You get four in a month and then quiet for a quarter, and the quiet arrives without warning. I wrote about that pattern separately in why your best clients stopped referring you.

Analysing your site is not a marketing chore. It is the only way to do the first two things yourself in the months the third one does not happen: to know who resembles your customers, and to say what you are worth in words they recognise. The technical score, the keyword report and the visit count are all useful. None of them does that.

Axel Joseph Auguste