◔Visitors, figures and A/B tests

With audience measurement. There are two routes, and the choice has consequences:

• Privacy-friendly measurement — without identifiers, without recognising returning visitors. Provides page views, referrers and device type. Usually needs no consent banner.
• Classic tools such as Google Analytics. Provide more, but set identifiers and therefore need consent — with everything that entails.

For the vast majority of businesses, the first route is entirely sufficient. It answers the questions you actually have — “is anyone coming at all?”, “from where?”, “which page is being read?”.

If you measure nothing, that is a decision too. But you are flying blind when you want to know whether a change has made any difference.

Three things reliably, everything else with caution:

• Is anyone coming? The order of magnitude — ten, a hundred or a thousand page views a month.
• From where? Search engine, typed in directly, a link from elsewhere, social networks. This is the most useful figure of all.
• Which pages? What is read and what nobody opens.

What you cannot read from them: why someone did not call, whether the colour works, whether the text convinces. Figures say what happened, never why.

And the most important restraint: with fifty visitors a month, any change of ten per cent is chance. Compare months, not days.

That depends so heavily on the business that any general figure is misleading. As a rough guide for small businesses operating regionally:

• Under 50 a month: the site is practically invisible. Almost always because it is not linked anywhere and there is no entry on a map service.
• 100 to 500: the usual range for a trades business with a well-kept listing.
• Over 1,000: either a larger catchment area, ongoing advertising or a blog that gets found.

More important than the figure is the ratio: how many enquiries come per hundred visitors? If there are none at all, it is not down to the volume but to the site.

And: do not count yourself. The most frequent “visitor” to a small website is its owner.

With small numbers: nothing. That is chance, not a signal.

Two rules of thumb:

• Under a few hundred page views a month, fluctuations of 20 to 30 per cent are normal. A single post in a local group doubles a day.
• Compare like with like. A month with public holidays is no comparison with the month before.

What genuine signals look like:

• A sustained decline over several months.
• A sudden crash to almost zero — then something is usually broken: site not reachable, search engines locked out, domain expired.
• A rise after a measure that holds.

Everything else is weather, not climate.

Go through it in order — the first three explain almost everything:

1. Is the site reachable? Open it in a private window, ideally over mobile data.
2. Is the domain still running? An unpaid domain is switched off silently.
3. Have you changed addresses? Renamed pages without a redirect cost visibility immediately.
4. Is the site still in the search results? Search for your company name.
5. Season? Some businesses simply have less demand in July or December.

If everything is in order and the decline persists: it may be due to a change at the search engine. The only remedy is to keep going — good content, correct details, clean technology.

Sorted by effect, for a business operating regionally:

1. The entry on the map service. Well kept, with photos, opening hours and reviews. For many businesses the single most important source — often more important than the website itself.
2. Being linked everywhere. Email signature, vehicle, invoice, business directories, trade guild, suppliers.
3. One page per service, with the place in the text.
4. Posts that answer questions people search for.
5. Paid ads — work immediately, but stop working as soon as you stop paying.

What does not work: stuffing more search terms into the text, redesigning the home page, changing provider.

Not from the visitor count — search engines do not pass the terms on to it. You need the search engine’s console, a free tool where you prove once that your website is yours.

What you see there:

• Which terms your site appeared for in the results list.
• How often and in which position.
• How often it was clicked.

This is the most useful information there is about a website — because it tells you what your customers really type in, and that is almost never what you assume.

The most common insight from it: you already appear for terms you never thought of. Then a dedicated page for them is worthwhile.

As a rough guide: one to three in a hundred visitors get in touch, more with well-matched enquiries. For a page that answers exactly the right question, it can be considerably higher.

If you are below that, there is almost always one of four causes:

• The contact is hidden. Phone number not in the header area, form on a subpage.
• There is no prompt. No call to action, no invitation.
• The price is missing. Anyone who cannot estimate what it costs does not enquire but keeps looking.
• The visitors are the wrong ones. If you are found for something you do not even offer.

Do not calculate in percentages when you have fifty visitors. Count enquiries.

Only from a scale that most small websites do not reach: a few hundred visitors per version, so that a difference stands out from chance at all.

With fifty visitors a month, an A/B test is not a gain in insight but a coin toss with extra effort.

If you have the numbers, test what carries the most weight:

• The headline at the very top.
• The button label.
• Whether a price is stated or not.

Not: colours, font sizes, spacing. That rarely changes anything and takes the same time.

And always only one thing at a time — otherwise you will not know afterwards what made the difference.

Three things:

• In the privacy notice. What is measured, with what, how long the data is kept.
• Consent, as soon as the tool sets identifiers or recognises returning visitors. Without consent, it may then not even load.
• Data processing agreement (AVV) with the provider of the tool, if it is a third party.

With tools whose servers are outside the EU, the question of data transfer comes on top — that is a topic of its own with its own case law.

The easiest way to spare yourself all this: privacy-friendly measurement without identifiers. Fewer figures, but no banner, no agreement, no discussion.