Skip to content
PageSpeed 100 as the delivery default
Analyse

Website Metrics That Matter for Small Businesses

Six numbers instead of a data graveyard: which website metrics actually trigger a decision, and how they turn into a monthly sheet with just three values.

14 min read KennzahlenAuswertungAnfragenDatenschutz

In many small firms there is a login somewhere to an analytics dashboard that nobody has opened in months. That is not negligence, it is a reasonable response: if what you find there is a wall of curves, bars and two dozen metrics, none of which suggests a single action, you eventually stop looking. The way out is not more data, it is less. A business without a marketing department needs six numbers, and on a monthly rhythm only three. This article shows which ones they are, why visits, time on page and bounce rate are usually misleading on small sites, how to count enquiries where they actually arrive, and how a single sheet of paper per month turns into a decision. Data protection is part of the design, not an afterthought: measuring works without personal data and without third-party advertising networks.

Website metrics: the monthly sheet for small businessesThree numbers, one comparison month, one action instead of a data graveyardMonthly sheet, JuneThree numbers that trigger a decisionQualified enquiriesInbox, phone list, calendar14Prev. month 11 (+3)Calls from the websitePhone link on mobile9Prev. month 7 (+2)Appointment requestsForm and callback list5Prev. month 5 (0)Totals without personal data — names stay in the inboxVisits from a phone68 %of all visitsCheck mobile first, desktop afterLoad time: the key pagesHome1.4 sServices1.7 sContact1.2 s2.5 s referenceMeasured on a phone, not on the office desktopReference: 2.5 s to the largest elementThe monthly routine: count, compare, one action1CountRecord enquiries where theyarrive: inbox, phone, calendar2Ask the originOne question at first contact:How did you hear about us?3CompareOnly with the previous monthand last year, not with days4One actionExactly one change per month,or the effect stays unclearMeasurement period of at least four weeks — allow for holidays, breaks and weatherNo personal data, no third-party ad networks

Why the usual numbers mislead on small sites

Having a website is standard practice among small and mid-sized firms. 94 percent (Bitkom) of the trade businesses surveyed name their own website as a measure for being visible online, followed by listings in online directories at 88 percent (Bitkom). Across all sectors, 69 percent (Federal Statistical Office of Germany) of companies in Germany had a website, and among firms with one to nine employees roughly 65 percent (Federal Statistical Office of Germany). So the question is no longer whether a site exists. It is whether anyone in the business can say what that site actually does. This is exactly where most analytics fail, because they deliver numbers built for a retailer with six-figure traffic, not for a firm with three hundred visits a month.

The reason is arithmetic, not carelessness. With three hundred visits and twelve enquiries a month, the enquiry rate is four percent. If the number drops to nine enquiries the following month, that is twenty-five percent less on paper, but three enquiries in reality — roughly the amount explained by a week of holiday closure, a heatwave or a long weekend. Large percentage swings on a small base are noise. React to them and you change something different every month, and by the end of the year you cannot say what worked. There is a very practical reason why many firms hold back here: 58 percent (Bitkom) say they simply lack an overview of digital applications.

Visit counts

They rise with every search engine update, every crawler pass and every shared link, without a single job resulting from it. On a small base they fluctuate heavily and say little about whether the workshop is booked out. As a sole measure of success they are of limited use.

Time on page

The number does not measure interest, it measures confusion. Someone who finds the opening hours in eight seconds produces a short session and is perfectly satisfied. Someone who searches for four minutes and then leaves produces a long one. Without knowing the visitor's goal, the direction cannot be interpreted.

Bounce rate

On a page with a phone number, an address and directions, a bounce is often the success case: the information was found, and the call follows. With a few hundred views a month the value also fluctuates so much that comparisons between two months are rarely reliable.

How to recognise a usable metric

Three conditions have to come together. First: the number arises where work happens anyway — in the inbox, on the phone or in the calendar. Second: there is an action that changes when the number goes up or down. Third: the number is large enough that a swing of one or two does not reverse it. Test any metric against these three points and you will discard half of the usual reports before opening them for the first time.

Six numbers that trigger a decision

What remains is a manageable list. It is deliberately built so that each number matches a concrete question in the business and none of them requires an additional tool. Four of the six values arise without any technology at all, simply by recording what comes in anyway. The other two describe the condition of the website itself and change slowly enough that a quarterly check is sufficient.

Qualified enquiries per month

The one number that counts on its own if you only keep one. Qualified means: it concerns a service the business offers, in an area it covers, with a reachable contact detail. Everything else is mail, not demand. The article on turning contact form submissions into enquiries describes the path in detail.

Calls placed from the website

On a phone, a call is the shortest route from page to job. If the number is a tappable link and you ask briefly in the first conversation how the contact came about, you have the figure without any technical effort. In trades and hospitality it is often higher than the number of form submissions.

Form submissions per page

The distribution matters more than the total. If ninety percent of submissions come from the contact page and the three service pages contribute nothing, that is a finding about the service pages, not about the form.

Appointment requests

An appointment is already a preliminary stage of a job and therefore more meaningful than any click. 48 percent (Bitkom) of trade businesses now offer online appointment booking. How appointment requests fit into daily operations is covered in the article on appointment requests for service businesses.

Share of visits from a phone

This number does not decide success, it decides the order of work. If the share is two thirds, every change is checked on a phone first. The article on mobile usability for business websites describes what matters here.

Load time of the key pages

Measured on a mobile device, not on the office desktop inside your own network, and only for three to five pages: home, the strongest service page, contact. A reference value of 2.5 seconds until the largest visible element is a workable anchor.

Splitting these into rhythms matters more than it sounds. Check all six values monthly and you rebuild the same data graveyard, only smaller. Three of the numbers belong on the monthly sheet, two belong in a quarterly slot, one is reviewed twice a year. The overview below sorts the values by where they arise and which decision they actually trigger.

MetricWhere it arisesWhich decision it triggersRhythm
Qualified enquiriesInbox, phone list, calendarIs capacity utilised? Does the site have to do more?monthly
Calls from the websitePhone link, question at first contactIs the contact route reachable enough on a phone?monthly
Appointment requestsCalendar, callback listDoes the slot offering match actual demand?monthly
Form submissions per pageSubject tag or confirmation pageWhich page carries its weight, which is decoration?quarterly
Load time of the main pagesMeasurement on a mobile deviceDo images, embeds or delivery need work?quarterly
Share of visits from a phoneSite condition dataIs everything checked and maintained on a phone first?twice a year

Count enquiries where they arrive

The most reliable measuring point in a small business is not a tool, it is the inbox. What lands there is what actually happened: a message, a callback request, an appointment enquiry. Count those events once a month and you have a number that is subject to no extrapolation, no sampling and no estimate. That is a substantial difference, because analytics systems work with approximations and an inbox does not. In practice it takes a handful of decisions, made once and then held to.

  1. A dedicated email address or a fixed subject tag for everything arriving via the website. That way the channel can later be filtered without thinking about it.
  2. A phone list next to the handset or a note in the job book: date, service, one line on origin. Nothing more is needed.
  3. A fixed counting day each month, for example the first working day. Move the date and you move the comparison period, losing comparability.
  4. A written definition of what counts as qualified. It is noted once and stays the same, even in a weak month.
  5. A separate line for existing customers. Otherwise repeat business mixes into a number that is meant to describe new business.

What counts as a qualified enquiry

A workable definition has three parts: the requested service belongs to the offering, the location lies within the area served, and there is a way to reply. Job applications, cold sales mail, partnership offers and enquiries outside the catchment area are recorded separately, not deleted. They are a finding in their own right, because a high number of unsuitable enquiries points to unclear statements about services or service area on the site.

Origin can be captured without any technology, using a single question at first contact: how did you hear about us? Four rough categories are enough — search, recommendation, directory, and signage or vehicle. The answer is imprecise, and that is understood. But it is available, it costs nothing, and it covers exactly the cases that any technical measurement loses: the call placed after a glance at the site, the visit following a recommendation, the customer who copied the address out of a directory. In practice, after a few weeks around three quarters (project experience) of enquiries fall into one of the four categories, and no more precision than that is needed for the decision.

85 percent of trade businesses see a clear expectation of constant availability among their customers, and 87 percent observe markedly higher price sensitivity.

Bitkom, study report Digitalisation of the Skilled Trades 2025

That expectation explains why reachability weighs more heavily as a metric than reach. Firms that name 76 percent (Bitkom) time savings and 75 percent (Bitkom) higher visibility with customers as benefits of digital applications ultimately measure both against the same question: does the enquiry arrive, and does it arrive early enough to be answered? The article on generating enquiries as a tradesperson describes the typical breaking points along that path.

The monthly sheet: three numbers, one comparison month, one action

On a monthly rhythm, the list of six metrics becomes a sheet with three numbers. Not four, not eight. Three, because three values can be held in your head and because with three values the question of cause is still answerable. Add one comparison month and exactly one action for the coming four weeks. Take two actions at once and you cannot say next month which one worked, so you start from scratch again.

monthly-sheet.txt
Monthly sheet  <month / year>

Number 1  Qualified enquiries ......... ____   (prev. month ____)
Number 2  Calls from the website ...... ____   (prev. month ____)
Number 3  Appointment requests ........ ____   (prev. month ____)

Special circumstances this month
  Public holidays / school breaks ... ______________________
  Weather / season .................. ______________________
  Own campaigns ..................... ______________________

One action for next month
  What ............................. ______________________
  On which page .................... ______________________
  By when .......................... ______________________

The comparison month is the previous month, plus once a year the same month of the previous year. The year-on-year comparison is the more valuable of the two, because it removes seasonality: a roofing business in February and the same business in July operate in two different worlds, and so does a restaurant. Comparisons between individual days or weeks do not belong on this sheet, because the case numbers are too small. An example month makes the reading clear.

NumberJuneMayReading
Qualified enquiries1411Plus three. Within the usual fluctuation, no pressure to act.
Calls from the website97Plus two, second month in a row. The phone link near the top of the page is working.
Appointment requests55Unchanged while enquiries rise. This is where the action of the month applies.

In this example the action falls on the third line, not because the number dropped but because it stayed flat while the other two rose. That is the real contribution of the monthly sheet: it does not show peaks, it shows divergence between numbers that ought to move together. A fitting action here would be to make the appointment request visible as a route of its own alongside form and phone, and to count again the following month.

The rule in one sentence

Three numbers, one comparison month, one action. Anything beyond that increases effort faster than insight.

Derive targets from your job value

A number without a target is an observation, not a metric. And the target does not come from an industry benchmark, it comes from your own costing. The starting point is the annual cost of the website: the one-off investment spread over its useful life plus the running costs. Set against that is the contribution margin per job, not the revenue. Anyone who wants to read the path from cost to decision in more detail will find it in the article on one-off and ongoing website budgets.

target-value.txt
Annual cost of the website
  = one-off investment / useful life + running costs per year

Contribution margin per job
  = average job value x contribution margin rate

Jobs needed per year      = annual cost / contribution margin per job
Enquiries needed per year = jobs needed / conversion rate
Monthly target            = enquiries needed / 12

Example: 1,800 euros annual cost, 1,200 euros job value,
30 percent contribution margin, 40 percent conversion rate
  ->  5 jobs  ->  around 13 enquiries per year  ->  around 1 per month
Average job valueContribution margin (30 percent)Annual cost of 1,800 euros covered fromMonthly enquiry target (40 percent conversion)
400 euros120 euros15 jobsaround 3
1,200 euros360 euros5 jobsaround 1
4,000 euros1,200 euros2 jobsbelow 1
12,000 euros3,600 euros1 jobbelow 1

The fourth column is the real yield of this calculation. It translates annual costs into a number that sits on the monthly sheet and can be checked without reaching for a calculator. For small-ticket services the target is higher, but more enquiries are in circulation there anyway. For high job values it drops below one, which means: a single job per quarter carries the site, and the spread between months is so wide that the year-on-year comparison remains the only sensible yardstick.

Calculating with revenue makes the target look too easy

A job worth 1,200 euros does not cover annual costs of 1,200 euros, because material, labour time and travel are inside that figure. The calculation uses the contribution margin, that is, the share left after variable costs. If you do not know that rate, use a conservative placeholder and sharpen it later. Targets calculated too generously produce false comfort, and that is more expensive than an uncomfortable number.

Measurement period, season and confounders

Four weeks is the minimum before a change is assessed. The reason lies in the structure of demand: enquiries arrive in waves, tied to weekdays, pay days, school holidays and weather. A period shorter than a full monthly cycle mostly measures chance. For businesses with fewer than ten enquiries a month, eight weeks is the more dependable choice, even if the wait is uncomfortable. The alternative is not a faster insight but a wrong one.

  • Note public holidays and bridge days. A month with three public holidays has roughly fifteen percent fewer working days than a normal one.
  • Record school holidays in your own state, and for firms with private customers those of neighbouring states as well.
  • Log weather patterns if the service is weather-dependent. Roofing, gardening, outdoor dining and bicycle repair depend on it noticeably.
  • Enter your own campaigns: a newspaper article, a trade fair, a poster, a piece in the local newsletter.
  • Note company holidays and periods of illness, because during those weeks calls go unanswered and forms go unreplied.
  • Record technical outages, for example an unreachable page or a form that failed to deliver messages for a while.

A further confounder is automated traffic. A considerable share of requests to small sites comes not from people but from search engines, security scanners, preview services and language models reading content. For the visit count that is decisive; for the enquiry count it is irrelevant, because an automated request does not submit a form with meaningful text. That is precisely where the underrated advantage of enquiry-based measurement lies: it is inherently immune to this distortion, without a single filter rule to maintain.

The most common mistake

It is not the absence of data that costs decisions, it is reacting to periods that are too short. Rebuild after ten days and you are measuring weather and weekdays, not your own website.

Mobile share and load time: the two technical values

Two numbers describe condition rather than success. The share of visits from a phone sets the order of every check, and the load time of the key pages decides how many visitors reach the contact route at all. The link between loading behaviour and business results is well documented: one provider improved the value for loading the largest content element by 31 percent (web.dev, The business impact of Core Web Vitals) and recorded 8 percent (web.dev) more sales. In another case, an improvement of 70 percent (web.dev) led to 76 percent (web.dev) fewer abandoned page loads.

Load time to the largest element

The reference value is 2.5 seconds, measured on a mobile device over a mobile network. Check it only in the office over Wi-Fi and you are measuring your own connection. What sits behind this technically is described in the article on static delivery and PageSpeed.

Share of mobile visits

By sector the value ranges between roughly half and three quarters. It is not optimised, it is read: it determines which device every change is approved on first, and which font sizes and spacings are up for discussion at all.

Unreachable pages

A third, frequently forgotten value. A form that no longer delivers messages, or a subpage returning an error, has the effect of a closed shop with the lights on. The article on a maintenance routine for small firms sorts the check points.

Further evidence points the same way: halving load time coincided in one case with a 50 percent (web.dev) better bounce rate, and an improvement of 55 percent (web.dev) with a 23 percent (web.dev) longer average session duration. These figures come from considerably larger properties and cannot be transferred one to one to a firm with five pages. The direction is stable enough, however, to justify a simple rule: measure load time before polishing content, because a slow page devalues every text improvement.

What the performance report contributes and what it does not

Alongside your own counting there is a free source for visibility in search. The Search Console performance report provides four metrics (Google Search Central): clicks, impressions, click-through rate and average position. The data goes back 16 months (Google Search Central), the tables are limited to 1,000 rows (Google Search Central), and queries made by only a very small number of people within a few months are anonymised for privacy reasons and therefore omitted from the tables (Google Search Central). That is exactly where the differences between chart totals and table totals come from, which puzzle so many people.

Two levels that complement each other

The performance report describes the level before: what is being searched for, and is the site shown for it? The monthly sheet describes the level after: how many of those contacts turn into enquiries? Together the two levels form a chain in which causes can be assigned. Look only at the first level and you optimise visibility without jobs. Look only at the second and, when enquiries fail to appear, you cannot tell whether the business was found in the first place.

How to read that search data in detail, which patterns offer the greatest leverage and how it becomes a quarterly routine is covered in the article on evaluating search queries. For the monthly sheet, one single figure from that report is enough: the number of clicks compared with the previous month. Everything else belongs in the quarterly slot, because positions and click-through rates move more slowly than a monthly rhythm can depict anyway.

Measuring without personal data: legal basis and consent

Counting means processing data, so the principles of the General Data Protection Regulation apply. Article 5 requires purpose limitation, data minimisation and storage limitation, among other things (Regulation EU 2016/679, Article 5). That is easy to satisfy for a monthly sheet, because the three numbers on it are totals without any personal reference. The enquiries themselves are a different matter: an email with a name, a phone number and a request is processing of personal data and needs a legal basis under Article 6 (Regulation EU 2016/679, Article 6), as a rule the initiation or performance of a contract.

Personal data shall be adequate, relevant and limited to what is necessary in relation to the purposes for which they are processed.

General Data Protection Regulation, Article 5(1)(c)
  • The monthly sheet carries totals only. Names, addresses and phone numbers stay in the inbox and in the job system.
  • The origin question at first contact is noted as a category, not as a quotation. A tick next to “search” does not create a personal profile.
  • Reach data for the site is collected without cross-device recognition and without passing anything to third-party advertising networks.
  • Storing or reading information on a visitor's device requires consent under Section 25 of the German Telecommunications Digital Services Data Protection Act.
  • Deletion periods for enquiries are defined and observed, separated into commissioned and not commissioned.
  • The privacy notice describes what is measured, on which basis that happens and how long the data is kept.

The practical advantage of this design is bigger than it first appears. Measurement without personal data and without third-party advertising networks manages in many cases without consent-requiring access to the device, and correspondingly loses no data base to refusals in the consent dialogue. How such a dialogue is built to be both legally sound and economical is described in the article on cookie consent and the GDPR.

From finding to action in thirty minutes

The routine takes less time than costing a quote. It consists of five steps worked through in this order, and it ends with one sentence that sits in the calendar. What it does not contain is just as important: no new metrics, no discussion about tools and no second action.

  1. Count: go through the inbox, the phone list and the calendar and enter the three numbers. Ten minutes.
  2. Contextualise: note the month's special circumstances, that is holidays, school breaks, weather and your own campaigns. Five minutes.
  3. Compare: place the previous month beside it, and once a year the same month of the previous year as well. Five minutes.
  4. Explain: formulate one assumption per deviation, in a single sentence, without embellishment. Five minutes.
  5. Decide: name one action, record the page it applies to and set a date. Five minutes.

After three months this produces something no analytics package can deliver: a memory. You then know what an average month looks like in your business and recognise outliers without doing the arithmetic. That memory is the basis for the larger decisions, for instance the question of when a new website pays off, or the choice between a site builder, an agency and an AI-assisted build. Both decisions turn out considerably more sober with three monthly sheets in hand.

Reach and condition are built into XICflow

The four countable values arise inside the business, no software required. The two condition values — mobile share and loading behaviour of the key pages — are supplied by the platform itself in XICflow. Reach and condition data are part of every site, without an additional counting system to embed, maintain and cover contractually. That is less a question of convenience than of completeness: what does not have to be set up cannot be missing later. Which building blocks belong to it is shown in the overview of what XICflow includes.

  • Reach and condition data per site, without a separate counting system and without an additional contract.
  • Contact routes as building blocks: a form with delivery confirmation, a tappable phone number, appointment requests as a route of their own.
  • Static delivery as the basis for short load times on exactly the pages that appear on the monthly sheet.
  • Legal texts, consent and accessibility are set up in every generated site rather than bolted on afterwards.
  • Changes to texts and pages are possible without a service provider, so the action of the month takes effect within the same month.

If you want to see how contact routes are laid out in finished sites — form, tappable phone number and appointment request side by side rather than buried in body copy — the XICflow example sites offer several business types to click through. And if you want to bring in search-side metrics as well, the reading of the four values is set out in the article on evaluating search queries.

What counts in the end

A website is well evaluated when someone in the business can say, within thirty minutes, how many enquiries came in during the month, where they came from and what will be done differently next month. Everything else is supplementary information.

Sources and studies

This article is based on data from: the Google Search Central documentation on the Search Console performance report, in particular the four metrics clicks, impressions, click-through rate and average position, the 16-month data window, the row limit of the tables and the anonymisation of rarely used queries; the web.dev (Google) collection titled “The business impact of Core Web Vitals”, which brings together documented relationships between loading behaviour and business metrics from several case studies; the Bitkom study report “Digitalisation of the Skilled Trades”, based on a representatively weighted telephone survey of 504 skilled trade companies in Germany conducted in 2025; the survey by the Federal Statistical Office of Germany (Destatis) on the use of information and communication technologies in enterprises 2025, in particular the share of companies with their own website by size class; and the General Data Protection Regulation (Regulation EU 2016/679), specifically Article 5 on the principles of processing and Article 6 on lawfulness of processing, supplemented by Section 25 of the German Telecommunications Digital Services Data Protection Act on consent for access to terminal equipment.