Analytics for a small business: fewer charts, more answers

When I ask a small business owner how many inquiries their website brought in last month, I usually hear one of two answers. The first: “No idea, a few, I guess.” The second: “I’ve got Google Analytics,” and then it turns out the last time anyone looked at it was the day it was installed.

Both end the same way: you’re making decisions on gut feel. This piece is about measuring little, but measuring the right things. Every claim links to the tools’ own docs or to regulators, not to someone’s opinion. One caveat: the GDPR section describes the law and regulators’ positions. It isn’t legal advice.

Why pageviews lie to you

Analytics tools count a ton of things on their own. GA4’s enhanced measurement collects pageviews, scrolls to the bottom of the page, outbound link clicks, file downloads and form interactions without you lifting a finger. It looks impressive. The problem is that none of those numbers answers the question you actually care about: is the site bringing in customers?

A thousand pageviews a month could mean twenty jobs, or zero. Long time on page could mean someone is reading with interest, or that they can’t find your phone number. A number you can’t act on come Monday is just noise.

Events that mean money

Instead of measuring everything, pick a few moments where a customer takes a step toward paying you. Google calls these key events: actions that are “particularly important to the success of your business.” They used to be called conversions; at Google, “conversion” now means something you measure for ads.

For a service business, that’s usually:

• inquiry sent (a form, an email from a button);
• phone number tapped on mobile;
• call booked (a calendar, a callback);
• quote accepted;
• invoice paid.

GA4 even has a ready-made name for the first one: generate_lead, meaning a form submission or a request for information. PostHog and other tools let you name events whatever you like. The name matters less than keeping it single and consistent.

Try it yourself. On each card, pick how you’d measure it:

1. 01A customer submitted the inquiry form

   On-site eventServer-side eventA pageview is enoughDon’t track it

2. 02Someone tapped the phone number

   On-site eventServer-side eventA pageview is enoughDon’t track it

3. 03Someone opened the Contact page

   On-site eventServer-side eventA pageview is enoughDon’t track it

4. 04A customer accepted a quote

   On-site eventServer-side eventA pageview is enoughDon’t track it

5. 05An invoice was paid

   On-site eventServer-side eventA pageview is enoughDon’t track it

6. 06Someone scrolled to the bottom of the page

   On-site eventServer-side eventA pageview is enoughDon’t track it

7. 07The cursor hovered over the menu

   On-site eventServer-side eventA pageview is enoughDon’t track it

Pick an answer on each card.

The funnel: where people drop off

Once you have a few events, put them in order: visit, inquiry, call, job. That’s your funnel. PostHog’s docs put it simply: in any process, more people start than finish. The point is to see which step loses the most.

Here’s the counterintuitive part. One extra percentage point isn’t worth the same everywhere. Pick a step and add some points:

Where do you improve?

InquiryCallJob

By how many percentage points+1

1. Visitsper month 1,200
2. Inquiry2.5% of visits 30
3. Call60% of inquiries 18
4. Job35% of calls 6.3

As is: 6.3 jobs a month. The biggest leak is between a visit and an inquiry.

One more point on inquiries (from 2.5% to 3.5%) means 40% more jobs. The same point on calls (from 60% to 61%) is under 2%. Multiplication is ruthless: you gain the most where the rate is lowest.

What does that mean in practice? Before you pay for more traffic, check how many of your current visitors get in touch at all. If it’s 1–3%, it’s often cheaper to improve the site (a visible phone number, a shorter form, a “starting at” price) than to buy more visitors.

Events that happen off the site

The two most important steps, the quote and the payment, don’t happen on your website. They happen in email, on the phone, in your invoicing software, at the bank. A tool running in the browser will never see them.

The fix is to send events from the server. GA4 has the Measurement Protocol for this: a plain HTTP request that adds an event to your data. Google notes it’s meant to supplement browser tracking, not replace it. PostHog has server-side libraries for the same job, and Matomo has its HTTP Tracking API.

The key detail: a server-side event has to be linkable to the inquiry. The simplest way is to give every inquiry an ID when the form is saved and send it with every step after that. Then you know not just how many payments came in, but which inquiries they came from.

Phone calls are their own case. The site sees the tap on the number, but not the call. If most customers call you, log calls in your own system, or even a spreadsheet, and count the “call” step from there.

Cookies, consent and GDPR

This is where most of the myths live, so let’s take it one step at a time. Again: not legal advice, just what the law and regulators say. These are the rules that apply to a business in Poland.

The cookie rule isn’t only about cookies. In Poland, it’s Article 399 of the Electronic Communications Law (which replaced the old Telecommunications Law provision on November 10, 2024). It covers storing information on a user’s device and accessing it. Consent isn’t required only when that’s necessary to transmit a message or to provide a service the user asked for. The European Data Protection Board’s Guidelines 2/2023 (adopted October 7, 2024) make clear the scope also covers browser local storage, tracking pixels and URL-based tracking.

Analytics usually isn’t “necessary.” France is the exception: its regulator, the CNIL, exempts audience measurement tools from consent as long as they’re used only for that, produce only anonymous statistics, don’t combine data with anything else and don’t track people across sites. Poland has no such exemption and no specific guidance on analytics; Matomo’s overview lists it among the countries where it’s safer to assume consent is needed, or to assess the risk yourself.

So you have two honest options:

• A tool that stores nothing on the device. Plausible uses no cookies or browser storage; it counts unique visitors from a hash of the IP address, the browser and a salt that rotates daily, so it can’t link the same person across days. PostHog has a cookieless_mode option: no cookies, no browser storage, with the ID computed on the server. The price is accuracy: you won’t see that someone came back a week later.
• A full tool plus a consent banner. GA4 or PostHog with cookies, but only after consent. Keep in mind some people will decline, so your numbers will run low, and by different amounts on different devices.

Either way, GDPR still applies: IP addresses and identifiers are personal data, so you need a legal basis for processing, a section in your privacy policy and a data processing agreement with the tool’s vendor. If the data needs to stay in the EU, pick EU hosting; PostHog and Plausible both offer it, and you can run Matomo on your own server.

One more little-known thing about GA4: in the standard version you set the data retention period to 2 or 14 months, and it applies to exactly the explorations and funnel reports. Standard aggregated reports stay. If you want to compare your funnel year over year, check that setting today.

One view, four numbers, every week

The last step is the least technical and the one most often skipped. Tracking nobody looks at is worth as much as no tracking at all. So I suggest one view with three to five numbers, reviewed on a fixed day each week.

Something like this:

Week 42 ← →

• Inquiries 14 +1
• Calls booked 5 −4
• Quotes accepted 3 −1
• Payments 11,600 zł −700 zł

In week 42 there were fewer calls even though inquiries went up. That’s the one thing to look into on Monday.

A few rules that work:

• weekly, not daily. At a small business, daily numbers are mostly chance. A week smooths out the noise; a month is too long to react quickly;
• always the same order, from visit to payment. Read from the top and look for the spot where something slipped;
• the change next to the number. “5” on its own says nothing; “5, down from 9 last week” says a lot;
• one thing to look into. If you finish reading the view and don’t know what to do differently, it has too many numbers, or the wrong ones.

Any of these tools can build that view: in GA4 it’s a custom report or a funnel exploration, in PostHog a dashboard with a few charts, in Plausible and Matomo goals on the main dashboard. None of them know about payments, though, until you send them from the server or add them from your invoicing software.

Where to start this week

You don’t have to do it all at once. Start with one event: inquiry sent, counted after the form saves. After a month you’ll have your first rate, visits to inquiries, and that one usually tells you the most. Then add calls, then quotes and payments.

And one last tip: when you change something on the site, change one thing and give it a few weeks. With a dozen or so inquiries a week, a one-week difference is almost always chance.