Why your analytics dashboard is misleading you — and what SMBs should watch

Reading time: 6 minutes · For owners of SMBs in the March region, on the Obersee and around Lake Zurich

Your dashboard does not lie with wrong numbers. It lies by staying silent. GA4, your online shop, your accounting tool — they all show you clean totals and reassuring trends. And they leave out exactly the one number that would actually change your next decision.

This is not a flaw of one particular product; it is a property of almost every standard report: it summarizes. And whatever gets summarized disappears. In this post I will show you, using a real case, how dangerous this quiet incompleteness is — and how a single afternoon is enough to find out what your dashboard is keeping from you.

The dashboard answers nine of your ten questions

And that is exactly the problem: a tool that answers nine of your ten questions will happily let you forget you ever asked the tenth. The dashboard looks complete — and that impression is its most dangerous property.

An example from my own work (anonymized): a business was convinced — “we mostly sell bundles”, packages of several products. The whole pricing and advertising strategy was built around it. The dashboard showed order totals and conversion rate, but nowhere the number of items per order. The standard connector between shop and reporting simply did not carry the product level across.

One focused look at the raw data — around 320 orders — showed: 1.14 items per order. Bundles were not the rule but the rare exception. An assumption that pricing and advertising had been built on for months, overturned in a single afternoon. No new tool, no migration — one query in the right place.

Why this hits SMBs hardest

Large companies have data teams that find gaps like this sooner or later. In an SMB that corrective does not exist: what the dashboard shows is what counts. Standard connectors pipe totals into the dashboard and drop the detail level — individual products and services, customer segments, repeat buyers. You see confident, complete-looking numbers, with no hint that whole dimensions remain invisible.

A dashboard is rarely wrong. It is just incomplete — and that is more dangerous.

The same pattern exists well beyond web analytics, by the way: the “post” button in social media only tells you half the truth too — more on that in the post on the kitchen-party strategy.

Three questions before any metric

The fix is not a new tool and not a big BI project. It is a habit of mind. Before any number you want to base a decision on, ask three questions:

• What does this number not show me? Every total hides a distribution. “Revenue per month” says nothing about which service brings it in.
• Where is the missing information? Usually closer than you think: in the shop’s CSV export, in a standard accounting report, in the GA4 raw data.
• Would it change my decision? If yes, get the number before you decide. If no, skip the exercise.

An afternoon of data archaeology: a fiduciary example

Take a fiduciary firm in Pfäffikon SZ. The website dashboard says: visitor numbers rising, contact page getting clicks, all green. What it does not say: which service the enquiries are actually about. An afternoon with the email inbox and the client list paints a different picture: most new enquiries concern private tax returns — small-ticket, seasonal, low-margin. The lucrative bookkeeping mandates for corporate clients come in almost exclusively through referrals. So the website “works” — but for the wrong business.

That one insight flips the priority list: homepage, service pages and the Google presence get realigned towards corporate clients, and the report now tracks “enquiries from corporate clients” instead of “visitors”. None of the big numbers on the dashboard would have triggered that decision — only the level beneath them. How I work on questions like this with local businesses: see the marketing consultant Pfäffikon page.

The numbers SMB dashboards typically keep quiet about

Working with Swiss SMBs, the same five blind spots come up again and again:

• Items or services per order: do you really sell packages — or, in truth, single services?
• Repeat-buyer share: how much revenue comes from existing customers? This number is missing from almost every standard dashboard.
• Where the enquiries that became orders came from: not visitor sources, but the path of the customers you actually won.
• The product or service level: totals look complete — the distribution underneath tells the real story.
• Contribution margin instead of revenue: your biggest revenue driver is not always your biggest profit driver.

Data does not replace your gut — it challenges it

The point is not to distrust your gut. The bundle business had a plausible story — it just was not true. The gut poses the hypothesis; the raw data delivers the proof or the refutation. That division of labour is what good decisions are made of. And an afternoon of data work like this can be sped up considerably with AI support — how, you can read in How AI takes the load off an SMB back office.

The rare skill is not setting up dashboards. It is recognizing when a report is quietly incomplete — and having the discipline to look beneath the comfortable summary. That is part of marketing leadership that takes ownership of the numbers; what that collaboration looks like is on the Services page.

Frequently asked questions

Do I need a new analytics tool for this?

No. In almost every case the missing number already sits in your systems — in the shop, in the accounting software, in the inbox. You do not need new software; you need the right question and an export.

How often should an SMB question its metrics?

Once a quarter is usually enough: one fixed afternoon in which you test a central assumption of your business against the raw data. A manageable effort with a big lever.

What if nobody in the company has time for data work?

Then bring in an outside pair of eyes. Many businesses between Lachen, Pfäffikon and Rapperswil are in exactly that position — the answers sit in their own data, only the time to dig them out is missing. Half a day of external support per quarter is often enough.