What the consent banner actually changes in the numbers
In Germany, § 25 TDDDG requires consent before a website stores or reads information on a user’s device, unless this is strictly necessary. Conventional analytics tools set exactly that kind of identifier. Anyone who declines in the banner is therefore not recorded, or only partially. From that moment, the tool sees just a slice of the actual traffic.
What matters is not how large that slice is, but how it comes about. Consent is not randomly distributed. It depends on the device, the browser, the entry channel and how familiar someone is with the brand. On top of that, some browsers restrict cookies by default, and ad blockers stop analytics scripts from loading at all.
The result is not an even undercount but a skewed sample. Some metrics survive this well, others become useless. Which ones depends mainly on where the number comes from.
Web analytics without cookies starts with the data you already have
The most reliable numbers are not produced in the analytics tool at all. Orders, revenue, returns and enquiries are recorded in the shop system, the inventory system or the CRM, where customers and transactions are managed. These systems count every transaction, regardless of what anyone clicked in the banner.
Absolute business figures. How many orders came in, the average basket value and which products sold are all fully recorded in the back end. The consent requirement has changed nothing here.
Proportions within the recorded visitors. Which pages are viewed more often relative to others, where visitors drop off and which content is read for longer can still be read from the sample. As long as the skew affects all pages similarly, the ranking stays usable. To decide which product page to rework first, that is usually enough.
Trends under stable conditions. If banner, tool and settings stay the same for months, changes over time are meaningful. The absolute figure is wrong, but its direction is right. In practice, this condition is broken more often than people assume.
One simple test helps with every metric: can it be determined without the analytics tool? If so, the answer should come from the back end.
Which metrics are systematically skewed
The trouble starts with numbers that claim absolute quantities of people or attribute causes. With these, the skew in the sample feeds straight into the result.
Visitors and users. The number of “users” in the analytics tool is no longer a count of people. It shows how many consented and were recognised. Someone who consents but uses a browser with short cookie lifetimes is counted again on later visits. New and returning visitors can hardly be told apart any more.
Channel attribution. If consent varies by channel, the weighting of channels in the report shifts. Some visits also end up as “direct” or “unassigned” because their origin is lost. A channel can look weaker than it is, and budget moves in the wrong direction.
Conversion rate from mixed sources. This is an error that often goes unnoticed. Dividing orders from the shop by sessions from the analytics tool puts a complete figure against an incomplete one. The rate comes out too high, and the more visitors decline, the stronger the effect. Numerator and denominator must come from the same source.
Segment comparisons and tests. Comparing mobile with desktop, or one country with another, also compares consent rates. An A/B test whose variants trigger different consent rates ends up measuring something other than intended.
Why a better banner looks like growth
One of the most misleading developments in analytics reports has nothing to do with the business. When the banner is redesigned, its text reworded or its buttons reordered, the consent rate changes. The tool then shows more or fewer visitors, although actual traffic has stayed the same.
A typical sequence: a banner update goes live on a Tuesday, and marketing is not told. Two weeks later, the monthly report shows noticeably more visitors. The increase is credited to the campaign that happened to run at the same time. Orders in the shop stay flat, and nobody notices the contradiction.
The same happens when the consent tool is replaced, the tag manager is updated or browsers change their rules. Anyone who wants to read trends therefore needs a change log recording the date and content of each adjustment. Many analytics tools allow such notes directly on the timeline. Without them, every trend line is hard to interpret.
What Consent Mode does and what it only estimates
Google’s Consent Mode, an interface that tells Google tags the visitor’s consent status, is often seen as the fix for the gap. It does not close it. It makes tags behave differently depending on consent. In the advanced variant, they send cookieless signals even after a refusal, from which Google models the missing values.
These modelled values are projections, not measurements. They assume that visitors who decline behave much like those who consent. With a skewed sample, that assumption is questionable. Google also requires minimum volumes of daily users and events before modelling applies. Many small websites never reach these thresholds.
Legally, too, the advanced mode is not a given. Whether signals may be sent without consent is assessed differently and belongs in the data protection review. Anyone running Google Ads in the European Economic Area now needs Consent Mode for certain measurement and audience features. That is a reason to set it up. It is not a reason to trust its numbers blindly.
When web analytics without cookies or server-side tracking is not worth it
Two responses to the gap are frequently recommended. Cookieless analytics tools count page views without setting persistent identifiers. Server-side tracking routes measurement data through your own server before it reaches analytics or advertising services. Both have limits that product descriptions rarely mention.
Cookieless does not automatically mean consent-free. § 25 TDDDG covers any access to the device, not just cookies. Whether a tool can do without consent depends on what it actually reads, calculates and stores. That calls for an individual assessment; it is not a product feature. Such tools also usually cannot recognise returning visitors. They deliver reach, but not journeys across several visits.
Server-side does not mean complete. Your own tracking server partly bypasses ad blockers and can make identifiers more stable. It changes nothing about the consent requirement. In return, it brings its own operations: hosting, updates, monitoring and someone who understands the data flow. If the server fails, nobody may notice until the report stays empty.
For a small business, neither usually pays off when three things coincide: little traffic, a small advertising budget and decisions that are made from back-end figures anyway. More precise web analytics then does not lead to better decisions. It is different when a substantial ad budget depends on conversion signals. There, a clean server-side data flow can genuinely improve how campaigns are steered.
What a robust tracking setup looks like for a small business
A sound setup starts by asking which decisions are actually based on numbers. For many small businesses there are only a few: which channels get budget, which pages are reworked, whether a change had an effect. For each of these decisions, the right source can be defined.
Business figures from the back end. Revenue, orders and enquiries are read where they originate.
Origin through your own signals. UTM parameters, additions to campaign links that name the source, can be stored with the order if the shop system supports it. An optional field asking “How did you hear about us” on the enquiry form often gives more honest answers than any attribution in the tool.
Analytics tool for proportions. It answers questions about rankings and flows on the website, not about absolute quantities.
A log of changes. Every adjustment to banner, tags or tools is recorded with its date.
A monthly comparison is enough as a check. If orders in the back end and conversions in the tool move in the same direction, the measurement is intact. If they diverge, it is usually the measurement that is broken, not the business.
After the switch, measurement remains a maintenance task
A tracking setup is not finished once it is configured. Browsers change their rules, consent tools are updated, advertising platforms change their requirements, and new pages need their own measurement points. Once the sources have been cleanly separated, such breaks are spotted sooner, and decisions rest on numbers that hold.
How we build and look after websites and shops technically is described under Services.