• Shopware Gold Partner & Shopify Partner

Chapter 3

The truth about data sources: Where your numbers should really come from

Table of contents

A metric only helps you if you know what it means and where it comes from. Most stores do not have one central reporting system but five sources that partly contradict each other. What matters is not which source is "right" but which one leads for which question.

The five data sources of an online store

1. The store backend. Shopware, Shopify or another system. Every order appears here exactly once, with order value, items and customer number – regardless of cookie banner, browser and marketing tracking. Nothing is estimated, modeled or distributed proportionally. For revenue actually realized, payment status, cancellations and refunds are reconciled.

2. ERP and accounting. Inventory management, your accounting system, the monthly management accounts from your accountant. This is where reality's corrections come in: purchase prices, returns, credit notes, payment fees, shipping costs. What is left here is the real return. For the closed net revenue of a month, reconciled accounting is the authoritative figure.

3. Google Search Console and SEO tools such as SISTRIX. Organic visibility, rankings, impressions and clicks. Both work independently of the cookie banner and deliver reliable data.

4. The ad platforms. Google Ads, Meta, Microsoft Ads. What you spent is exact. The conversions attributed to you are attribution, not proof of additional purchases (see Chapter 2).

5. Google Analytics 4. The tool many regard as "the" data source. For behavior and trends it is valuable; for absolute numbers it is the most overrated source.

Why GA4 only sees part of your business

Three effects determine how much of your store GA4 sees:

  • The cookie banner: depending on design and audience, between a third and more than half of visitors decline in many stores. These visitors are missing from GA4 entirely or are estimated via Consent Mode.

  • Browsers and ad blockers: Safari expires cookies set via JavaScript after seven days at most. Anyone who does not return within that time counts as a new visitor on their next visit – longer purchase paths break apart. Ad blockers stop the GA4 script from loading at all.

  • Technical errors: script errors, loading problems and faulty integrations after updates. They are often only noticed when someone reconciles the numbers with the store.

How large the gap can become is shown by a study Orbit Media carried out on its own website: for visitors who did not see a banner, GA4 was missing 15.8 percent of traffic; for visitors with a banner, 55.6 percent. That is a single case, but it shows the order of magnitude.

For purchases, the gap is often smaller than for traffic: people who buy consent more often, and many stores also send purchases server-side. Even so, purchase measurement in GA4 is almost never complete.

Does that make GA4 worthless? No. GA4 is good for trends, behavior and channel comparisons, poor for absolute numbers. As long as banner, audience and setup stay the same, trends are reliable. Anyone who reads GA4 revenue as their actual revenue is comparing a partial measurement with reality.

The trap of mixing sources

An obvious thought: orders from the store are complete, sessions only exist in GA4 – so divide one by the other. That is exactly what produces a fantasy number.

A simplified model: in reality, your store has 100,000 sessions and 2,000 orders, a conversion rate of 2 percent. GA4 captures 60,000 of those sessions and 1,200 purchases – also 2 percent. If you now divide all 2,000 store orders by the 60,000 GA4 sessions, you suddenly get 3.33 percent. Your store has not sold a single extra item.

The rule that follows: the numerator and denominator of a ratio must come from the same measurement basis. Calculate the conversion rate with purchases and sessions from GA4. Use the orders from the store to check how completely GA4 captures purchases.

The data source pyramid

The pyramid ranks the sources by how independent they are of tracking, consent and attribution:

Level

Source

Strong for

Depends on cookie banner

1

Store backend

Orders, order value, new and returning customers, repeat purchases, customer value

no

2

ERP / accounting

Realized net revenue, cost of goods sold, returns, fees, contribution margin, inventory

no

3

Search Console + SISTRIX

Visibility, rankings, search clicks

no

4

Ad platforms

Ad spend (exact); attributed conversions – no proof of additional purchases

partly

5

GA4

Sessions, behavior, conversion rate, channel mix, drop-offs

strongly

What is in the store and in accounting is reality. What is in GA4 is a partial measurement of that reality.

This does not mean a rigid ranking for every number, but clear ownership: define a leading source for every metric. Orders from the store, closed net revenue from accounting, search clicks from Search Console, ad spend from the platforms, conversion rate and behavior from GA4. Settle this once and you end the "So which number is right?" discussion in every meeting.

Three questions for every reporting meeting

  • Which source does this number come from – and is that the leading source for this metric?

  • Do two differing sources really measure the same thing – with the same revenue definition and the same period?

  • Which share is measured, which is attributed or estimated?

These three questions take two minutes. They prevent a large share of the wrong decisions described in Chapter 1.

How to make the measurement in your store more reliable is covered in the guide Server-Side Tracking for Shopware and Shopify. How arboro handles the implementation is explained on the page server-side tracking.

What you can check today

  1. For the last 30 days, compare the orders in the store backend with the purchases in GA4. How large is the difference?

  2. Match order numbers with the transaction IDs in GA4. This shows you which purchases are missing, appear twice or are sent with the wrong values.

  3. In your last monthly report, write next to each number which of the five sources it comes from.

  4. Check whether ratios such as the conversion rate are calculated from one measurement basis – or whether store orders are divided by GA4 sessions.

  5. Define a leading source for every metric in the report.

Frequently asked questions

For orders the store backend, for realized net revenue reconciled accounting. Both are independent of cookie banner, browser and attribution. For search clicks, Search Console leads; for ad spend, the ad platform.

Because GA4 only shows purchases that are sent via the configured tracking or added through modeling. Declined consent, ad blockers, browser restrictions and technical errors mean that this is not your full store revenue.

No. The numerator and denominator then come from measurements of different completeness, and the conversion rate looks artificially high. Always calculate ratios from one measurement basis.

Yes, for trends, behavior and channel comparisons. As long as the banner and setup stay the same, trends are meaningful. For absolute revenue figures, the store backend and accounting apply.

No. Search Console shows impressions and clicks directly from Google Search, regardless of whether visitors consent on your website.