The summary risk assessment
An anomaly is not yet a finding.
An analytical procedure used by the tax administration to examine digital data for anomalies — not a form of audit laid down by statute and not a means of evidence in its own right. It does not appear in the Abgabenordnung, the German fiscal code. What it produces are statistical anomalies. Whether a power to estimate follows from them is not decided by the procedure but by § 158 Abs. 2 Nr. 1 AO: whether the circumstances of the individual case give cause to object to substantive accuracy.
The difference between an anomaly and a finding
A statistical procedure always answers the same question: how likely is this pattern under a given assumption? It does not answer whether the assumption holds for this business. A departure from the expected value is therefore a reason to look more closely — it is not a finding about the accuracy of the records. The step from anomaly to finding has to be reasoned separately.
What the result must be measured against
The provisions in this table are German law. Most European legal orders know the same principle — set out verbatim and by country in the register of legal bases.
| Yardstick | Legal basis | What it means |
|---|---|---|
| Precedence of the accounts | § 158 Abs. 1 AO | Proper accounting shall be taken as the basis. |
| The threshold | § 158 Abs. 2 Nr. 1 AO | Only in so far as there is cause to object to substantive accuracy. |
| The scope | § 162 Abs. 1 Satz 1 AO | Estimation only in so far as nothing can be established or calculated. |
| All circumstances | § 162 Abs. 1 Satz 2 AO | Including those that explain a statistical anomaly. |
| Favourable circumstances too | § 88 Abs. 1 Satz 2 AO | The authority must also take the favourable circumstances into account. |
| The Czech Republic for comparison | § 92 Abs. 5 lit. c daňového řádu | There the statute says it expressly: the authority proves the facts that disprove the completeness of the records. |
Why the explanation usually lies in the business
Statistical expected values assume an even course of business. Real businesses are not even: opening hours change, menus change, staff change, tills are replaced, payment methods shift. Almost every anomaly has an operational counterpart — it only has to be found and evidenced. That is work on facts, not statistics.
And where the objection does not carry
Declaring a statistical procedure unfit in general does not get far — as a selection instrument it is not meant to achieve anything other than finding cases that deserve a closer look. The effective objection is not aimed at the procedure but at the inference: that an anomaly already gives rise to a power to estimate. § 158 Abs. 2 Nr. 1 AO requires that inference to be reasoned separately.
What we contribute
We go through every anomaly individually and look for its operational cause in the data. For every point we record whether it can be explained or whether it stands — including where it stands.
The legal provisions cited here are given for information and do not replace legal advice.