The Missing Comparison
Why industry comparison fails in low-capacity administrations, and what should replace it
A trading business records its own sales, purchases and stock. Gross margin follows by arithmetic, and clusters by trade. A business far outside its band is not hiding — the anomaly sits in data the authority already holds. This study asks why it so rarely reads it.
The comparison not made
Each trade holds a characteristic gross-margin band. Filings arrive; only the few an analyst reaches are compared against it.
Overview
Commercial performance ought to be among the more legible things a revenue authority observes. Sales, purchases and stock movements are recorded by the business for its own management purposes, before any authority asks for them, and the relationship between them — gross profit as a percentage of sales — clusters industry by industry within a fairly predictable band.
A wholesaler whose margin sits far below the wholesale norm is not concealing the fact. The anomaly is present in data the authority already holds, and the arithmetic that reveals it is not complicated. It should therefore be comparatively easy to notice. This study asks why, in practice, it so rarely is.
The answer proposed here is that the failure is not one of law, rate, or enforcement will. It is a failure of comparison. Administrations hold the transactional data from which a benchmark follows, and know in general terms what normal looks like in a given trade — yet lack any governed mechanism for setting the second against the first at the granularity at which the anomaly actually lives: the individual taxpayer within an individual product classification.
The central claim
The compliance risk foregone through weak industry comparison is, in the main, not risk that has been assessed and excused. It is risk that has never been measured — and unmeasured risk is recoverable through administrative design, whereas a case an officer has examined and set aside is a matter of judgement.
Diagnostic framework
Process reconstruction of the as-is triage cycle, failure-mode decomposition against a problem tree, and design prescription derived by inverting the root condition.
Key diagnostic insights
- 01 Each comparison requires its own extraction, cleaning and formula design, rebuilt by hand — so coverage silently contracts to what one analyst can manage, and no one records what was left unexamined.
- 02 Analysts select different comparison groups and tolerances, none documented. The same business could be flagged by one officer and passed by another, for reasons unconnected to its actual performance.
- 03 A modest divergence across many taxpayers in one trade amounts to more than a large divergence in one. Sampling is tuned to the conspicuous outlier; a uniform two-point shortfall across a sector attracts no attention at all.
The root condition
Industry comparison is treated as an ad hoc analytical exercise, performed by individuals upon artefacts the institution does not control, against standards the institution has not specified and cannot reproduce — rather than as a governed computational process for which the institution is accountable.
The reference design
The prescription is organised around a single invariant, from which the architecture follows.
The invariant
What the taxpayer reported is never overwritten. What the approved standard specifies is written alongside it, never over it. The divergence between them, and the classification that follows, are themselves records — not transient calculations. Triage is the act of interpreting that divergence, not of producing it, and never of concluding from it alone.
The three-record model
Reported · D
What the taxpayer stated
The business's own figures and the margin computed from them, immutable after intake. Cannot be altered by the authority under any circumstance.
Reference · C
The standard applied
The approved benchmark for this classification, selected by governed rule, with the reference-set version stamped on the record.
Divergence · V
The signal
V = D − C, retained as a record rather than derived on request, with its classification under the tolerance then in force.
What should be measured
An implementation should be evaluated against outcomes rather than delivery milestones. Each indicator below is observable, and each degrades visibly when the design is failing.
| Indicator | What it measures | Healthy trend |
|---|---|---|
| Standard convergence | Whether comparable taxpayers now receive comparable treatment. | Outcomes converge over time rather than depending on which analyst handled the case. |
| Selection provenance | Whether audit effort is directed by measured divergence or by other means. | A rising share of examinations originate from classified cases — and the share is known rather than assumed. |
| Rejection rate | Data quality at source. | A rise on introduction as the gate begins to bite, then sustained decline. |
| Confirmation rate | Whether classifications survive officer review. | Stable and moderate. Very high suggests thresholds set too conservatively; very low suggests the reference set needs revision. |
| Traceability | Whether any classification can be reconstructed on demand. | Complete, and tested by reproducing a historical case at random. |
Study reference & lineage
Unit of analysis: The compliance-risk triage cycle of a national revenue authority.
Method: Process reconstruction, failure-mode decomposition, design prescription.
Applicability: Revenue administrations in developing and transitional economies.
This study advances a general argument about administrative design. It does not describe, endorse, or evaluate any commercially available system.
© 2026 The Missing Comparison — concept study prepared by Sujoy Maitra.