12 Common Benchmarking Mistakes (Self-Audit Checklist)

The mistakes below are not exotic. They are the recurring failures in benchmarking files examined by the TPO and the IRS — the errors that turn a sound study into an adjustment, usually at the step where the authority rebuilds the pool and finds the study’s choices were made somewhere else, or not at all. Use the table as a pre-filing self-audit: each row is a question the file should be able to answer with evidence.

#	Mistake	Consequence in examination	The fix
1	Wrong industry code — the tested party mapped to the registered activity instead of the actual function, or at a level too broad	The TPO rebuilds the pool on the function’s code; the pool is different; the range moves	Pin the 5-digit NIC code to the function (see search design); document the basis; widen stepwise only with a stated reason
2	Database ratios used as-is — the pool’s PLIs taken from the database’s own definitions, not rebuilt on the study’s operating lines	Every comparable’s PLI is off by the definition gap; the whole range is a construction the TPO can recompute	Pull the lines, rebuild the PLI on the study’s definitions for tested party and pool identically; document the mapping
3	Mixed PLI definitions — the tested party on one operating-cost definition, the pool on another (or the PLI changed mid-study)	The comparison is apples to oranges; the TPO re-runs on a single definition and the position changes	One cost definition, used everywhere; stated in the Local File; consistent across years
4	Related-party revenue leakage — comparables with significant affiliate revenue enter the pool as “independent”	The comparable’s economics are the group’s economics, not the market’s; the pool is contaminated	Screen related-party revenue from the notes; reject (or segment-adjust) with the reason logged
5	Consolidated financials for a segment function — a conglomerate’s group totals used as a comparable for a single function	The function’s PLI is buried in the group’s mix; the comparable is not comparable	Use segment data where disclosed; where absent, reject with “function not separable from group”
6	Loss-makers handled silently — excluded without the categorization, or included where the PLI does not compute	The pool’s before/after is unexplained; the TPO’s reconstruction differs from the file’s	The loss-maker framework: count, categorize, treat, show both distributions
7	No working capital adjustment — or the adjustment applied to the pool but not the tested party (or vice versa)	The PLI gap includes a working-capital difference that is not a functional difference	The WC adjustment on both sides, same tax rate, same period; the calculation in the file
8	Range shopping across years — the PLI, the pool or the period changed year to year to keep the tested party inside	The trend analysis shows the movement tracks the tested party’s position, not the economics	A stable method across years; where the tested party’s function genuinely changed, document the change as the reason
9	Size gaps left in the pool — a comparable at 10× the tested party’s scale accepted without comment	Scale economies (or the absence of them) distort the PLI; the TPO drops the outlier and the range tightens	The size screen with stated bounds; surviving gaps accepted with a reason or adjusted
10	The qualitative screen as a formality — every database company accepted on the name and the code, without the function/product/risk review	The pool contains adjacent-function companies the TPO identifies one by one; the matrix credibility collapses	The company-by-company review on the comparability factors — the qualitative screen is the study’s core defence, not its paperwork
11	Outliers left in without analysis — an extreme PLI in the pool accepted because “it is in the database”	The outlier stretches the quartiles; the TPO’s reconstruction excludes it and the range shifts	Identify the extremes, examine the reason (one-off, data error, genuine structure), and document the keep/exclude decision
12	The matrix without reasons — accepts and rejects without the specific, evidence-grounded rationale per company	The matrix is a list, not a method; the TPO (or the IRS) re-runs the screen and rejects the whole pool on credibility	Every disposition carries its reason, grounded in the evidence seen; overrides carry their own rationale — see defending the matrix

The self-audit, in order

Work the list top to bottom against the study file:

1. The code — is the tested party’s NIC/NACE code the function’s code, and is the basis stated? (row 1)
2. The lines — are the PLIs rebuilt from the lines on one definition, for tested party and pool? (rows 2, 3)
3. The contamination — related-party revenue and segment leakage screened out, with the reasons? (rows 4, 5)
4. The loss-makers — counted, categorized, treated, shown before/after? (row 6)
5. The adjustment — working capital on both sides, computed in the file? (row 7)
6. The stability — same PLI, same pool logic, same period across years, with any change explained by a functional change? (row 8)
7. The size — bounds stated, surviving gaps reasoned? (row 9)
8. The review — every company examined on the factors, not the name? (row 10)
9. The extremes — outliers examined and decided? (row 11)
10. The reasons — every disposition, including every override, carrying its grounded rationale? (row 12)

A study that passes all ten questions has a file that can be rebuilt by the examiner and come out the same — which is the entire standard of defensibility. A study that fails any of them knows, before the TPO does, where the adjustment will come from.

See also

• Defending the Accept-Reject matrix
• Quantitative screening
• Qualitative screening
• The benchmarking pillar guide