Multi-Year Averaging: The 1-Year vs 3-Year Data Choice
Multi-year averaging defined: the choice between the single-year and the multi-year (3-year) comparables data — the simple, weighted and per-period methods and the consistency discipline.
Definition
Multi-year averaging is the choice of the data period the comparables’ PLIs are computed on: the single year (the current year’s values, one PLI per member) against the multi-year (typically the three-year — each member’s PLI averaged over the periods, on the stated method). The averaging smooths the one-year noise (the one-time item, the cycle’s trough, the member’s single-year anomaly) into the member’s structural position — and the guide has the methods and the edge cases in full. The glossary’s content is the choice’s shape and the consistency discipline that makes it defensible.
| The method | The computation | The use |
|---|---|---|
| The single year | One PLI per member, on the current year’s values | The method where the current year is the study year and the one-year data is the stated basis (the trend is monitored separately, the multi-year data trend read) |
| The simple average (3-year) | Each member’s PLI = the mean of the three years’ PLIs (the unweighted) | The standard multi-year method — the three years’ noise averaged, the member’s structural position |
| The weighted average (3-year) | Each member’s PLI = the mean weighted by the base (the revenue, the costs — the weight per year) | The variant where the years’ scale differs (the growth year, the contraction year) — the weight is the base, stated |
| The per-period (the trend) | The PLIs kept per year, the trend read (the direction, the cycle) — the range per period, or the averaged range with the trend note | The method where the trend is the comparability question (the tested party’s trajectory vs the pool’s) — the averaging with the trend preserved |
The consistency discipline (the tested party selection and the PLI reference): the averaging method is applied consistently to the tested party and the pool — the tested party’s PLI averaged on the same method as the pool’s (the tested party on the 3-year simple average, the pool on the single year, is a comparison the method does not support), and the method is stated (the periods, the weights, the trend treatment) in the PLI block of the file. The working read: the averaging masks the trend risk where the tested party’s trajectory differs from the pool’s (the tested party declining, the pool stable — the averaged PLI hides the direction) — the per-period read (the trend) is the supplement where the trajectory is the question.
Example
The study runs on the 3-year simple average: each of the 12 pool members’ OP/S is computed on each of the three years, then averaged (the unweighted mean) — the member’s structural OP/S, the one-year noise smoothed. The tested party’s OP/S is averaged on the same method (the three years, the unweighted mean) — the consistency stated. The pool’s distribution (the 12 averaged values): the IQR 2.1%–3.4%, the median 2.8%. The tested party’s averaged OP/S is 2.9% — inside, 0.1 above the mid-point. The trend note (the per-period read): the tested party’s OP/S is 3.2% → 2.9% → 2.6% (the decline), the pool’s is stable (2.8% → 2.9% → 2.8%) — the trajectory difference is noted in the file (the trend risk, the documentation’s read), the range is on the averaged values, stated.
See also
FAQ
Single year or 3-year — which is the default? The 3-year (the multi-year) is the practice default for TNMM — the one-year noise (the one-time item, the cycle) is the comparability risk the averaging removes, and the OECD’s guidance (the OECD guidelines position) is the multi-year data where available. The single year is the stated variant (where the multi-year data is not available, or the current year is materially different on the documented facts) — the choice is stated in the file either way, with the rationale.
Simple, weighted or per-period — how is the method chosen? On the data’s character: the years’ scale similar → the simple average (the unweighted, the standard); the years’ scale differing (the growth, the contraction) → the weighted (the base’s weight, stated); the trend the comparability question (the tested party’s trajectory vs the pool’s) → the per-period (the trend preserved, the range per period or the averaged range with the trend note). The guide has the method-by-method mechanics and the edge cases (the missing year, the one-time year, the method change).
Does the averaging have to match the tested party’s period? Yes — the consistency discipline: the tested party’s PLI is averaged on the same method as the pool’s (the same periods, the same weights, the same trend treatment) — the comparison is the averaged tested party against the averaged pool, on the stated method. The tested party on the single year, the pool on the 3-year average, is a comparison the method does not support — the TPO’s and the appeal’s first question on the data period. See the multi-year data guide’s consistency section.
Run the screens as a study, not a spreadsheet
Quartyl applies the method, PLI and screening steps above as a pipeline — and keeps a documented reason for every exclusion.
Related docs
Multi-Year Averaging in TNMM: Simple, Weighted and Period Methods
Single-year vs multi-year data in TNMM: when the OECD requires looking beyond one year, how simple and weighted averages are computed, and when averaging masks a trend.
Read docTested Party: Definition, Selection Logic and Documentation
The tested party defined: the entity whose result is benchmarked against the comparable pool — selected as the least complex party, and documented as a decision.
Read doc