Qualitative Screening: The Company-by-Company Review

A Transfer Pricing Officer does not re-run your quantitative filters. They read your notes. That is why the qualitative screen is the heart of defensibility: it is the only part of the study where a reviewer is simultaneously checking the substance (is this company comparable?) and the credibility (did anyone actually look?). Boilerplate notes fail both checks at once.

What qualitative screening is — and is not

It is a company-by-company factual review of every candidate that survived the quantitative screen, against the tested party’s functional analysis. It is not:

• a second pass of the ratios (that stage is done);
• a legal opinion on comparability (it is the evidence an opinion rests on);
• a judgment about whether the company is a “good” comparable (the verdict follows from the facts, stated per comparability factor).

The output is the Accept-Reject matrix: every surviving candidate with a disposition and a reason, written at review time.

The five things to verify per company

#	Verify	What good looks like	Red flags
1	Product / service line	The company’s core offering is the same kind of offering as the tested party’s, at the same level of specialisation	A product the tested party does not make; a different value chain position (manufacturer vs distributor)
2	Revenue mix	A high share of revenue is in the comparable line (practice: ≥ ~80%)	The in-scope line is a minor segment; the entity-level PLI reflects the whole conglomerate
3	Customers & market	Same customer type (B2B/B2C), same geography, comparable concentration	Serves a captive single group; retail where the tested party is institutional
4	Assets & scale	Tangible asset profile consistent with the tested party’s function	Heavy capital assets behind a “services” revenue line; headcount profile implying different operations
5	Geography & operations	Operations in the same or comparable jurisdictions; no unexplained cost-base differences	A cost base driven by a different wage/inflation environment with no adjustment in sight

Two of the five failing is usually a reject. One is a flag with an adjustment or an explicit note. None is a keep.

The evidence hierarchy

The note’s credibility is the credibility of its source. Rank sources this way and cite what you actually used:

1. Statutory / audited filings — segment notes, annual reports, filings that disclose revenue by line. The strongest; use wherever available.
2. Database company profiles — the benchmarking database’s own classification and notes. Fine for the first pass; not a substitute for verification on the companies that matter.
3. Website, LEI/GLEIF entity data — what the company itself says it does; corporate linkage and legal-name history. Best for product line and operations checks.
4. Analyst / reviewer notes — your own observations, with the date and the source examined. Required whenever 1–3 are silent.

A note that says “per the 2024 annual report, 84% of revenue is from X services” is evidence. A note that says “company is comparable to the tested party” is a conclusion with no legs.

Accept, reject, flag

Disposition	Meaning	Reason standard
Accept	Comparable; enters the pool	The five checks, summarised per factor with the source
Reject	Not comparable; excluded from the pool	The specific failed check(s), with numbers: “62% of revenue outside the service line; entity-level PLI reflects group head-office activity”
Flag	Comparable with a known difference that will be addressed	The difference and how it is handled (adjusted in stage 5, or noted as immaterial with the basis)

The reason standard is the whole game. “Diversified conglomerate — 62% of revenue outside services” is defensible. “Not comparable” is the single most common finding in a transfer pricing examination of an Accept-Reject matrix, because it gives the reviewer nothing to argue with and nothing to accept.

The reasonable-efforts standard

OECD practice (and Indian TPO practice) does not demand proof of non-comparability for every candidate — it demands that you sought it with what was reasonably available. The documentation of that effort:

• what sources were checked per company (the hierarchy above);
• what was available and what was not (“no segment disclosure in filings; website and LEI data reviewed instead”);
• the date of the review — the screen is contemporaneous, not reconstructed.

Absence of data is not a reason to reject; it is a reason to say what you did instead. “Could not verify” followed by a keep-with-flag is a legitimate, documented position. “Could not verify” with no follow-up is not.

Common failures

1. Notes written after the range was known — the matrix was completed after seeing that a particular company helped the range, and the reasons were shaped to match. The date and the order of the working papers tell this story; contemporaneity is the defence.
2. One note, twelve companies — the same sentence pasted for the whole pool. A reviewer who finds three identical reasons for three different businesses has found the study’s weakest page.
3. Rejecting on a ratio, keeping on a product — quantitative and qualitative reasons mixed together, so neither standard is met.
4. Flags that never resolve — a flagged company that enters the pool with no adjustment and no closing note. Every flag needs a destination: adjusted, or closed as immaterial, with the basis.

FAQ

How much of the revenue mix must be in-scope? There is no statutory percentage; practice clusters around 80% as the line where an entity-level PLI stops reliably reflecting the comparable function. Below that, the company is a conglomerate, not a comparable — unless the in-scope line can be isolated, which is an adjustment, not a keep.

Is AI-assisted screening acceptable? Where the system’s accept/reject recommendation is backed by the underlying evidence and a human reviews and signs the rationale per company, it is the same standard as manual screening: evidence per company, reason per company, review contemporaneous. What it must not do is replace the review — an unreviewed machine disposition is not a defensible disposition.

Do I document companies the search never found? The search criteria and the population they returned belong in the file (the quantitative stage’s log). Companies outside the search are outside the study; the criteria are the boundary, and they must be stated so a reviewer can see where it is.