The Competitive Set Inclusion Standard: Direct, Adjacent, Substitute, or Out

You just surfaced a company and someone asked the question that stalls every research meeting: does it count? This is the standard for answering that once, defensibly. It is for strategy and research teams, talent-intelligence analysts, and operators sizing a market who need to take a single company and assign it a tier - direct, adjacent, substitute, or out - from public evidence, with a score a second analyst would reproduce.

Every ranking page defines direct, indirect, and substitute competitors, then hands you a feature-comparison table. None answers the prior question: does this company belong in the set at all, and if so where. That is the gap this document fills. I score competitive relevance on four public axes you can actually verify, so a team stops arguing about who counts and grades each candidate the same way.

Why "who counts" is contested by design

The judgement is contested because the competence to make it is spread across two job families, not one. In Refolk's index of professional profiles, 247 people in the United States hold Competitive Intelligence Analyst or Manager titles, while 178 US product marketing managers list competitive intelligence as a skill without carrying the title. The people who decide who counts sit in product marketing as often as in a dedicated CI seat.

That diffusion is the whole problem. When one person in a dedicated analyst role and another in product marketing look at the same company, they reach different tiers unless a shared rubric forces the same evidence and the same scale. The rubric is not bureaucracy. It is the only thing that makes two people agree.

The scarcity gets worse outside the US. Refolk's index holds only 24 dedicated CI-title professionals in the United Kingdom against 247 in the US, roughly a tenfold gap. Thin analyst pools mean fewer people can carry tacit knowledge about who competes with whom, which makes a written, reproducible standard more valuable in a small market, not less.

The tier definitions and where sources disagree

A direct competitor sells a substitutable product to the same buyer persona, for the same job to be done, at a roughly comparable price. An adjacent competitor targets the same category with a product different enough to act as a partial substitute. A substitute (also called a replacement or phantom competitor) solves the problem from a different category entirely, spending budget your customers could have spent with you. Out means none of these hold to a scoreable degree.

Sources agree on the direct end and diverge sharply in the middle. The table below shows three cited definitions side by side, and the divergence is the point.

Notice that one source classifies by category, one by buyer, and one by budget. That is why teams argue: they are silently using different tests. I resolve it by taking a position. Budget beats category as the discriminant, because buyers allocate money, not taxonomies. The category model produces false positives - two tools that share a G2 category but never meet in a deal. The budget test dissolves them: if a prospect chose between you and them with the same budget, they are direct; if they solved the problem differently and you were never in the running, they are indirect.

Ignore the count folklore while you are here. The widely repeated figure - businesses average 25 competitors including 10 direct, 10 indirect, and 5 replacement - appears verbatim across sites but no source cites a primary study. Treat it as folklore. The one survey that exists shows only 16.92% of teams track more than 10 direct competitors, and practitioner guidance is three to five direct competitors, since more than five usually means you have let indirect competitors leak into the direct tier.

The four axes and what each signal proves

Score four public axes: product overlap, buyer/ICP overlap, geography, and talent/hiring overlap. Each has a verifiable public signal and a specific way it lies. Knowing the lie is what separates a signal from a guess.

Product and category overlap. The signal is co-listing and comparison-page data on review platforms. A "Compared To" field shows the competitors your product was directly compared to on a comparison page. What it proves: buyers put you in the same consideration set. What it fails to prove: that you lose deals to them. Co-listing is adjacency, not loss.

Buyer and ICP overlap. The signal is mid-funnel behavior on review platforms - prospects comparing vendors, reading reviews, or exploring competitor pages. What it proves: shared research interest. What it fails to prove: shared buyer economics. The same problem can have budget spread across departments, so an indirect competitor may be funded by a budget line your sales team never sees.

Geographic overlap. The signal is the buyer's company headquarters location from intent data. What it proves: you are chasing the same regional accounts. What it fails to prove: that regional presence translates to contested deals, since two vendors can dominate different segments in the same geography.

Talent and hiring overlap. The signal is job postings and technographic (tech-stack) changes crawled from public indicators. What it proves: a firm is building a similar capability. What it fails to prove: that they compete for your customers. A company hiring the same engineers may serve a different ICP entirely.

Here is the discipline that ties the axes together: never let a single axis carry a tier. Talent overlap in particular is the axis most often overstated. Pair every hiring signal with a buyer or ICP signal before it earns any weight toward "direct."

The talent overlap benchmark you score against

To score talent overlap you need a benchmark for how much CI and competitive hiring a company can plausibly carry, and that benchmark is thin. The table below, drawn from Refolk's index, shows where CI competence actually lives.

Two derived facts follow. The US-to-UK ratio for dedicated CI titles is roughly 10.3x, and PMMs carrying the CI skill are about 72% of the dedicated-title headcount in the US. The practical read: a company hiring a "competitive intelligence" role is a genuine capability signal, but a company embedding CI as a skill inside product marketing is far more common and easier to miss. When you score talent overlap, look for the CI skill inside PMM job posts, not only for the rare dedicated title.

Finding those people across a candidate's hiring footprint is exactly the kind of query that used to mean stitching together job boards and profiles by hand. Ask Refolk in plain English and it searches public LinkedIn, the GitHub graph, and the open web at once.

How to score: weight, rate, and combine

Use the Competitive Profile Matrix mechanics with one deliberate change. In the standard CPM you assign each critical success factor a weight, rate each competitor 1 to 4 per factor, multiply weight by rating, and sum for a total weighted score. A published worked example totals to 3.15 on that scale. That additive total is fine for ranking companies you have already decided belong in the set.

For the prior question - does this company belong in the direct set at all - additive scoring is the wrong tool, because it rewards a company that overlaps strongly on one axis and not at all on the others. A talent-only match can climb into the direct tier on the strength of one axis. So I borrow the multiplicative logic from signal scoring, where Fit, Intent, and Timing are multiplied so the composite drops to zero when any factor is zero.

The rule: use the multiplicative composite to decide the tier, and report the additive weighted total alongside it to rank within a tier.

Set weights that trace to buying reality. Weights chosen because the data is easy to find are the fastest way to a matrix that misleads. Anchor them in how often each factor decides a real deal - product and buyer overlap should carry the most weight, geography and talent the least. Two analysts must sign off on the weight table before any scoring starts, because a disputed weight table poisons every score built on it.

AXIS                 WEIGHT   RATING(1-4)   SOURCE URL / EVIDENCE
Product overlap      0.35     __            __
Buyer/ICP overlap    0.35     __            __
Geography            0.15     __            __
Talent/hiring        0.15     __            __

ADDITIVE TOTAL   = sum(weight x rating)          -> ranks within a tier
MULTIPLICATIVE   = product(rating/4)^weight       -> decides the tier

TIER BANDS (multiplicative, 0 to 1):
  Direct      >= 0.70    (weekly monitoring, battlecard)
  Adjacent    0.45-0.69  (weekly scan, no staffing)
  Substitute  0.25-0.44  (monthly, watch budget line)
  Out         < 0.25     (drop unless a trigger fires)

The procedure

This is the seven-step run for grading one candidate. Budget roughly half a day per candidate for evidence gathering; the rest is fast once the weight table is fixed.

The reproduction check in step six is the load-bearing step and the one teams skip. If two analysts working the same evidence land more than one band apart, the failure is not the analysts - it is an ambiguous rubric or thin evidence. Fix that before you trust any score in the set.

Two analysts landing more than one band apart is a rubric failure, not an analyst failure.