Competitive Intelligence Metrics

Competitive intelligence is the work of turning public information about competitors, buyers and the market into decisions somebody makes differently as a result. The weight sits on that last clause. Collecting is the cheap half and it is the half most programmes are measured on, which is the root of nearly every problem on this page.

In practice it covers four things: watching what competitors do, working out what those moves mean, getting that reading to the person who needs it while they can still act on it, and keeping the whole record current enough to be trusted.

What separates it from market research, and from espionage

Market research asks what a market wants. Competitive intelligence asks what named competitors are doing about it, and what you should do in response. The two overlap in method and differ in the unit of analysis: one studies a population, the other studies a handful of companies you meet in deals.

The boundary at the other end is legal rather than academic, and it is not blurry. Everything worth building a programme on comes from material that companies have published, filed, advertised or sold, plus the evidence your own deals generate. Misrepresenting who you are to obtain a quote, handling information covered by an agreement you are bound by, and exchanging pricing intentions with a competitor are each prohibited, and none of them is necessary.

Why that definition is what makes it hard to measure

Functions that are straightforward to measure own their own output. Marketing produces leads and counts them. Sales produces bookings and counts those. The output of competitive intelligence is a decision taken by somebody else, often weeks later, usually with several other inputs in the room.

That single structural fact explains most of what follows. It is why attribution is the hard half of any return-on-investment calculation here, why the operational metrics are more trustworthy than the outcome ones, and why almost every dispute about a competitive intelligence number turns out to be a dispute about its denominator rather than its arithmetic.

Search for competitive intelligence metrics and you will find lists of twenty or thirty, each presented with a percentage attached. Market share gains of half a percent to two percent a year. Sales cycles fifteen to twenty-five percent faster. Win rates eight to eighteen percent higher in competitive situations. The figures are specific, they are confidently stated, and almost none of them carry a citation you can follow to a method.

A benchmark you cannot trace is worse than none

That matters more than it looks. A benchmark is not a decoration on a metric, it is the thing that tells you whether your own number is good. Quoting a range whose definition you cannot inspect means you are comparing your carefully defined figure against somebody else’s undefined one, and the comparison will be wrong in a direction you cannot predict. The honest version of most of these rows is that no benchmark exists.

The denominator is where the metric actually fails

The second failure is more common and more fixable. Almost every metric in this field is a fraction, and almost every argument about one turns out to be an argument about its denominator rather than its arithmetic. Win rate is a good example: the division is trivial, but whether a deal counts as competitive is a decision somebody has to make and write down. Count only deals where a competitor was logged in the CRM and you will measure your logging discipline. Count every deal where a competitor was plausibly present and you will measure your optimism.

Market share fails the same way and more expensively, because the denominator is a market definition rather than a filter. Share of voice fails on the channel measured. Coverage fails on the list of what should have been covered. In each case the number moves for reasons nobody can reconstruct three months later, and a metric that moves inexplicably loses its audience faster than one that moves in the wrong direction.

A third group is genuinely soft, and that is fine

There is a third category worth naming separately, because it is the one most likely to be quietly abandoned. Some of these metrics are genuinely soft. Stakeholder confidence, decision confidence and rep confidence appear on most published lists, and none of them has a formula. That does not make them worthless. It makes them survey instruments, which means they need a fixed question, a fixed scale and a fixed population, and they need those before the first measurement rather than after somebody asks why the number moved.

The table above marks all three cases explicitly. Where a figure exists and can be sourced, it is named and dated. Where the threshold comes from the formula itself, it says so. Where the only sound comparison is your own previous quarter, it says that. And where nobody has published anything, it says that too, which is the entry most of these metrics deserve.

The most common mistake in measuring a competitive intelligence programme is reaching straight for the outcome. Influenced revenue and return on investment are the numbers a budget conversation wants, so they are the numbers people try to produce first. They are also the two that take longest to become trustworthy and are the easiest to argue with, which is how a programme ends up defending its measurement instead of using it.

The alternative is to measure the chain that produces the outcome, one link at a time. Each link is countable on its own, each one fails in a visible way, and a break anywhere explains the number at the end.

1. Is the evidence current?

Intelligence freshness: the share of tracked items verified inside the window you committed to. This is the earliest indicator in the chain and the one most worth watching, because everything downstream inherits it. A battlecard that is four months stale does not announce itself. It sits in the enablement library looking exactly like a current one, gets used in a live call, and gets a rep corrected by the buyer.

2. Did it reach the deal?

Competitive deal coverage: the share of competitive deals that received relevant intelligence before the decision, not after it. The word before is doing the work. A briefing delivered after a loss is a post-mortem, which is valuable for the next deal and worth nothing for this one. Coverage is also the metric that exposes an uncomfortable truth early, which is that most programmes are covering the deals they hear about rather than the deals that matter.

3. Did anyone use it?

Adoption: reps who opened a card in the period, over reps who had a competitive deal in the period. The denominator is the part people get wrong. Measured against the whole sales organisation, adoption rewards you for headcount. Measured against the reps who actually faced a competitor, it tells you whether the work is reaching the room. Neither version proves the card was used in the conversation, which is why this link in the chain needs a qualitative check alongside it.

4. Did the deal behave differently?

This is where the sales metrics belong: win rate against each named competitor, the gap between contested and uncontested cycle length, the discount gap, and average deal value where a competitor was present. These are lagging indicators and they are noisy at small volumes, which is an argument for reading them quarterly rather than for skipping them.

5. Did it show up in revenue?

Competitive-influenced revenue and then, only if you can defend the attribution, return on investment. Both belong at the end of the chain rather than the start, and both should be reported with the earlier links visible next to them. A programme that can show coverage rising, then adoption rising, then win rate against one competitor improving, has told a story that survives scrutiny. A programme that presents an ROI multiple on its own has told a story that invites it.

Metrics and KPIs get used interchangeably, and separating them changes how many of them you should be carrying. It is a different question from sales performance metrics, which measure the team rather than the intelligence reaching it. A metric is anything you can count. A competitive intelligence KPI is the much smaller set you have agreed to be judged on, each with a target, an owner and a review cadence attached. The test is not whether a number is interesting. It is whether anyone has committed to it.

Three or four KPIs, not twenty

In practice that means a programme can reasonably track a dozen or more metrics and should hold itself to three or four KPIs. A list of twenty KPIs is not a rigorous programme, it is a list of metrics that nobody has been willing to prioritise, and it produces the specific failure where every number is reported and none is acted on.

Which three or four you commit to depends on your programme's maturity. A new programme should take the two operational ones, coverage and freshness, because both are countable inside a month and neither requires an attribution argument. An established programme can carry an outcome KPI, usually win rate against a named competitor, because it has enough quarters of consistent tagging for the number to mean something. Committing to influenced revenue in month one produces a figure that gets revised twice and then quietly dropped.

Every KPI needs a defined direction of travel

The other property a KPI needs is a defined direction of travel. Some of these metrics are not better when higher. A discount rate falling in contested deals might mean your positioning improved, or it might mean you stopped competing for the deals you were losing on price. Cycle length shortening might mean better enablement, or it might mean a competitor withdrew from your segment. Any KPI that can move for two opposite reasons needs a second number reported beside it, or it will be read as good news once and as a mystery afterwards.

These two are worth separating carefully, because published definitions disagree and several of them state outright that the terms are interchangeable. They are not, and the difference is large enough to make a benchmark meaningless if you apply the wrong one.

The two calculations, on the same hundred deals

Take a hundred closed competitive deals, sixty of them won. The rate puts wins over every closed deal and reports 60%. The ratio puts wins over losses alone and reports 1.5. Identical deals, two denominators, two numbers that cannot be read as each other: parity sits at 1.0 on one scale and at 50% on the other, so a figure of 60 means something completely different depending on which was meant.

Why an unlabelled figure is unusable

The practical consequence shows up whenever you compare yourself to anything. A figure quoted as a win rate of 1.5 is a ratio that has been mislabelled, and a figure quoted as a win/loss ratio of 60% is a rate that has been mislabelled. Either one, taken at face value, will tell you that you are performing very differently from how you actually are. Before comparing your number to a published one, check which calculation produced it, and if the source does not say, treat the comparison as unavailable.

Use the rate to manage, the ratio to argue

There is a reason to keep both. The rate is the better internal metric because it behaves predictably and is bounded between zero and one hundred. The ratio is better in a room, because 1.5 against one competitor and 0.7 against another makes the asymmetry obvious in a way that 60% and 41% does not. Use the rate to manage and the ratio to communicate, and label both every time.

Both calculations share the same exclusion. Open deals do not belong in either denominator. Including them lets a growing pipeline improve the number without a single additional win, which is the most common way a competitive win rate gets quietly inflated.

Most of the benchmark gaps in the table above cannot be filled, because the data does not exist. One of them can, and it happens to be the most useful one, because it is the earliest link in the chain.

We took the four platforms this category is usually shortlisted from, pulled 500 verified reviews across them, coded every one against a written evidence rule, and published the method next to the result. Two of the findings bear directly on measurement.

Stale content is not a marginal complaint

The first is that stale content is not a marginal complaint. On one of the four platforms it was the sharpest single criticism, appearing in 13.0% of that product’s reviews. These are customers of tools bought specifically to keep competitive information current, describing the information as out of date. That is the strongest available argument for treating freshness as a tracked metric rather than an assumption, and for measuring it on your own content regardless of what you bought.

Alert volume is the failure mode, not the achievement

The second is that the only complaint shared by all four competitive intelligence platforms at material frequency was alert noise, at 3.3% of reviews for one, 10.0% for another, 5.0% and 6.0% for the remaining two. Alert volume is the thing most programmes instinctively measure, because it is the easiest number to produce and it always goes up. The review data says that volume is the failure mode rather than the achievement. A metric that rewards detection without rewarding triage produces exactly the outcome those reviewers are describing.

A formula is only as good as the two figures you put into it, and this is where most measurement projects stall. Roughly half the metrics here draw on internal systems you already own, and the other half need a number about a competitor that nobody is going to hand you.

The internal half is mostly one system

The internal half is mostly one system. Win rate, cycle length, discount rate, average deal size, coverage and influenced revenue all come out of a CRM, and all of them depend on the same prerequisite: a field recording which competitor was in the deal, populated often enough to be worth analysing.

That field is empty by default in every major CRM, and the practical work of measuring competitive performance is mostly the work of getting it filled. Below roughly half coverage, nothing built on it survives a challenge. Getting competitive intelligence out of a CRM works through which figures survive that constraint, and how to structure the field so it yields counts rather than free text somebody has to read.

The external half needs a public source

The external half is harder and more interesting. A competitive pricing index needs a competitor’s prices at a matched configuration. A displacement rate needs to know which incumbent an account is leaving. Relative market share needs a figure for the largest competitor. None of these arrive by asking, and all of them are available in public sources if you know which one holds what and how far it can be trusted.

Survey inputs need the same discipline as numeric ones

The qualitative inputs deserve the same rigour as the numeric ones. Competitive confidence is a survey, and a survey with a drifting question produces a trend line that measures the question rather than the confidence. Write the wording down, keep the scale fixed, and hold the population constant even when it is inconvenient. Win/loss interviews are the other half of this picture, and the only input that explains why a number moved rather than confirming that it did.

The published lists are long because length is cheap and because a longer list looks more rigorous. Length is the opposite of rigour here. Every metric you track is a commitment to define it, collect it, defend it and explain it every time it moves, and that cost is paid every reporting period rather than once. Three filters remove most of a long list quickly.

Can you define the denominator today, in one sentence?

If not, the metric is not ready. This removes market share from most early-stage programmes, not because share does not matter but because defining the market is a project in itself and the metric is unusable until it is done.

Would a plausible movement change a decision?

Pick a number, imagine it moving twenty percent in either direction, and name the decision that changes. If nothing changes, the metric is describing the programme rather than steering it. Alert volume usually fails this test. Coverage usually passes it, because a fall in coverage has an obvious response.

Can you still produce it in six months?

Metrics that depend on somebody remembering to do something manually decay on a predictable schedule. A metric produced by a saved report survives a busy quarter. A metric produced by an afternoon of spreadsheet work does not, and its absence from the next review will be read as a result rather than as a gap.

What survives is usually the same short list: coverage and freshness on the operational side, win rate by named competitor and the contested cycle gap on the outcome side, and one adoption number showing the work is reaching the people it was written for. Five numbers, each with a visible denominator, is a programme people can argue with productively. Thirty is a document.

A board conversation is not a smaller version of the internal dashboard. It has less time, less context and a different question, which is whether this function should keep its budget rather than whether it is running well.

The three things that belong in the slide

Three things do that job. The first is one outcome number with its denominator and its period stated plainly, usually win rate against the competitor that appears in the most pipeline. The second is the direction of travel over four quarters rather than one, because a single quarter in a competitive sample is noise and presenting it as a trend is the fastest way to lose credibility when it reverses. The third is the decision the number is asking for.

State the attribution rule rather than burying it

Report the attribution rule alongside the figure rather than in an appendix. A stated rule that a reader can discount is far more persuasive than a clean number whose derivation is hidden, and the question will be asked either way. The same applies to the metrics you decided not to track: saying that you do not report competitive intelligence ROI because the attribution cannot yet be defended is a stronger position than presenting a multiple you cannot support.

Bring last period’s actions with you too, reported honestly including the ones that did not work. A competitive programme that only ever reports improving numbers is a programme whose numbers have stopped being believed.

Every metric on this page has the same failure mode, and it is not the arithmetic. It is that the inputs go quietly out of date. Freshness is the clearest case, because it measures decay directly, but the problem reaches further than that: a pricing index built on prices that changed last month is wrong in a way that looks exactly like being right, and a battlecard adoption figure rising against stale cards is worse than no figure at all.

Manual collection loses this race by design. The changes that break a competitive metric arrive silently, on somebody else’s schedule, and a quarterly refresh means the number spends most of its life describing a market that has moved. Flares watches competitor pricing, product, positioning and hiring continuously and dates every change, so the inputs behind these formulas stay current without anyone remembering to check. The measurement is still yours to define. What stops decaying is the evidence underneath it.