Turning Market Maps Into Pipeline

By Cécile George · July 2026

TL;DR

AI has made market maps sharper than ever, yet fewer than one in ten companies has scaled it anywhere (McKinsey, July 2026).

Most segmentation fails commercially for a simple reason: it stops at the map. Nothing reaches the seller's week.

The fix is routing, not more modelling: every priority account gets an owner, a reason to call, and a weekly review. Start with one product line or region.

Done this way it pays quickly: segmenting one client's agent base produced €7m of additional sales in six months and 3x the pipeline from target customers.

Budget for the human part: every dollar spent deploying AI needs roughly three on change management.

The map is not the deliverable

A market map only pays when it changes what a seller does on Monday morning. Most segmentation work never gets that far. It produces an accurate picture, everyone agrees the picture is accurate, and then it sits there. The money is lost at that point, not in the modelling.

McKinsey surveyed nearly 4,000 buyers and sellers across 13 countries in July 2026 and found that fewer than one in ten organisations has scaled AI in any given function. That covers far more ground than segmentation, but the shape of the failure is familiar. Plenty gets built. Very little of it reaches the week.

In segmentation specifically we think the cause is narrower, and more fixable, than most transformation programmes assume. What a seller needs is not a map but a routed opportunity: a named account, someone who owns it, a reason to call this quarter, something to open with.

Where the value leaks

We keep meeting the same four failures. None of them is a modelling problem.

The map never reaches the seller

The analysis lands as a deck, or as a dashboard nobody has a reason to open. Sellers carry on with the accounts they already know, because nothing has arrived in the place where they plan their week.

Nobody decided who calls whom

Knowing which accounts matter is a different question from deciding who owns them, in what order, through which channel. Somebody has to make that call, and until they do the list is only a suggestion. So sellers do the reasonable thing and stick with what they know.

The seller is given a name but not a reason

Hand a good seller an unfamiliar account and they will want to know what changed, what the account probably needs, and how to open the conversation. Without that, the call doesn't get made. A few weeks later the list gets blamed for being wrong.

It goes stale, and trust goes with it

One out-of-date list is enough. A seller opens the map, finds last year's contact at an account they called in March, and stops opening the map. Winning that back costs more than the original exercise did.

What routing looks like in practice

The fix is not a better model but a different endpoint. The same analysis stops when it has produced a map; it should stop when it has produced routed opportunities sitting in somebody's week.

A routed opportunity is a short brief rather than a row in a spreadsheet: why this account fits, what changed recently, what it probably needs, who owns it, what to open with. Short enough to read in a minute. Specific enough to act on before Friday.

An opportunity that has sat in someone's name for three weeks with no contact is either wrong or unowned, and somebody has to say so with that person in the room.

The weekly review is what keeps it honest, and it is less comfortable than it sounds.

There is a reason this step gets skipped, and it is structural rather than intellectual. Large analysis programmes are staffed and priced for the analysis. The people who built the map are usually gone by the time anyone has to decide which fifteen accounts a particular seller will call next quarter, which is roughly the point where the work stops being analytical and starts being awkward. Reassigning accounts. Changing coverage. Sitting in a review every week until the habit holds. None of it is hard to understand. It is slow, and it needs somebody in the room while it happens.

What we see in portfolio companies

Two cases, in businesses with almost nothing in common.

In an industrial casting business selling through 125 or more agents across 50 or more markets, segmenting the agent base produced €7m of additional sales in the first six months, and three times the pipeline from target customers against baseline. The segmentation was the cheap part. What moved the number was rebuilt agent coverage and a weekly review with a name against every target account.

In a premium paper supplier entering Asia, we assessed 50 markets bottom-up. That assessment only mattered because it resolved into three go-to-market archetypes with specific people accountable for each. Two subsidiaries opened; regional sales doubled and stayed there.

Neither result came from a more precise map. Both came from someone owning what happened next, week by week.

Why this matters more on a hold-period clock

For a PE-backed business the question isn't whether this year's segmentation is sharper than last year's. It's how many quarters pass before it moves a number. A map finished in month four that only changes selling behaviour in month eighteen has spent more than a year of the hold on analysis.

Which argues for a smaller first slice than most programmes choose. One product line, or one region, routed properly and reviewed every week, will teach you more and bank more than a complete map nobody acts on.

How to wire it in

Start from the motion, not the model. Decide what a seller should do differently, then build backwards. If you can't name the behaviour, the analysis has nowhere to land.

Prove it on one slice. One product line or one region, routed end to end. Watch what actually happens before you extend coverage.

Give every opportunity a reason. Fit, trigger, likely need, opening. Sellers act on accounts they understand, not on names.

Put it in the weekly rhythm. Win rooms and pipeline reviews anchored on routed opportunities rather than on the map.

Refresh on a clock, and measure conversion. Track how routed opportunities convert against baseline. That number is the only thing that tells you whether any of this worked.

A budgeting note to end on. The working estimate is that every dollar spent deploying AI needs roughly three on change management, and most organisations plan it the other way round. That matches what we see. The modelling is rarely the expensive part. Getting a commercial team to work differently is, and it's the part that's hardest to buy in a fixed-scope engagement.

Sitting on a market map that has not changed anyone's week?

→ See how it worked for 125+ agents