GEODESIC

The industry has not kept pace with the people who buy from it.

The average CMO now lasts 4.1 years, the shortest tenure in the C-suite. They are asked to commit budget faster than a study cycle can answer, so they commit without the answer and live with the result.

That gap is the problem. Not rigour, not craft, not the instruments. The distance between when the question gets asked and when an answer can arrive.

Send us one old dataset Already fielded, already reported. No client involved, nothing to sign.

One space, not twelve reports.

Your tracker, your concept tests, the open ends nobody had time to code, the proprietary files that never leave the building. Read together, against each other, in the same units.

The measured spaceA triangulated sphere with a route crossing it, standing for the single space every source is placed into.

More inputs than any single study could carry, and one output. Every source is placed in the same space, in the same units, so what comes back is one readable structure rather than twelve findings somebody has to reconcile by hand.

The volume is not the problem and never was. The problem is that none of it accumulates. Your instruments stay where they are; what gets added sits on top of them, and the distance between any two things becomes a number rather than an impression.

Out of it come four things nobody could hand you before: where the brand sits, which positions are reachable from there, what the move costs, and which step to take first.

The mathematics is not new. Metric multidimensional scaling has been in the peer-reviewed literature for fifty years, which means you can go and check it.

What comes back, and the room it has to survive

The brief goes to a CMO, a business head, a partner agency. It has to be short enough that they read it, specific enough to act on, and defensible when one of them pushes back in front of the others. That is the standard it is built to.

You get the one-page brief, the data file your team works from, a typing tool for new respondents, and the ledger of what we tried to knock down and what survived. Every finding carries its evidence, including the weak ones.

You brought it into the room. The credit stays with you.

Where it applies, and where it does not

Six layers of marketing measurementSix stacked bands. The second band, structure, is highlighted in gold as the one Geodesic owns; the other five name who does them. ListeningSocial and search platforms Structure Where everything sits relative to everything else Geodesic Trade-offConjoint and choice vendors CausalityAttribution and experiments AllocationMarketing mix modelling ContinuityYour tracker

Not attribution. Not media mix. Not incrementality. If that is the question we will say so on the call, rather than after you have paid for it.

How it runs

The four data levelsFour stacked bars of increasing length, one per level: public information only, plus research already owned, plus sales and behavioural data, plus data collected for the question. Each names what it asks of the client. Level one public only nothing asked of them Level two plus research they own an NDA Level three plus sales and behavioural data a security review Level four plus data collected for the question a contract and a field budget Levels two and three carry most of the value, because they are made of things already paid for.
Level four takes weeks rather than days and is quoted separately. It gets recommended where the question genuinely cannot be answered without it, and ruled out just as plainly when it can. We need raw data rather than topline, because words and numbers go into the same space. White label is normal; your client never has to hear our name.

Four parts. Three of them are machinery.

The four partsThree grey blocks for the mathematics, the written rules and the execution layer, and a gold block for judgment. The mathematics published, checkable The written rules set before the question The execution layer does the arithmetic Judgment checks every step machinery a person Commercial AI is the third box. One component of four, working inside the second, deciding nothing.

What is underneath, and who has checked it

Metric multidimensional scaling in a non-Euclidean space, published for fifty years, built by Dr. Joseph Woelfel across Wisconsin–Madison, Chicago, Illinois at Urbana–Champaign, Michigan State and Michigan, programmed and first commercialised at Albany. In 2009 the U.S. Army funded a RAND review of methods for measuring and changing attitudes at scale. Of the social-science approaches surveyed, this was the one RAND recommended be tested for operational use — the report calls it the closest any approach came to a usable framework for planning and assessing attitude change. Expected utility modelling and automated content analysis were rated already mature; this one was not. That distinction matters and we will not blur it.

It was never a mathematical problem. It needed more computation than anyone would spend on a marketing question. That changed. Nothing else about it did.

Sixteen citations, and what is still open

What you should be asking, that we are not advertising

Measured

Position and distance. True by construction of how the data is collected, not a claim about our skill.

Held out

Concept results, checked against data the model was not shown.

Still open

No published hit rate against realised outcomes. Two-analyst reliability not run. Until it is, we will not call this objective.

What breaks it

Ask the wrong objects and you recover a clean space answering a question nobody cares about. The geometry solves whatever it is handed.

Three ways in

Reanalysis

A set you already fielded and reported. There is a known prior answer, so either this reaches something the original could not, or it does not.

Subcontract

You hold the client. We do the structure layer and hand back the file, the brief and the ledger.

Joint pitch

For a brief you would not otherwise win, or would win on price. Named or not.

Everyone is getting faster. Nobody is getting righter.

Send one old dataset. Tell us where it breaks.

Send us one old dataset