Playbook
    AI & Data
    Craft

    Part of The company brain: where your organization's knowledge lives

    Start with one number

    Two departments, two figures for the same thing. How to turn that one contested number into an agreement a team dares to steer on, in a single afternoon.

    Max van GenderenFounder of Datahub, data and AI architecture4 min read
    Share on
    Start with one number

    There is a moment in almost every monthly meeting when two people quote a different figure for the same thing. Logistics shows delivery reliability at 94 percent. Someone from sales shakes their head; going by what customers report back, it looks closer to 85. Both are looking at real data. They are simply measuring something different, without anyone ever having agreed on it.

    Two figures for the samethingLogistics measures 94% deliveryreliabilitySales hears closer to 85% from customersBoth real data, different measurementThe meeting argues about the numberOne agreed definitionWhich date counts: promised or requestedPer order or per order lineA partial delivery counts, or does notThe meeting makes the decisionThe difference is not in the data, but in a measurement nobody ever agreed on.

    The reflex is usually to start big: a data platform, a project, a team that puts everything in order. There is also a smaller first step, and it starts with that one contested number.

    Pick the number people argue about

    Not the most important number, but the one where the friction is: the figure two departments have two versions of. That disagreement is a sign that the meaning underneath was never captured. Which is exactly why it makes a good starting point, the gain is visible immediately.

    Write down what it means

    Sit down with the people who use the number and make the choices out loud. Do you measure on time against the date you promised yourself, or against the date the customer originally asked for? Do you count per order or per order line? Does a partial delivery count as on time? There is no right answer, only the agreement you make and write down.

    Writing it down is the real work, and it takes an afternoon rather than a quarter. What you end up with is a definition in plain language that everyone who uses the number can read.

    Give it an owner

    Appoint one person who is accountable for it, not a committee. That person adjusts the definition when needed and tells everyone who uses the number. Without an owner a definition starts drifting again after a few months, and the next meeting has two figures once more.

    Compute it from the agreed definition

    You do not need a platform for this yet. Take the data you already have and apply the definition in a spreadsheet. The number now comes out of the agreement, and anyone can follow how it was built up: which lines count, which drop out, over which period. From now on you can show, alongside that number, where it came from.

    01Pick the numberWhere the friction is, not the biggest one02Write down what it meansPlain language, an afternoon of work03Give it an ownerOne name, not a committee04Compute it from the definitionA spreadsheet is enough to start05Use itNumber and meaning arrive togetherFive steps, one number: from meaning to owner to source to a decision.

    Use it, and let it grow

    Put the number on the table in the next meeting with its definition attached. Last time''s argument does not happen, because the figure and its meaning arrive together. The team can make a decision instead of bickering about the figures.

    With that you have walked the whole path in miniature: from meaning to owner to source to a number you dare to steer on. That is a company brain in its first form. The next number goes faster, and when the spreadsheet gets tight, the platform slots in underneath. But it started with one afternoon and one number, and that is how an organisation grows without getting bigger.

    About the author

    Max van Genderen

    Founder of Datahub, data and AI architecture

    Max works on data foundations for logistics, retail and manufacturing: the governance, meaning and access layer that analytics and AI agents lean on. He designs the Datahub architecture, leads client implementations, and writes most of the articles and research pages on this site.

    Why this source

    • Designs and implements data foundations at logistics, retail and manufacturing organizations
    • Owns the foundation scan: the first-party measurement behind our research pages
    • Author of the pillars 'The company brain' and 'Managing intelligence'

    Writes about: Data governance · Semantic layer and data modelling · Private AI and AI agents · EU AI Act and data rules

    Reviewed by: DatahubDatahub editorial team

    More about the team

    Next step

    Want to see what's already inside your organization?

    Leave your details. We'll reach out and plan a scan. Within thirty days you'll see one concrete result.

    No newsletter, no reselling. Just this conversation.

    Comments

    Comments are reviewed by the editors before they appear.

    Use your Google or Apple account, or your business email address.

    Sign in to comment