The method

Four layers under every product.

Whether you buy a single dashboard or the Full Programme, the work rests on the same four layers. Here is what each one fixes and what you get from it.

Layer I · Identity resolution

The Fidelity Layer

The problem
Guest checkouts, old email addresses and typos turn one customer into three.
What we do
We match the records that belong together, using probabilistic matching (Python’s Splink) and common sense, and give each real person one Master Customer ID.
What you get
A truer count of your customers, and one history for each of them.
Layer II · 0–9 RFM scoring

The Decile Engine

The problem
Averages hide who actually drives the business, and who is quietly slipping away.
What we do
We score every customer 0 to 9 on how recently they bought, how often, and how much they spend.
What you get
You see your “999” Brand Champions, and the big spenders who have gone quiet, so effort goes where it counts.
“999” Brand Champion Big spender going quietIllustration
Layer III · Inventory and demand

The Velocity Layer

The problem
Cash sits on shelves in stock that isn’t moving, while best-sellers run out.
What we do
We age your stock to find the dead lines, and forecast demand for each product from your own sales history.
What you get
Clear reorder numbers, and less money tied up in stock you can’t sell.
Layer IV · Marketing attribution

The Illumination Layer

The problem
Ad and email spend goes out, but nobody can say which part brought the sales in.
What we do
We tie each campaign, channel and email to the orders and revenue that followed it.
What you get
You can see which spend earns its keep, and which to cut.

One product in the roster

The Full Programme

For founders who want the whole build. It runs in four stages, from the diagnostic to AI advisory.

  1. Stage 0

    Diagnose

    The 7-day diagnostic: where revenue is leaking and where the data can’t be trusted, with a written list of fixes.

  2. Stage 1

    Foundation

    A BigQuery warehouse with every source in one place, and a Looker Studio command centre on top.

  3. Stage 2

    Predictive growth

    Churn prediction models that flag customers who are drifting, and automated cross-sell.

  4. Stage 3

    AI advisory and pricing

    Generative AI advisory systems that work from your own data, and dynamic pricing.

The tools

What we build with

  • Storage
    Google BigQuery
    Serverless data warehouse
  • Pipelines
    Airbyte / Fivetran
    Automated data loading
  • Modelling
    dbt
    Tested, modular SQL
  • Science
    Python
    Splink, Prophet, XGBoost
  • Dashboards
    Looker Studio
    Your command centre
See all productsBook a 7-day diagnostic