Crux
Book a call

Data Foundation

Your metrics and KPIs defined in a single layer, and optimised for AI

We join every platform in one warehouse, model it around how your business works, and check it every night with hundreds of tests.

Book a call

Ownership Yours from day one. The models are yours from the day we write them. If you leave, the whole instance goes with you at no charge.

Your Data Foundation● Validated every night
Shopify Meta Ads Google Ads GA4 Klaviyo Amazon YOUR WAREHOUSEModelledaround your businessValidatedhundreds of testsDefined onceevery metric Dashboardswhere to look AI analystin Claude
01

Included

What's included

  1. 1

    All of your data sources ingested

    Shopify, Meta, Google Ads, GA4, Klaviyo and Amazon (plus more), joined in one warehouse.

    Data sources
  2. 2

    Modelled around your business

    Your COGS, attribution rules (opens in a new tab), contribution margin definitions and targets, set in a sheet your team can change without code, so the model fits how your business runs.

    Data modelling
  3. 3

    Validated every night

    Hundreds of tests catch missing orders, revenue mismatches, date gaps and changed fields before they reach a dashboard, and the warehouse reconciles to each platform.

    Data validation
  4. 4

    Every metric defined once

    CAC, new customer conversion rate, CM1, CM2, CM3 and the rest, defined and described in one place for your business, so every dashboard and every AI answer uses the same definition.

    Data definitions
  5. 5

    Dashboards that say where to look

    The full dashboard suite: KPI health reports, acquisition deep dives, retention and LTV, and product and order deep dives. All built from operator experience.

    Dashboards
  6. 6

    An AI analyst in Claude

    Your own custom MCP connector, available to the whole team. Ask the next question in plain English and get the answer in seconds, with the data sources behind it named.

    AI analyst
02

Your numbers

Your numbers, set up around how your business runs

Your configuration sheet sets the rules the model follows, and your team can change it without code. Every number that uses a change updates at the next refresh.

  • Track unique discount codes.
  • Track COGS, fulfilment, contribution and margin costs per order or item.
  • Assign your own attribution rules.
Fig. 1 · Your config sheet, your numbers
Crux config · Google Sheets
COGS
SKUProductUnit cost
KIT-01Starter kit£12.40£13.10
REF-01Refill pack£12.40
Fulfilment
Pick and packper order£3.80£4.20
Packagingper item£0.85
Discount codes
CodeChannelPartner
SPRING10EmailNewsletter
GEMMA15InfluencerGemma

Illustrative: you change a cost or add a code, and every number that uses it changes at the next refresh.

  1. Attribution (opens in a new tab) built from six platforms and modelled on your survey answers, because the platforms shouldn't mark their own homework.

    Attribution

  2. Your COGS, fees and contribution margin are modelled in, so growth is measured in profit, not just revenue.

    Contribution margin

  3. Every KPI is flagged urgent, caution or opportunity against its own last 12 weeks, with daily pacing against budget.

    KPI health
03

Validated

Checked every night, before anyone looks

Hundreds of tests catch missing orders, revenue mismatches, date gaps and changed fields before they reach a dashboard, and the warehouse reconciles to each platform.

Fig. 2 · Last night's run
Nightly validation · 02:14 412 of 412 passed
  • Orders reconcile to Shopify 1,284 of 1,284
  • Ad spend matches Meta and Google to the penny
  • No missing days 365 of 365
  • No duplicate orders 0 found
  • Refunds match Shopify £2,104
  • Every metric defined and described 48 of 48

Illustrative: example questions and figures.

04

Built for AI

Your AI is only as good as the data behind it

Your dashboards and your AI analyst read the same validated data and the same definition of every metric, so they never give you two different answers.

Fig. 3 · Your own custom attribution model
Orders, one month · 5,000 Where they came from
LAST CLICK Meta 900TikTok 300Influencer 200Google 750Email 450Brand search 700Direct 1,100Unknown 600 2,400 unplaced · 48% AFTER THE WATERFALL 1,2505501,150900650250 500 unplaced · 10% ① Attribution waterfall six sources, checked in order AFTER RE-ATTRIBUTION Meta 1,500TikTok 650Influencer 1,300Google 900Email 650 0 unplaced · 0% ② Modelled re-attribution split by your survey answers

Illustrative: one month of 5,000 orders. The waterfall checks six sources in order, then re-attribution splits what is left by what surveyed customers said.

  1. We remodel attribution from your post-purchase survey, so orders left as Direct or Unknown are credited to what actually drove them. We don't let the ad platforms mark their own homework: each one reports the sales it claims, so the same order is often counted by two or three of them.

    Attribution

  2. Built for AI from the ground up: every table and metric is defined and described, so your AI workflows get the right answer first time, on fewer tokens.

    Data definitions

  3. Questions that used to wait days for an analyst are answered in seconds, from the same numbers your dashboards show.

    AI analyst

  4. Every number and definition is traceable to its source.

    Traceable
05

See it working

Browse the workflow library: the questions your Claude analyst answers, grouped by the job they do. Read how we trace each order back to the channel that earned it, in the attribution walkthrough. (opens in a new tab)
06

Next

Start with your reporting

We go through how you report today before we price anything, then scope the build against what we find.

Book a call