Data Readiness Is Four Questions, Not a Data Strategy
Mid-market AI projects fail on join keys, not on model quality. The fix is one identifier per entity, not a warehouse.
Bence FedorIT Solution ArchitectThe four questions
Does the fact exist in a system at all, or does it live in somebody's head and inbox? If a critical field was never captured, no tooling recovers the history — you can only start capturing it today.
Is there a stable unique identifier connecting your systems — a company ID, property ID, booking reference — or are you matching on names? Name matching across legal entity, trading name and local-language spellings fails constantly, and fails silently. Is the history long and consistent enough to compare periods, given that a migration two years ago usually means your three years of data is really fourteen months? And who grants read access, and how long does that take? That is frequently the longest single line in the plan.
The fix is three days of unglamorous work
Designate one golden system per entity type — customers in the CRM, properties in the operational system, invoices in accounting — and give every other system a foreign key pointing at it.
Adding an identifier field and backfilling a few thousand records takes two to three days and unblocks everything downstream.
What not to do about it
Responding to this diagnosis with a data warehouse or a business-intelligence programme costs six months and meaningful money, produces no operational change, and leaves the join-key problem intact with a dashboard on top.
Pick one customer that appears in your CRM, your accounting system and your operational system. Check whether any single field lets a machine match those three records automatically.
Designs the API and integration layer that automation depends on, across systems never built to connect.
Meet the team →Questions we get asked
Do we need a data warehouse before doing AI?
Almost certainly not. What you need is one agreed identifier per entity and the discipline to enforce it at creation. That is three days of work; a warehouse is six months and leaves the underlying problem intact.
Why do AI projects fail on data in mid-market companies?
Join keys. Systems hold the same company under a legal name, a trading name and a local-language spelling, with no shared identifier, so matching fails silently and every downstream output inherits the error.
How much history do we need for useful analysis?
Enough consistent history to compare like periods. Check for system migrations first — a change two years ago usually means your apparent three years of data is really fourteen months.
Related insight
The AI Assistant That Does Everything Does Nothing
An agent's reliability collapses as you add systems it must be simultaneously correct about. The only agents that reach production…
AutomationYour Software Is Fine. The Gaps Between It Cost You Forty Hours a Week.
The recoverable waste in most operating companies sits in the transfers between systems, not inside any one of them. No vendor is …
AutomationAutomate a Broken Process and You Buy a Faster Broken Process
Automation makes a process permanent. Any step you encode without questioning becomes structurally harder to remove than it was th…
Where is your business still running manually?
We review your processes, identify the highest-impact automation opportunities and estimate their financial impact — before anything is built.