Data Assessments: Know the Problem Before You Migrate It

The gap between knowing and knowing where

Ask a data owner whether their customer master is clean and you will get an honest, useless answer: probably not entirely. Everyone knows the data is imperfect. Almost nobody can say where the problems concentrate, how extensive they are, or which of them carry business risk and which are cosmetic.

That gap is where migration budgets go to die. Programmes scope cleansing effort on assumption, discover the truth during the first mock load, and spend the rest of the timeline negotiating between the go-live date and the data. An assessment closes the gap before the programme commits to either.

What an assessment actually measures

A useful assessment does not produce a generic score. It measures the dimensions that decide whether data will survive its next use: completeness, accuracy, conformity, integrity and uniqueness, each tested against rules the business defined, not defaults a tool shipped with.

Profiling tells you what is there. A business rule tells you whether it should be there. A field can be 100 percent populated and 40 percent wrong, and only a rule that encodes the business meaning will catch it. The same applies to duplicates: exact matching finds the easy ones, and fuzzy and semantic matching find the ones that cost money.

Cross-object validation matters just as much. Customers without valid sales areas, materials without units of measure, vendors without bank details in the roles that pay them: individually valid records that fail as a set. Assessment at object level misses what assessment across objects catches.

From weeks of workshops to days of profiling

When we started advocating assessments, the honest objection was cost. Profiling meant extracts, scripts and workshop after workshop, and by the time findings landed they were stale.

That objection has aged out. Automated profiling reads the source directly. AI-assisted classification names thousands of columns in minutes and flags where its own confidence is low. SAP-aware rule packs test customer, vendor and material objects against the checks that migrations actually fail on. Our own focused assessment runs one domain, Customer, Material or Sales Area, in four to five weeks, and ends with findings, prioritisation and a management summary rather than a maturity model.

Assess before you migrate, not during

The highest-value moment for an assessment is before a migration is scoped, because the findings change the scope. Knowing that 88 percent of customer records lack a valid account group turns a vague cleansing workstream into a sized, owned, scheduled piece of work. Knowing which duplicates cluster where determines the Business Partner merge strategy before the first mock load, not after the third.

Assessment converts assumption into evidence. Evidence converts a hopeful migration plan into a credible one. We have been making this argument publicly for more than ten years, and the programmes that took it seriously are the ones whose cutovers were boring.