Why ERP selection fails before the demo

By the time a shortlist exists, the outcome of an ERP program is largely fixed. The demos, the scoring matrix, the negotiation — all of it operates on assumptions made weeks earlier, usually informally, usually undocumented.

This note describes the pre-selection phase we run before a client speaks to a single vendor, and why it is the cheapest part of the project to get right.

01Start from process, not modules

A module list is a vendor artifact. It tells you what a product ships, not what your organization does. The first deliverable of selection should be a process inventory written in your own language: who initiates a purchase request, who approves it, where it stalls, and what the finance team does at month-end to compensate.

We run this as interviews, not workshops. Workshops produce consensus; interviews produce the truth, because the person doing the workaround will describe it when their manager is not in the room.

If a process cannot be described without naming a spreadsheet, that spreadsheet is a requirement.

What the inventory contains

• The trigger and the closing event for each process
• Every handoff between departments, with the format it travels in
• Volume: transactions per month, peak periods, growth expected
• The known failure modes and how staff currently recover

02Characterize the data before you value the software

Data quality decides the timeline more often than functionality does. Before scoring vendors, we profile the systems of record: row counts, orphaned records, duplicate identity rates, and how many fields are free text where they should be coded.

-- the query that ends most optimistic timelines
SELECT COUNT(*) AS rows,
       COUNT(DISTINCT national_id) AS unique_ids,
       SUM(CASE WHEN national_id IS NULL THEN 1 ELSE 0 END) AS missing_id
FROM   students;

If the unique identity rate is below 95 percent, remediation is a project of its own and belongs in the plan with its own budget line. Discovering that during migration is what turns a six-month rollout into eighteen.

03Score against your weightings, published in advance

A scoring model written after the demos is a rationalization. Publish weightings before the first vendor call, and have the sponsor sign them.

04Decide who owns the system on day one

Name the internal owner before contract signature — a person, with time formally allocated, not a committee. The single strongest predictor of a healthy system two years post-launch is whether someone inside the organization had authority over it from the beginning.

05What good looks like

At the end of pre-selection you should hold four documents: a process inventory, a data profile with a remediation estimate, signed scoring weights, and a named owner with a phase plan. Vendors then respond to your specification rather than shaping it.

That is the whole trick. Selection is not a shopping exercise. It is the last moment you can define the problem before someone else defines it for you.