The first three articles in the series built a mechanism: because the man-day model priced input, it had to sell diagnosis cheaply; once AI made mechanical work cheaper, the quality of the decision became the only real differentiator. One question remains: what does this mechanism look like in a project? This article does not ask a new question — it describes the sequence. How does an ERP journey that says diagnosis first progress, step by step?
The Five Steps of the New Path
- Measure: Before watching demos or collecting proposals, put the current state into numbers — how much of the processes is documented, how clean the master data is, who owns which decision. Without measurement, every remaining step rests on guesswork.
- Diagnose: Turn the measurement into findings. "Inventory accuracy is low" is an observation; "in which warehouse and in which product group are post-count correction entries concentrated" is a diagnosis. The diagnosis tells you which problem is worth solving.
- Write the criterion: Write down the project’s definition of success before it even starts, with a number and a date. "Go-live" is not a success criterion; "the month-end close drops from five working days to two" is a criterion. The criterion must be verifiable independently of the vendor.
- Build the phases by dependency: Let technical dependency, not appetite for modules, determine the order of phases. If a phase does not work without the output of the phase before it, the order is not up for debate.
- Tie acceptance to evidence: At the end of each phase, the record should say not "it works" but "this scenario was run with this data by this person and passed". If acceptance is not tied to evidence, the handover meeting turns into an argument.
The Difference Between the Old Path and the Diagnosis-First Path
The two paths do not do the same things in a different order; they make different decisions at different levels of knowledge. The difference is clear in the table:
| Decision | Old path | Diagnosis-first path |
|---|---|---|
| Scope | At the proposal stage, with the least information | After measurement, based on findings |
| Product selection | First question: which brand? | Last question: which capability, against which criterion? |
| Phase order | According to the vendor’s delivery schedule | According to technical dependency |
| Acceptance criterion | "We went live" | Written scenario + evidence + approving person |
| Weight of the budget | On implementation and mechanical work | On diagnosis and decision quality |
| Owner of the risk | Entirely the customer | Shared, depending on the criterion |
A Capability Category, Not a Product Name
The most visible consequence of the diagnosis-first path is that product selection changes its place in the sequence. The decision is made not with a brand name but with three parts: which capability, against which criterion, with which precondition. "Bill of materials management" is a capability category; "can revision history be kept in a multi-level bill of materials" is that category’s criterion; "product master data must have been de-duplicated" is its precondition.
This triad also generates the questions to ask the vendor — and those questions become the skeleton of the requirements specification. The brand question does not disappear; it is moved to the end of the sequence. Because whether a product is the right choice is a question that can only be answered once the capability and the criterion have been written down. In projects that proceed in the reverse order, the specification is written backwards from the feature list of the product already chosen.
Phase Order Is Built on Dependency
In most projects, the phase plan is a matter of negotiation: which module is switched on first is decided according to the appetite of departments or the vendor’s delivery schedule. On the diagnosis-first path, this order is a technical constraint. A few concrete examples:
- Production execution is not set up before the bill of materials and work order flow are working — execution consumes the output of the bill of materials.
- Costing does not produce meaningful results until stock movements are recorded correctly; cost is a derivative of the movement.
- A sales report is not reliable until customer and product master data have been de-duplicated — duplicate records quietly corrupt the report.
When these dependencies are violated, the project does not stop; something worse happens — a phase built in the wrong order appears to work, and the problem surfaces only once data has accumulated. Tying the phase order to dependency prevents this delayed error from the start.
Acceptance Is Tied to Evidence
The most expensive moment in a project is the handover meeting: the argument between "this was in scope" and "this is additional development" decides which side the invoice lands on. The only way to head off this argument is to tie acceptance to evidence from the start.
In practice, this means: the result of each acceptance scenario is recorded not only as "passed/failed" but, if it failed, classified by its cause — missing configuration, missing data, an out-of-scope request, or a genuine defect? These four causes send the invoice to four different parties. The name of the approving person is written next to it. When no classification is done, all the "failed" entries fall into a single pile, and at the handover meeting who is right becomes a matter not of the record but of who shouts loudest.
Where Do You Start?
The first of the five steps is measurement; and measurement has a starting point too. To see your own project’s risk profile, you can use the 10-question ERP Risk Score assessment: process maturity, management ownership, change management, data quality, internal resources, scope clarity, budget realism, appetite for customisation, schedule and team experience — ten dimensions are scored together with their weights, and you fall into one of four risk bands. The result is an estimate, not a diagnosis: it tells you where you need to look, not what you need to do. But you will have taken the first step of the five-step path.
This article is the fourth in the growing «From Man-Days to Value» series.
1. Is Consulting That Sells Man-Days Coming to an End?
2. Let’s X-Ray the Promises of AI
3. What Has AI Made Cheaper in ERP, and What Has It Made Valuable?
4. Diagnosis First: What Does the New Path in ERP Look Like in Practice? (this article)
5. The Unmeasured Layer of Digital Transformation: Leadership
6. Why Do Reports
Stay on the Shelf in Digital Transformation?
7. (coming soon)
To make sure you do not miss new analyses, subscribe to the MindDX-Digital Excellence newsletter.
