A requirement list describes what the plant needs to do. It does not describe screens, and it does not describe the old system with new words. Write it from the processes, along the order flow from inquiry to invoice.
One need per line
| Weak | Better |
|---|---|
| Flexible planning | The planning run proposes orders per material from demand, stock, lead time, lot size rule and safety stock, and shows the pegging to the customer order. |
| Modern warehouse functions | Stock is managed per storage location, and a count can be booked per location without blocking the whole warehouse. |
| Good reporting | Planned and actual load per work center and week can be exported without programming. |
The gear unit plant collected 180 requirements in workshops with the departments, numbered, each with a category, a priority and the process it came from.
| Category | Requirements |
|---|---|
| Planning and scheduling | 34 |
| Production and shop floor feedback | 30 |
| Warehouse and inventory | 26 |
| Purchasing | 22 |
| Sales and order processing | 20 |
| Finance and costing | 18 |
| Master data and change | 16 |
| Technology, integration and reporting | 14 |
Ninety-six are must, 62 should and 22 nice to have. Must means: without it, the process does not run or a legal duty is not met. Every special requirement is questioned once, because often the process is not special, only the habit.
From long list to decision
Ten knockout requirements from the must list cut the market to eight vendors. The written inquiry with the full list, answered per line as standard, configuration, development or no, left five. Functional fit and a first cost frame produced a short list of three.
Then comes the part that decides. Each vendor gets one day, a script and a data set four weeks in advance: 40 materials, 8 bills of material and 8 routings from the cleaned wave 1. Six scenarios, each scored by the key users of the department it belongs to, not by the project lead.
A customer order for a variant gear unit from entry to the delivery date. A planning run with a housing shortage and the rescheduling that follows. A production order with feedback, scrap and rework at the turning cell. Goods receipt in boxes, putaway, and a count of a location with a deviation. A new material from request to release, and a bill changed with an effective date. The cost of a production order, planned against actual.
Five criteria, one score
| Criterion | Weight | Vendor A | Vendor B | Vendor C |
|---|---|---|---|---|
| Functional fit (requirement list) | 35 % | 8 | 7 | 9 |
| Demonstration with own processes | 25 % | 6 | 9 | 5 |
| Total cost over seven years | 20 % | 6 | 7 | 4 |
| Implementation partner | 10 % | 7 | 8 | 6 |
| Technology and integration | 10 % | 8 | 7 | 8 |
| Weighted score (0 to 10) | 7.0 | 7.6 | 6.6 |
Vendor C had the best functional fit on paper, 95 of 96 must requirements in standard, and scored 5 on the demonstration day: three scenarios needed workarounds that the key users refused to live with. In a free presentation nobody would have seen it.
Cost over seven years includes everything, not the licence alone. For vendor B: software EUR 980,000, implementation partner EUR 720,000, internal effort of 9,000 hours at EUR 55 which is EUR 495,000, interfaces and migration EUR 180,000, training and operation EUR 225,000. Total EUR 2.6 million, against EUR 2.9 and EUR 3.6 million for A and C.
Two reference visits per vendor, at plants of similar size, close the picture with the questions nobody answers in a sales meeting: how long did go-live really take, and what did you underestimate.
Contract terms are a matter for your purchasing and legal departments. The system landscape around the ERP system is covered in the Digital Shop Floor series.