In this guide 6 sections
Test four things on your own data. Show the drawing revision a job ran on, the stock sitting with a subcontractor, a short shipment split into two dispatches, and a complaint traced to its lot. Demos skipping these hide the weak spots.
Ask which revision each job actually ran on
Every demo shows a clean work order. Few show a changed one. Revision B may become C after release, so ask whether the system warns the planner, freezes the old copy or quietly swaps the file under a running job. Make the vendor show it.
Operators print what the system hands them, and a stale PDF at the lathe means a whole tray of scrap by lunchtime. Ask to see the print on the floor terminal, with the revision letter in the header and the approval date beside it.
Then ask who approved it. A revision without a name and a timestamp, tied to every job it touched, is only a rumour that somebody edited a file.
- Can two revisions of one part number be open on different jobs at once?
- Does a revision change reach the inspection plan and the routing?
- Is the superseded drawing still viewable for old lots?
- Who can release a revision, and is that role restricted?
Check how material moves to an outside processor and returns
Plating, heat treatment, powder coating and outside machining all send goods out of the plant, and the stock stays yours even while it sits in someone else’s yard. Ask how the software shows it.
A good answer is a returnable delivery note carrying quantity, process and expected return date, linked to the work order and visible to the buyer. A weak answer is a stray remark. Listen for the difference.
Suppose the processor returns fewer pieces than you sent, with rejections and process loss lumped together and no line of their own to explain the gap. Your stock count then drifts a little every month, and nobody in stores, accounts or purchase can say where the missing pieces went.
- Is there a status for stock lying with a subcontractor?
- Does an overdue return raise an alert to the buyer?
- Can the processor’s rate and invoice be matched to the delivery note?
- Does the scrap at the processor reduce the right lot?
Test partial dispatch with a real order
Orders seldom leave on one truck. Ask the vendor to split a sales order of 500 units into 200 now and 300 later, then watch the invoice, the packing list and the balance quantity.
Check the lot numbers. If the first dispatch draws from lot A and the second mixes lots A and B, the packing list should say so without anyone typing it.
Then try one nasty case. Dispatch, cancel the invoice, redo it, and watch whether the balance quantity and the invoice number survive without a manual fix from someone in accounts. Many tools stumble here.
Follow a complaint back to its lot
Months after dispatch, a customer writes that a batch of bracket assemblies cracked, and the vendor must start from the customer name and invoice alone. Ask it to reach the heat number, the machine, the shift and the inspection record without leaving the screen or opening a spreadsheet.
Count the clicks. If the trail needs three exports to a spreadsheet and a phone call to stores, nobody will use it when a customer is shouting. Speed matters more than polish.
Ask where the corrective action lives. An email thread teaches the plant nothing. A cause field on the complaint does.
- Can a return be logged against the original dispatch line?
- Does the complaint link to the inspection record of that lot?
- Is there a field for cause, such as material, tooling or handling?
Run the same script with every vendor
Write the test steps before the first call, using an awkward part, a job sent out for treatment and an order that ships in two parts. Give every vendor the same sheet.
Score each answer as shown live, promised later or not possible, and insist that every promise carries a written date, signed by someone senior at the vendor.
Use last quarter’s three worst jobs.
Vendor demo scorecard
Five rows cover revision control, outside processing, partial dispatch, complaint trace and reports, with three columns per vendor headed shown live, promised with date, not possible. A notes box names the sample part.
Open a blank worksheet to printQuestions owners ask
Should the vendor demo use our data?
Yes, and send two or three real part numbers, a subcontracted job and a split order beforehand so the vendor cannot hide behind tidy sample records. Sample data hides odd naming, long part numbers, and jobs that skip an operation, which is exactly where many tools stumble once your own records go in.
How many vendors should we shortlist?
Three vendors is enough. A fourth adds another week of demos, another set of notes and another round of follow-up calls while the buying decision slips further away. Run the identical script with each, then lay the scorecards next to each other and read the promised-later column before anyone mentions price.
Who from our side should attend the demo?
Send the planner, the stores person, someone from quality and whoever raises invoices, because each of them spots a different gap in the demo. Each will live with the tool daily. Ask the stores person to bring a recent delivery note from a subcontractor, and ask the invoice clerk to bring a split order from last month.
Is a custom build better than a packaged product here?
It depends on how unusual your outside processing is, because packaged tools cover the common flow of send, process and return without much fuss. A custom build suits trades with odd steps, such as multi-stage plating or a customer who owns the tooling, but it needs a clear written specification first.