Forty-five minutes,
and you set them.
Email the agenda and the scenarios two or three days before the meeting. This is not adversarial; it is the opposite. A vendor who has seen your scenarios can prepare, which means you see the product doing your work rather than watching someone improvise. It also tells you something immediately: a vendor who ignores the agenda and opens with a forty-minute feature tour has told you how the relationship will go.
Keep the feature tour out of the meeting. Every vendor has a recorded one and it is the least informative forty minutes available, because it is the same for every school. Your forty-five minutes are for the things that are specific to you, and there are not many of them.
Bring real data,
including the messy parts.
Take a real week of your timetable, three real families (names changed), one invoice with an instalment plan, one with a failed payment, and a teacher absence. Then ask the vendor to run these live, in this order, because each builds on the last.
1. Enrol a student. Into a named class, with a payment, ending at a portal login. Watch whether a parent could have done it alone. 2. Change something. Move that class to another day and ask to be shown the teacher's view, the family's view and the invoice afterwards. This is the scenario that reveals the architecture. 3. Mark a register. Ask to see it as the teacher would, on a phone if that is how your teachers work. 4. Handle a failed payment. Show me what the family sees and what the office sees. 5. Produce Friday's picture. Arrears and falling classes, without exporting to a spreadsheet. 6. Answer a data request. A parent wants everything held about their child — show me how, and tell me who inside your company has to be involved.
The sixth one is worth asking of everyone, and the answer is often more revealing than the demo. Our own answer, for the record, is on the vendor questions page: SprintUp can export everything held about an account, and today that operation is run by us rather than by your administrator. Ask every vendor the same thing and see who tells you before you have to find out.
Numbers you can compare
across three demos.
Impressions do not survive a fortnight of demos; counts do. Keep a sheet with these five columns and fill it in as each scenario runs.
The phrase to listen for is "you would". "You would just add the student here, and then you would go into billing and add them there as well" is a description of a workflow, delivered smoothly, that contains a double entry you would otherwise not have registered. Count each one. Three or four in a single demo is the signature of a product assembled from modules rather than built around a student record.
The most informative
fifteen seconds.
In roughly every second demo, something will not work: a page will not load, a feature behaves differently than the presenter expected, a scenario turns out to need a setting nobody switched on. What happens next is the single best predictor of what support will be like, and it is free information.
Good: "That is not how it works — here is what we would actually do, and here is the limitation." Also good: "I do not know, I will find out and send it in writing." Less good: changing the subject, or explaining at length why your scenario is unusual. Worst: asserting that it works and moving on quickly. You are not testing whether the product is perfect — none are — you are testing whether the vendor tells you the truth when it is inconvenient, because in month six every question you ask will be an inconvenient one.
A trial on your own data,
or it did not happen.
No demo, however well run, survives contact with your actual term. The final step of the meeting is arranging trial access with your own data in it — not the vendor's sample school — and agreeing what you will test: usually the same scenarios plus a week of real use by the person who will use it daily. Two weeks is enough for a small school. A month is enough for anyone.
Use the trial to test the two things a demo cannot show. Support: send a real question mid-trial and time the answer — not to a sales contact, but through the channel a customer would use. And friction over repetition: a workflow that takes three screens is fine once and grinding on the fortieth enrolment, which only becomes apparent when someone does it forty times. Where a product offers a genuinely free tier rather than a time-boxed trial, that is worth more, because you can keep one real class running on it while you decide. SprintUp's free plan exists partly for this reason: one educator and ten students, permanently, with no card.