← All Kits · All Guides · Data Migration Series

Stage 6: User Acceptance Testing

Stage 6 of 8 · The client's own people check real records · Part of the Analyst Prep Kit

User acceptance testing is where the client's staff open real records in the new system and say whether they are right. It is the last gate before anything becomes permanent, and it fails for one reason more than any other: people are asked to look around instead of being asked something specific.

The short version. Give named people named records and a named question. Vague invitations produce silence, and silence at this stage becomes a dispute later.

Ask a question, not for a favour

Which of these two do you think gets a reply? Decide, then read.

"The sandbox is available, please have a look and let us know if anything seems off."

"Please open these ten clients, check the visit history for 2023 against the old system, and tell me by Thursday whether anything is missing."

The second one gets a reply, because it names the action, the records, and the day. That is not a communication trick, it is the finding from Gollwitzer and Sheeran's meta-analysis of ninety four tests: an intention converted into a specific when and how produces a medium to large improvement in follow through. Apply it to every request you make of a client.

Who should test

Not the project sponsor and not IT. The people who use the records every day. A scheduler will spot in four seconds that a visit type looks wrong, and nobody senior to them will.

Ask for two or three of them by name at kickoff, and warn the client then that these people will need real time set aside. Testing squeezed into the gaps of a working day is testing that does not happen.

What to give them

Build the test pack yourself. Do not make the client invent it.

Include the de-duplicated records on purpose. Those are the ones where a wrong decision does the most damage, and the client's staff are the only people who can confirm the merge was right.

What counts as a pass

Agree this before testing starts, or you will be negotiating it while people are annoyed.

A pass is that the records match what was agreed in the field map. It is not that the new system works exactly like the old one, and it is not that data the client never had has appeared. That distinction is why the signed map matters, and it is worth restating in the email that opens UAT.

Findings sort into three buckets: a genuine migration defect that you fix, a difference that was agreed in the map and needs explaining, and a new request that is out of scope. Name the bucket for each item when you reply, politely and every time.

The gate

The stage closes when the client confirms in writing that the sample is acceptable. In practice this is almost never a formal signature. It is an email from you saying what was tested, what was found, what was fixed, and asking them to confirm it now meets expectations, with the go-live date restated.

Copy the person who signed the contract. Not as a threat, as a courtesy. It also means the confirmation exists somewhere other than one inbox.

The one habit

Never end UAT on silence. No reply is not approval, and it is the single most common way a migration turns into an argument three months later.

When you have been asked to test something at work, what made you actually do it?

Up next

Previously: Stage 5: The Dry Run  ·  All eight stages

The checking work is SQL, and it is not advanced SQL.

Counting rows, grouping to find duplicates, and joining to find orphans covers most of what a migration asks of you. The SQL Kit teaches those moves in order, with the data in front of you.

Open the SQL Kit →

Or start typing straight away: open SQL Drill, thirteen queries that each add one thing to the last.

The whole run, in one place.

You are moving a client’s records into a new system, and the part that worries you is the week the client goes quiet. The Data Migration Playbook is all eight stages in order, 63 pages, from the first kickoff meeting through hypercare, so you can run one without finding the important question three weeks too late.

The Data Migration Playbook, $29 →

References

  1. Gollwitzer, P. M., & Sheeran, P. (2006). Implementation intentions and goal achievement: A meta-analysis of effects and processes. Advances in Experimental Social Psychology, 38, 69–119.
  2. Matthes, F., Schulz, C., & Haller, K. (2011). Testing & quality assurance in data migration projects. 27th IEEE International Conference on Software Maintenance (ICSM), 438–447.