The eight checks tell you what to test. Which of them will actually hurt depends on what your record follows. There are four common shapes — and picking yours before anyone builds a form saves you a rebuild in two years.
The routing page for Connected Data Intelligence. The eight checks tell you what to test. This page tells you which of them will bite hardest in your situation.
Before choosing a tool, decide what the record follows. There are four common shapes, and each one makes a different subset of the eight checks difficult. This is why two organisations can buy the same platform and only one of them ends up with something usable — they needed different things from it, and neither of them named which.
Pick your shape before anyone builds a form, because you will live with it for years. Two of these can look like the same request on paper and behave nothing alike: following a cohort for a decade and running a ninety-member network are both "we need to track our data," and almost nothing about the two set-ups is the same.
| Record shape | The question it exists to answer | What is genuinely hard | Checks |
|---|---|---|---|
| A · One person, followed for years | Did this person change, and did it last? | Recognising the same person years later, without asking them to remember anything | 02 · 04 · 08 |
| B · Several people describing one person | How is this person seen, and by whom? | Reading what people wrote while keeping who wrote it private — one problem, not two | 07 · 02 · 05 |
| C · Many programs, one picture | What is true across everything we run? | Comments arriving faster than anyone can read them, and programs that each defined things their own way | 05 · 03 · 08 |
| D · A network where each member sees their own part | What is true for me, and for all of us? | Many local people needing their own slice of one shared set of answers | 01 · 03 · 07 |
The quickest route is to answer one question: who or what does a row in your data describe?
If two of those sound right, you are probably not confused — see the section below.
A global skills body — 90+ member countries, around 1,600 competitors
Surveys sat in three different form tools, so no event could be compared with the one before it. One survey got 285 answers back from 1,600 competitors — enough to say something, not enough to stand behind. They want to ask a question at the event and return six months later, and to let around ninety country representatives see their own region's answers without asking anyone for a spreadsheet.
Breaks first on 03. The job is to reach everyone, not to sample — which is why paying by the response is the wrong shape here.
A patient association — 11,000+ members, English and Spanish
Their long-term data is collected by an outside research firm on a fixed schedule, and the association does not own the result. Their own earlier surveys asked people to invent a personal code; they forgot it, typed something different next time, and those answers could never be matched up. Open questions were left out altogether, because there was no way to use the answers.
Breaks first on 02. Nothing you do later rescues a round you cannot join to the one before it.
The King Center — Atlanta, seven programs, 10,000+ voices
Their data sat in separate platforms, and with nobody on staff to analyse it, the open-ended feedback simply never got read. Once before-and-after surveys from different cities were brought into one view, work that used to take months started happening during the training session itself. Read the King Center story.
Breaks first on 05. The comments were there the whole time. Nothing could read them at the speed the programs actually ran.
A leadership program — each participant rated by their manager, peers and team
Several people describe one colleague. The ratings have to stay attached to the right person while what each rater wrote stays private to them. The most useful finding is usually where two groups see the same person completely differently — which an average erases. See also 360 feedback software.
Breaks first on 07. Reading the comments and protecting who wrote them is one problem, not two.
Two of these are named and two are not — the King Center story is published and the 360 pattern has its own page, while the other two organisations are still mid-project. The shape matters more than the logo.
Most organisations of any size are, and it is not a problem as long as you name which one is primary.
A case-management organisation following clients for years is shape A. If each client is also described by a case worker, a clinician and a family member, it is shape B at the same time. A funder running several programs across a network of local partners is C and D together.
The rule is that the record follows your primary shape, and the others become attributes on it. For the case-management example, the record follows the client through time — that is the spine — and the different people describing them are held as a group attribute on each observation, not as separate records. Getting that the wrong way round is a common and expensive mistake: build the record around the rater and you will never be able to answer a question about the client.
If you genuinely cannot tell which is primary, ask which question you will be asked most often. Whatever the answer to that question is about, that is your record.
Because the shape determines which of the eight checks are hard, and every tool is good at some and weak at others. Choosing software before naming your shape means finding out during your second reporting cycle that the thing you most needed is the thing it does worst.
You can add a shape. Changing the primary one usually means rebuilding the record, because it changes what a row means — which is why it is worth ten minutes of argument now rather than a migration in two years.
These four cover most stakeholder-evidence work, but the useful part is the exercise rather than the list. Decide what a row describes, then work out which of the eight checks that makes hard. That reasoning transfers even if your situation is unusual.
Ideally yes, because the shape decides how identity works and identity is the thing you cannot retrofit. If you have already started, the shape still tells you what to fix first.
ChatGPT, Claude, and Gemini are fine for a quick test — but not for an answer you'll put in front of a funder or board. When it has to hold up, run it in Sopact Sense.
Try it in Sopact →
