Theoretical models

A theoretical model is the framework used to measure one of the two fixed sections. It defines which categories exist and how they combine into the index.

It is not an aesthetic preference: it is the instrument. Changing it changes what gets asked and how the index is computed.

One per index, not one per questionnaire

Each questionnaire points at two models: one for Satisfaction and one for Climate.

That follows directly from the two indices being parallel and never combined. They are two different questions about the same people, so choosing the theory behind one says nothing about the other. See Satisfaction and Climate.

What you choose is the Satisfaction model. Climate is measured with a single model, the same for every account, and has no selector.

The curated catalog is not edited

The catalog models ship with the product and are curated: their questions cannot be rewritten, and that is deliberate.

Comparability rests on a strong claim: two companies choosing the same catalog model measure exactly the same thing, with the same instrument. If every account could rewrite those questions, that claim could not be made about any pair of organizations, and cross-company comparison would go with it.

What is specific to your organization has two homes. Individual questions go in Section 3, which is editable precisely so this never has to be (see Questionnaires and inheritance). And if what you want is to measure Satisfaction with your own framework, you can write a custom theory.

Custom Satisfaction theories

A custom theory is a Satisfaction model written by your organization: its categories, its questions and how much each one weighs. Once created it works like any catalog model: it is chosen in Settings → Survey forms and in the study wizard, it has its own questionnaires, and it produces results, history and reports like the rest. In the selectors it carries the «Custom» label.

You create one from Settings → Survey forms → Custom theories, blank or by copying a catalog model (or another custom theory) as a starting point. Copying creates nothing until you save.

What you decide

• The categories and their questions, with whatever names and wording you want.
• How much each category is worth in the index. All the same, or the percentage you give each one (together they must add up to 100 %).
• How much each question is worth within its category, chosen per category: by order of importance (the first one weighs most), all equal, or the percentage you type. While you type, the editor shows how many index points each question is worth; the final figure is computed by the system when you save.
• The scale of each question: 1 to 3, 4, 5, 7 or 10, with a default scale for the theory. Only the 1 to 4 scale carries words (Nada, Algo, Mucho, Completamente, which are part of the instrument and are not translated); the others are answered with numbers.

It freezes once a study uses it

While no study has launched it, a custom theory can be edited freely. As soon as a study launches it, it freezes: categories, questions, weights and scales can no longer change, because changing them would reinterpret results already measured. The name and description can still be edited.

To change it afterwards, duplicate it: the copy is a new, editable theory that starts its own history (see below for why).

A custom theory is only comparable with itself. Nobody else measures with it, so its results are read against its own series, inside your organization, and never enter any cross-company comparison.

Custom theories are included in every plan.

Changing model starts from zero

This is the part that surprises people, so it is worth knowing beforehand.

Results from one model do not mix with those from another. An account that has measured Satisfaction with one model for three years and tries another does not see three years of history in the new one: it sees day one. The same holds for a custom theory and its copy.

That is correct. They are two different instruments, and joining their series would assert a trend between two things that were not measured the same way. In practice a model behaves like a separate space: each has its own history, and they are independent. In Statistics you can choose which stretch to look at when your history went through more than one model.

Practical consequence: choose the model at the start and change it deliberately, never to try it out. A model change costs that index's history, and that cost is not recoverable.

If you want to see what another model looks like before committing, do it in a test account, not the one that has been measuring.

A model is not deleted: it is retired

When a model stops being offered it disappears from the selectors, but nothing already measured is touched. Studies measured with it keep their results, and their series still names it.

It had to work this way: deleting it would leave published results with no way to say which instrument produced them, which is the same as not being able to interpret them. The same goes for a custom theory: one no study has used can be deleted, and one already used is retired and can be reactivated.

Each model has its own questionnaire setup

In Settings → Survey forms, the selector at the top picks the model, and everything below belongs to that model: its questionnaires and which branch inherits which. Each model has its own setup of the tree, and inheritance only reaches questionnaires of the same model. Starting a new model therefore means starting with a tree without questionnaires: that is the normal state, not an error.

The study chooses the model, in step 3 of the wizard, and that decides which questionnaire each branch gets. Once launched, the study no longer changes model. What cannot happen is one study measuring with two models at once: the index would be the average of two instruments, which is not a number that means anything.