Scoring and bands
In the portal: Sequences. Scoring and bands are set on the sequence, not on an atom.
What it is
A questionnaire is an ordinary sequence. What makes it a questionnaire is scoring: a rule that turns the answers into one number, and optionally a word for what that number means.
Two settings do all of it. A strategy says how the answers combine. Bands say which number ranges get which word. Everything else on this page follows from those two.
The score matters beyond the questionnaire. It is what a journey branches on, and it is what draws the line on the participant's Progress tab, so a sequence with no scoring can only ever record answers, never react to them.
What it looks like
Nothing changes on the capture screen. The participant answers the questions exactly as they would in an unscored sequence, and they are shown no running total.
The score appears afterwards, on the Progress tab, as a line running left to right with a dot per submission and the bands shaded in behind it. The band shading is behind the line, never in front, because most scored sequences have no bands and the line has to read on its own.
A score chart needs at least two points in the window being viewed. With exactly one scored submission the app shows the line "Only one scored response in this period." in place of the chart. So a one-off baseline questionnaire never draws a line, however long you wait: every answer on file belongs to the same single occasion, which makes them versions of one thing rather than points over time. Recurring delivery is what produces a line.
When you would use it
You add scoring when the total is the thing you care about, rather than the individual answers.
A worked example. A sample tenancy's four question confidence check asks each
question on a 0 to 4 scale. On its own that is four numbers nobody will read. Scored
as a sum with bands at 0 to 5 "getting started", 6 to 11 "building", and 12 to 16
"confident", it becomes one word the participant recognises and one number the
journey can branch on.
Add scoring where at least one answer is a number, a slider, a select or a yes/no. Free text and dates do not contribute, so a sequence made only of those records answers rather than a score.
The controls
Scoring is one optional block on the sequence.
| Setting | What it accepts | Default | What it does |
|---|---|---|---|
strategy | sum, average, max, min | none, required once you add scoring | How the contributions combine into one number. |
bands | a list of { label, min, max } | none | The words. Both ends are included, so a band of 0 to 4 covers a score of exactly 4. |
Cover the whole range, with no gaps and no overlaps. A gap gives a score with no word: with bands of 0 to 4 and 6 to 10, a score of 5 is stored with no band at all. Where two bands overlap, the first in the list wins.
The portal warns you about both as you author. Bands written in an import file are yours to check, so read them through before you load them.
Bands also decide the chart axis, and nothing else does. With no bands the line
fits itself to the data, so two answers one point apart draw as a dramatic climb.
With bands, the axis is pinned from the lowest band's floor to the highest band's
ceiling, so the same two answers read as the small change they are. A score that
lands outside the band range widens the axis rather than dropping off the chart. A
number or slider atom's own min and max do the same job for that individual
question's chart, so set those too.
The four strategies.
| Strategy | The score is | Reach for it when |
|---|---|---|
sum | every contribution added up | Total burden. This is what nearly every standard questionnaire uses. |
average | the sum divided by how many atoms contributed | The number should stay on the same scale as one question, so a shorter or longer form still reads the same. |
max | the single highest contribution | The worst single answer is what matters, whatever the rest say. |
min | the single lowest contribution | The weakest link is what matters. |
Which atom types can contribute. Five, and only five.
| Type | Contributes |
|---|---|
number | the figure the participant typed |
slider | the value they landed on |
single_select | the chosen option's score |
multi_select | the scores of everything ticked, added together |
boolean | the chosen option's score, or 1 for yes and 0 for no when no score is authored |
text, datetime, breathing and blood_pressure never contribute, and neither do the three types
that capture nothing. An atom that could contribute but was left unanswered simply
does not.
Give every single_select option a number, including the one worth zero. An
option with no score drops out of the total entirely rather than counting as zero.
Under sum that looks the same. Under average it does not, because the question
also leaves the divisor. A missing score on the "not at all" option is the commonest
cause of a total that reads wrong.
The other two behave differently. A multi_select always contributes, adding up
whichever of the ticked options carry a score, so an entirely unscored one
contributes zero. A boolean always contributes too, falling back to 1 and 0.
Weight. Every atom has a weight, which defaults to 1. The atom's contribution
is multiplied by it before the strategy runs. A weight of 2 on one question makes it
count double. It is a multiplier and not a percentage, so weights do not have to add
to anything. Setting one question to 2 and leaving the rest at 1 makes the maximum
possible total larger, which means the bands have to move too.
Where a score is computed
On the participant's phone, as they submit. The phone works out the number, looks up the band, and sends the answers, the score and the band together. So a questionnaire scores the moment it is finished, with or without a connection, and the score that reaches you is the one the participant saw.
The platform stores what it was sent and never recalculates it. So the number on file is the number the participant was shown, and it stays that way.
Three things follow from that.
A score is a record of the rules as they stood that day. Change the bands next month and every submission already on file keeps the band it was given. Nothing is silently rescored.
A sequence with nothing to contribute produces no score at all, rather than a zero that would read as a real answer. If nobody answered a scoring atom, or the sequence has none, there is no number and no band, so give any journey branch reading that score a path for the case where no score arrives.
A band is a lookup, not a calculation. The first band whose range contains the score wins. The list does not have to be in order.
Asking a question only when it is needed
A questionnaire does not have to ask everybody everything, and this is separate from scoring.
Every page and every atom can carry a condition. Inside a running sequence, that condition can read the answers given so far on earlier pages, by naming the atom. So a screening question answered above zero can reveal a follow-up page that nobody else ever sees.
Two rules make this behave.
A page whose atoms are all hidden is dropped, and the pages renumber. The progress bar counts what the participant will actually see.
A page condition can show or hide; an atom condition shows. So write variants as separate positive conditions rather than one positive and one negative.
Built-in screening instruments are a different mechanism
Standard clinical screening instruments are built into the platform, scored to their published rules, with their severity labels fixed. Nobody can shift a published threshold from the portal. A change to one of these is an engineering change, with clinical sign-off behind it.
Everything on the rest of this page is about the questionnaires you write. If you need a standard instrument scored to its published rules, ask for the built-in one rather than rebuilding it as a scored sequence.
Worked settings
A summed screening questionnaire. strategy: sum, each question a single_select of four
options scored 0 to 3, bands covering 0 up to the maximum possible total in
non-overlapping ranges. The score is the total and the band is the severity word.
A worst-item alert. strategy: max, the same options, no bands. The score is the
highest single answer, so one severe answer among nine mild ones still reads as
severe, which a sum would hide.
A confidence average. strategy: average, six sliders from 0 to 10, bands 0 to
3.9, 4 to 6.9 and 7 to 10. Adding a seventh question later does not shift what a
given score means, which a sum would.
Next
Two settings, a strategy and a set of bands, turn a set of answers into one number and one word. That number is what a journey branches on and what draws the line on the participant's Progress tab, so it is the difference between a sequence that records answers and one that reacts to them.
Conditions is where a score becomes a branch.
Charts is what the score and the individual answers draw for the participant.