How to read a People Intelligence Insight
Every Insight has an honesty classification that tells you how much weight it can hold, a date that tells you how current it is, and filters that only appear when they actually do something. This article explains all three, and what a card means when it has nothing to show.
The honesty classification
Every Insight in the People Intelligence Catalogue has a classification. In the catalogue it is the icon on each card (hover it for the meaning), and it is written out as a label when you open an Insight to add it. It is the Insight's own statement about what kind of claim it is making. There are four.
Factual. Directly observable: a count, a rate or a status read from a record someone created. How many mandatory trainings are overdue. How many people left last month. Where the current review cycle stands. If you disagree with a Factual Insight, the disagreement is about the underlying record, not about the method. These carry the most weight and are the right basis for operational action.
Diagnostic. Combines several signals to explain a situation: a behavioural engagement index, a team health index, a behavioural profile, a network centrality score. It is designed to point you at where to look. It is not a measurement of a person's worth, motivation or ability, and it should never be the sole basis for a decision about an individual. Treat a Diagnostic Insight as the start of a conversation.
Predictive. A validated forecast of something that has not happened yet: read it as a probability, never as a fact. A Predictive Insight can only enter the catalogue with its validation recorded. No Predictive Insight is offered in the catalogue at the moment.
Experimental. A hypothesis still under validation. It is offered because it is useful, but it must not drive decisions about people on its own. Use it to open a development conversation, never to close one. Experimental Insights carry an Experimental badge, in the catalogue and on the card on your dashboard. Today the two Manager Index Insights are marked Experimental.
On your dashboard, cards do not repeat the Factual, Diagnostic or Predictive label; only Experimental is shown on the card. Check the classification in the catalogue before you rely on an Insight.
What the card header tells you
data from a date: the date of the data the card was built on. When that date is more than a day old, the number of days is added next to it. Hover it to see the date of each source the card reads.
The help icon next to the title, when there is one: how the Insight was computed.
Experimental: see above.
Needs refreshing: the Insight was computed with an older version of the metric. Use Refresh Insight to recompute it.
Unavailable: the Insight behind this card was retired or changed. The last view is kept so the card does not vanish without explanation, but it is never recomputed. You can remove it.
Insight saved from chat: an older card that came from a Data Analyst Agent conversation. It is a frozen snapshot: it cannot be refreshed or filtered. Charts you pin in the Data Analyst Agent today stay in its Pinned charts list and are not added to this dashboard.
Why an Insight can disappear from the catalogue
Insights have a lifecycle. An Insight can be withdrawn from the catalogue, temporarily while something is corrected or permanently, for example when a better Insight replaces it. A withdrawn Insight cannot be added. A card you already had stays on your dashboard marked Unavailable, instead of disappearing or being silently recomputed with a different meaning.
The data date
The date comes from the source snapshot, not from the moment you loaded the page, so a card never claims to be current when its source is not.
Behavioural Insights (engagement, archetypes, team health, the organisational dimensions) are rebuilt by a weekly pipeline, usually at the weekend. Expect the date to move once a week.
Mandatory training Insights follow a daily extract and carry their own, more recent date.
Evaluation and talent-review Insights reflect the latest extract of your review data.
Cards are served from a cache so the page loads quickly. Use Refresh Insight on a card to recompute it against the latest available data. A card also keeps the language it was computed in (Portuguese or English, following your organisation's default language); refreshing it brings it into the current one.
Filters
A filter button appears on a card only when that Insight genuinely supports filters. It shows the current selection, for example Whole organisation. The filters are:
Team(s). Narrows the Insight to one or several teams.
Set of people. Narrows the Insight to a chosen set of employees, for a specific cohort you want to follow. Available on the Insights that rank people.
Team (talent review). On the 9-box. It uses the team recorded in the talent review rather than the platform team, because those are two different lists and mixing them would produce a population that matches neither.
Director and Manager. On the employee detail table. The director is derived from the manager's manager.
Review period. A named review inside a cycle, on the review status Insight.
Period. Only on the Insights that have a real time window: turnover, exits per month, learning activity, the onboarding cohort and the archetype history.
Compare with. On eNPS and the organisational dimensions: the current result is always shown, compared with the previous snapshot or with 3, 6 or 12 months ago.
Most Insights describe the current state and therefore have no period filter at all.
Team(s), Set of people and Team (talent review) are alternatives, because each one defines the population and a card has one population. The interface lets you use only one of them at a time.
To compare segments, add the same Insight twice with different filters. Two cards side by side is a comparison you can read; one card you keep re-filtering is a comparison you have to remember.
Small groups and suppressed results
Insights built on anonymous survey responses (eNPS, wellbeing by team, the Team Health Index, the organisational dimensions) are not computed for a segment that is too small, and the card says so: fewer than 10 people for eNPS and wellbeing, fewer than 5 for the Team Health Index and the organisational dimensions. Filtering a five-person team out of an anonymous survey would let you work out who said what, which would break the promise made to the people who answered.
Insights that list named employees behave differently: they do not suppress, because naming people is the point of those cards and a count threshold would protect nobody. What controls those is who is allowed to open People Intelligence at all. If you believe someone can see it and should not, raise it with your administrator immediately.
Cards that say they have nothing
A card with no data still appears, and tells you why. That is on purpose. A blank space is indistinguishable from a bug, and an Insight that silently disappears when it has nothing to say teaches you to distrust the whole dashboard.
No data: the card says there is no data for this Insight, or not enough to show it, for your organisation or for the population you selected.
Segment too small: see the section above.
Partially available: some Insights combine several independent sources and degrade item by item. The calibration headline band may show two real numbers and two explained absences, rather than pretending to be complete.
Five reading habits worth having
Check the classification in the catalogue. A Factual list is an action list. A Diagnostic ranking is a place to look. An Experimental card is a hypothesis.
Read the description before the chart. Every Insight states what it is not, and most misreadings are of things the card already warned about.
Check the date. A card several days old is marked as such. Refresh it before a meeting if its source has moved.
Check the population, not just the number. Several Insights cover only people with a formal review, or only people present in two sources at once. The card says which.
Remember that behavioural scores are internal. They are normalised within your organisation and adapted to the modules you use. They do not compare to another organisation, and they do not measure a person's engagement as a human being. They measure what happens on the platform.