What it is and what it's for#
The dashboards tell you how the programme is going. Smart Recommendations tries to answer the next question: so what do I do this week?
Every day, the system looks at your project's data, compares it with the previous week's and writes a handful of specific recommendations, ordered by priority. They are not system notices or threshold alerts: they are sentences written about what has happened in your particular project.
It is a starting point for a meeting, not an approved action plan. Each recommendation comes with the figure that triggered it precisely so that you can overrule it if you know the context and the figure is misleading.
How to set it up#
There is nothing to set up. Recommendations are generated on their own, once a day, for each active project. You don't choose the topics, you don't adjust thresholds and there is nothing to schedule.
What is worth knowing is what it looks at, because it explains why it talks to you about some things and not others:
| Area | What it watches |
|---|---|
| Logins | How many people sign in and what share of the workforce is active. |
| Completion | Challenges completed, late, not started and awaiting approval. |
| Content | Which specific challenges get completed and which ones get stuck. |
| Rewards | The points being handed out and what is done with them. |
| Momentum | Live streaks and the pace of the programme over time. |
The comparison is always the last week against the one before. That is why a recommendation can appear one day and be gone the next: it isn't a fault, it is that the figure behind it has changed.
In the first couple of weeks of a project the recommendations are worth little: there is no previous period to compare against. Wait until you have a month of activity before taking what they say seriously.
Where to look#
The screen is titled Smart Recommendations and shows, under the title, how many recommendations there are and which day they are from. The Refresh button reloads the day's recommendations; it does not force a new generation.
This screen is not linked from the panel menus: today you reach it through its direct address. If you can't find it under Analyse, it isn't that you are missing permissions.
How to read a recommendation#
At the top there are three counters — High priority, Medium priority and Low priority — summarising the split. Below, one card per recommendation.
The priority label
HIGH, MEDIUM or LOW. The cards come already ordered: what is at the top is what the system considers most urgent.
The category
LOGINS, COMPLETION, CONTENT, REWARDS or MOMENTUM. It is there to show at a glance whether the problem is one of participation, of content or of the points circuit.
What the data says
The WHAT THE DATA SAYS block is the observation: the figure or the comparison that triggered the recommendation. It is the part you can check for yourself in the dashboards.
Recommended action
The RECOMMENDED ACTION block is the specific proposal. It is the part to read with most judgement: the observation is usually right, the proposal may not fit your organisation.
The segment affected
Which part of the workforce it refers to: a site, a unit, a group. It is what makes the recommendation actionable, because it tells you who to talk to.
Always read WHAT THE DATA SAYS first and RECOMMENDED ACTION afterwards, in that order. If you start with the proposal, it is easy to accept it without checking the figure it rests on; and the figure is the only thing the system knows for certain.
What happens next#
- Nothing happens automatically. A recommendation does not send a notification, does not change a date and does not assign a challenge. It is a piece of text: what is done with it, you decide in whichever module applies.
- There is no status and no follow-up. Recommendations are not marked as done or as dismissed. Each day they are replaced by the next day's.
- No consultable history is kept. The screen shows the latest batch; there is no "what were you recommending three weeks ago" view.
- If you act, the effect shows up in the dashboards, not here. See Analytics and reports and The admin panel.
Frequently asked questions#
There are no recommendations at all. Is it broken?
Almost always no. If the project is very new or has very little activity, there isn't enough to compare and the screen says so instead of making something up. Press Refresh and, if it is still empty, wait for the next day's generation.
Can I ask for them to be generated right now?
No. Refresh reloads the day's recommendations, it does not trigger a new generation. The batch is produced once a day.
Why are there so few?
Because they are capped at the highest-priority ones. A list of twenty recommendations doesn't get read; the cut-off is deliberate.
There was a recommendation yesterday that isn't there today. Have I lost it?
Yes, in the sense that it cannot be recovered from the screen. The list is rewritten every day. If a recommendation strikes you as important, note it down at the time.
A recommendation looks wrong to me.
It may well be. Check the figure in the WHAT THE DATA SAYS block against the corresponding dashboard: if the figure is right and the proposal doesn't fit, just discard it. There is nothing to do to reject it.
Is information about my workforce shared in order to generate them?
Aggregate programme figures are used — how many people sign in, how many challenges are completed, how the points are going — not the content of what people write or their survey answers.
Limits and warnings#
- It is a reading, not a verdict. It is written automatically from aggregate figures and can get the interpretation wrong even when the figure is right.
- It is never a recommendation about a person. It talks about segments and about content. It must not be used as the basis for a decision about a specific individual.
- The comparison is over seven days. A one-off dip because of a bank holiday or a workload peak can trigger a recommendation that means nothing.
- The list is rewritten every day and no history is kept. Whatever you don't note down is lost.
- It does not replace looking at the dashboards. It only covers a few areas, and there are problems it does not detect because it does not look at them.
- A project with little activity produces poor recommendations. With very little data, any variation looks like a trend.