What Recommendations Look For

Once a night, Better.AI reviews your donors, recurring gifts, pledges and events, and puts what it finds on the Recommendations tab of your dashboard. Each finding arrives with the figures behind it, so you can judge it rather than take it on trust.

This article lists everything it checks, explains what to do with a card, and covers why the tab is sometimes empty.

How a finding is made

Better.AI does not read your database looking for interesting things. It runs a fixed set of checks, in plain database queries, and finds the records that meet them. Only then is a sentence written explaining the finding.

That order matters when you disagree with a card. Every recommendation traces back to a number, which means you can check it, and it means a card you think is wrong is usually a threshold that does not suit your organization rather than a mistake.

What it checks

Money you are about to lose

Money on the table

Something worth looking at

Duplicate records are the one to act on first. A donor split across two records has half a giving history on each, so they sit below every threshold above and are invisible to every other check. Merging them does not just tidy your data, it changes what Better.AI can see.

What to do with a card

Each card carries three controls.

Open takes you to the contact. It does not clear the card, because looking at a donor is not the same as dealing with them.
The check mark marks it done. The card is retired and will not come back.
The cross dismisses it. Also retired, and it tells us this check produced something not worth your time.

Both retiring actions are permanent for that finding. They are recorded separately because they mean opposite things: dismissals are how we learn which checks earn their place, so use the cross rather than the check mark when a card was simply wrong.

Why a card came back

A finding is retired at the size it was found. If the figures behind it move materially, for example a lapsed donor lapses further or a pledge falls further behind, it becomes a new finding and appears again. That is intended: the situation changed.

Today's suggested actions

The same findings, ranked for one person and shown as a strip on your dashboard. The ranking is arithmetic rather than an opinion, and every row shows the terms behind its position, so an order you disagree with can be traced to a number.

Rows are filtered to the contacts assigned to you where your account records ownership. Organization-level findings, such as an event running behind, appear for everybody. Each row turns into a real task on the contact in one click.

When the tab is empty

Three different things look the same at a glance, so the tab tells you which one you are looking at.

Better.AI is not switched on. Nothing has examined your data, so the tab offers to switch it on rather than reporting that nothing was found.
It has not looked yet. The header says when it last checked. If that says never, or a date that surprises you, click Check now to run the checks immediately rather than waiting for tonight.
Nothing needs your attention. A genuinely quiet night. The header will show a recent check, and there is a link to see examples of what would appear.

How the monthly sample affects this

The checks run every night whatever your plan, and every finding is stored. What your plan decides is how many findings get written up in sentences each month.

Past that number, cards still appear with the figures behind them. What they arrive without is the paragraph explaining them, which for most findings is a convenience over numbers you can already read. The Better.AI add-on writes up every finding.