The Insights Agent works best when you build the conversation in layers. One question rarely gets you to the insight; a few focused follow-ups usually do. We call it the progressive prompt method — and the Agent helps by suggesting the next question for you.

The Agent won't make the decision for you — but the data will point the way.

After each answer, the Agent proposes follow-up questions — for example, "Which broadcasts had the highest revenue in the last 12 months?" or "How has broadcast-attributed revenue trended over time?" Click one to go a layer deeper without retyping. This is the progressive method built into the product: each click narrows the question.

For each question, the Agent:

Broad questions can return a sample rather than the full dataset. When that happens, the Agent flags it — for example: "the results are based on a sample… the query hit the row limit… the figures reflect what's visible, not the full total." Each query reads up to about 1,000 rows. A narrower question stays within that limit and returns a complete, defensible number — which is why the four-step method gives you figures you can take into a meeting, while "how are we doing on marketing?" gives you a vague, and possibly partial, answer.

A multi-property group used the Insights Agent on their own, with no help from Bookboost. In three prompts, they found that 20% of OTA bookers who returned switched to a direct channel on their second booking — enough to justify a loyalty and OTA win-back campaign, with the opportunity confirmed before they spent budget designing it.

Rephrase or narrow the question — often a clearer metric name, a tighter timeframe, or a smaller scope is all it takes (and it avoids the sampling above). Use the thumbs-down on an answer to flag it. If a chart looks structurally wrong, raise a support ticket — see Getting support for Insights.

Contact us through the Talk to Us option on the left menu in the platform, or email support@bookboost.io.