"Rank my campaigns by revenue for September."

"Which campaign had the lowest click-through rate last month? Ignore campaigns with fewer than 1,000 impressions."

"Compare progress bars with popups over the last 30 days."

"Show campaign #12's daily impressions and orders for the last two weeks."

"Which of my experiments are running right now?"

"Did experiment #4 find a winner? Show each variant's conversion rate and lift over control."

"Is the difference in experiment #7 statistically significant?"

"How much extra revenue did Cooee bring in last month?"

"Compare conversion rate and revenue per visitor for visitors who saw Cooee and visitors held out of it, for Q3."

"Where do high-intent visitors drop off in my funnel this month?"

"Which pages do high-intent visitors leave from without ordering?"

"Which landing pages have the highest bounce rate on mobile?"

"Compare conversion for new and returning visitors from Instagram."

"Which products are viewed a lot but rarely added to cart? Only include products with at least 500 views."

"Show views, add-to-carts and units sold for my top 20 products last month."

"What were my revenue and orders last week, and how much came from Cooee?"

"Show daily revenue for October as a chart."

Give exact dates such as "1–30 September". Dates use your store's timezone.

Set a minimum (impressions, views or visitors) when ranking by a rate, so small numbers don't dominate the top of the list.

Ask the assistant to show its numbers before it draws conclusions, and check important figures against the Cooee dashboard before you act on them.

Call an experiment winner only when it is significant. The p-value must be below 0.05 and each variant should have a meaningful number of orders. See Reading lift and significance in experiment results.