Ask AI to find trends (dynamics over time). Return JSON {"data":"...","metric":"...","period":"...","compare":"..."}: data - context; metric - what we measure in dynamics; period — at what enlarged step to look at the dynamics; compare — which data slices (groups by attribute) to compare with each other.
The trend is visible only at an enlarged step. 365 raw daytime points are noise: every weekend jumps, and the eye drowns. Group by week or month and a line will appear. Mechanics: a trend is the direction of the average over time, and the average must be averaged on something. Lever: set the enlargement step yourself. If you don’t set it, the AI will either show a mess of 365 numbers, or average everything into one number, and the dynamics will disappear. Insider: ask not only for direction, but also for RATE - “by what percentage per month” - and the turning point when growth gave way to decline. Direction without tempo is useless: “growing” by 1% and by 40% are different solutions. Second trick: compare fractions, not absolutes. Online +20% with overall market growth does not mean that the online share has increased; ask for a channel's share of time. A typical mistake: looking at the trend based on daily numbers and being afraid of every weekend dip - this is seasonality, not a recession.
Unlock access to submit solutions for instant AI review.