Research

Analyze trends in spreadsheet data

Separate sustained change, seasonality, and one-off noise using explicit time periods

7 min setupTested with:ChatGPTReviewed: 2026-08-28

Add your context

Your text stays in this browser. AILesson Prompts does not send it to a model or server.

Metric and decisionTime fields and sample dataKnown events and limitations

Your prompt

Unfilled fields remain visible as placeholders, so you can still copy and edit the prompt

Design and interpret a spreadsheet trend analysis.

Metric definition and decision:
[metric]

Time fields, frequency, segments, and representative data:
[data]

Known events and limitations:
[context]

Validate time coverage and metric consistency before calculating change. Return period aggregation rules, formulas or PivotTable layout, absolute and percentage change, an appropriate moving average when justified, segment comparisons, and a chart recommendation. Distinguish sustained trend, seasonal pattern, level shift, and isolated spike. Treat links to known events as hypotheses unless the data establishes causality. Include checks for missing periods, changing denominators, partial current periods, and revisions.

Private by defaultPrompt assembly happens locally in your browser. Avoid placing confidential information into any AI service unless your organization allows it.

From input to outcome

A worked example

See how concrete context turns this recipe into a usable result

Actual input

Metric and decision: Weekly online conversion rate = completed orders / checkout starts. Use it to decide whether a recent decline needs investigation.
Time fields and sample data: One row per week from 5 Jan to 29 Jun 2026, with WeekStart, CheckoutStarts, CompletedOrders, Device. Overall rates for the last six complete weeks: 4.8%, 4.7%, 4.6%, 4.1%, 4.0%, 3.9%. Mobile checkout volume is roughly 70%.
Known events and limitations: The latest week is partial and must be excluded. Tracking definitions changed on 11 May. A mobile checkout release occurred on 1 June. No comparable prior-year data is available.

Example output

Exclude the partial week and split interpretation at the 11 May definition change; pre/post values are not automatically comparable. Calculate each rate from summed orders divided by summed starts, not by averaging row percentages. Add absolute percentage-point change and relative change.

The six complete weeks show a monotonic decline from 4.8% to 3.9%: −0.9 percentage points, or −18.75% relative. The larger step from 4.6% to 4.1% suggests a possible level shift rather than random week-to-week movement, followed by two smaller declines. This is evidence for investigation, not proof that the 1 June mobile release caused it.

Build a line chart with raw weekly rate by Device and overall rate; add a three-week moving average only as a secondary line. Mark 11 May and 1 June as annotations. Compare Mobile and non-Mobile using weighted rates and show weekly denominators. If the decline is concentrated on Mobile after 1 June and definitions are stable within that window, the release becomes a stronger hypothesis. Check missing weeks, denominator drops, delayed order completion, and later revisions before deciding.

Why this works

1
A fixed metric and period definition prevents apparent trends caused by changing denominators.
2
Separating pattern types produces more cautious and actionable interpretations.