ChatGPT Prompts for Marketing Analytics: 12 That Use Data

The most useful ChatGPT prompts for marketing analytics start from your own export, not a blank question. Paste a CSV of channels or campaigns, tell ChatGPT the periods and the one metric that matters, and ask it to check the data, explain what moved, and propose a test. The 12 prompts below follow a normal reporting week: clean, compare, explain, and act.

One rule for all of them: the model should only use numbers you gave it. Ask it to show its arithmetic, and spot check a few figures in your spreadsheet before you share anything.

Prompts to check the data first

1. Export sanity check

Here is my export: [paste]. List the columns and periods you see, any missing or zero values, and anything that looks like a tracking break. Do not analyze yet.

2. Metric definitions

For each column in [paste], write a one line definition and what it does not include. Flag any column where the name could mean two different things.

3. UTM and channel hygiene

Here are my top 50 source and medium pairs: [paste]. Group the ones that look like the same channel tagged inconsistently and suggest one naming pattern.

Prompts to compare periods

4. Period over period table

Compare [period A] and [period B] for [metric] by channel. Show absolute and percent change and cost per [metric] where spend exists. Show the formula you used.

5. Volume or conversion rate

For [channel], tell me whether the change in [metric] came from more or less traffic or from a better or worse conversion rate, using only these columns: [paste].

6. Small numbers warning

Which rows in this table have too few conversions to draw a conclusion? Explain your threshold as a judgment call, not a rule.

Prompts to explain what changed

7. Events timeline match

Here are events from the period: [launches, price changes, outages]. Which metric moves fit their timing? Label each "fits the data" or "needs checking", never "caused".

8. Attribution caveats

My numbers come from [GA4 last click, ad platform, CRM]. What could this source over credit or under credit in my channel table, and what should I check before cutting spend?

9. Funnel step drop

Here are counts for each funnel step by week: [paste]. Where is the biggest drop, and did it change between weeks?

Prompts to decide what to do

10. Three ranked actions

Based on this readout: [paste], give me three actions ranked by impact and effort: one budget question, one fix, and one test.

11. Test plan

Write a test plan for [idea]: hypothesis, primary metric, split, run length as a stated assumption, and the result that would change our decision.

12. Stakeholder summary

Turn this analysis into a five sentence update for [audience] with no jargon. Keep every number exactly as in the table.

Get the full channel readout in one prompt

The Marketing Analytics Readout: Channel Performance From Export prompt runs the whole sequence. Paste your export, the periods, your main metric, known events, and the attribution source. It returns a data check, a headline, a channel table with the arithmetic shown, a volume versus rate split, explanations labeled by confidence, attribution caveats, ranked actions, a test plan, and open questions. The example on the page compares August and September trial starts across four channels and traces a Paid Social drop to conversion rate, not traffic.

Tips for marketing analytics with ChatGPT

• Paste the data, not a description of it. Summaries hide the problems.
• Never let the model add industry benchmarks you did not provide.
• Remove customer names, emails, and other personal data before pasting.
• Keep the export small and focused on the question you are answering.

Related prompts

Fix messy campaign tags with the GA4 UTM Naming Convention SOP, and pin down KPI meanings with the Looker Studio Metric Definition Packet. Find more in the Marketing prompts hub, browse more prompts, or share a prompt from your reporting week.