YOUR NEXT SUPERPOWER · FREE COURSE

Make a weekly report you can trust

A checked summary, transparent calculations and a repeatable review routine.

MARKETING + PRODUCT + OPERATIONS · 45–60 minutes, suggested practice time

This course is free, including all prompts and worksheets. Use any suitable AI tool you already have; third-party tools may charge separately. This page does not run an AI model. Notes stay in this page session unless you explicitly save them in this browser; export to keep a portable copy. Avoid confidential information.

Your practice input · fictional data

week,channel,spend_usd,leads
1,Search,240,12
1,Social,160,8
2,Search,300,15
2,Social,200,5

STEP 1 OF 4

Define the metrics

Use the synthetic CSV. Spend is in USD; leads are counts. Cost per lead is spend divided by leads. Compare matching weeks and channels.

Use this prompt

Inspect this CSV before analysis. State columns, units, missing values and assumptions. Define total spend, total leads and blended cost per lead. Do not add outside data.

My approved input or previous output: [paste here]

Finish with: A metric dictionary and input-quality check.

My step 1 work and evidence

STEP 2 OF 4

Calculate and reconcile

Calculate week and channel totals. Recompute at least one result manually or in a spreadsheet. Never average channel cost-per-lead figures without weighting.

Use this prompt

Calculate each week’s total spend, total leads and blended cost per lead. Show the arithmetic, then show channel changes. Use null for division by zero. Separate observed changes from explanations.

My approved input or previous output: [paste here]

Finish with: A calculation table with independently checked totals.

My step 2 work and evidence

STEP 3 OF 4

Explain what changed

State the change and its limits. These rows do not establish why performance changed or whether a campaign is profitable.

Use this prompt

Write a five-bullet weekly summary using only the checked table. Include absolute and percentage changes, caveats and one investigation question. Do not invent causation, revenue or ROI.

My approved input or previous output: [paste here]

Finish with: A plain-language summary and one follow-up question.

My step 3 work and evidence

STEP 4 OF 4

Create a repeatable routine

Write who supplies data, who checks it and who approves sharing. An automation can prepare a draft after the manual routine is understood.

Use this prompt

Turn this checked process into a draft-only weekly routine with input validation, duplicate detection, an owner, a human approval step and stop conditions. Do not connect accounts or send reports.

My approved input or previous output: [paste here]

Finish with: A reporting checklist with an owner and a stop rule.

My step 4 work and evidence

Ready for a human review?

Totals and denominators were checked independently.

Observed changes are separate from causal guesses.

The report has a human reviewer before sharing.

Checked boxes reflect your own review; they are not an independent validation.

Choose your next move.

Explore these starting points; review current capabilities and permissions on each listing: Julius AI.

Open the companion kit ↗ · Compare on a fixed reporting task ↗

Work through your project with Ashley ↗

Get more practical missions in Power Notes ↗