ANALYTICS & REPORTING

Channel mix comparison

Use when you need to compare channels fairly rather than by whichever one the tool credits.

channel-mix-comparison.md

Download .md

Compare channel performance for {{BUSINESS}} without pretending attribution is settled.

Channel data: {{CHANNEL_DATA}}
Costs by channel: {{COSTS}}
Attribution model: {{MODEL}}
Sales cycle length: {{SALES_CYCLE}}

Output:
1. Table: Channel | Spend | Attributed conversions | Cost per conversion | Share of total conversions | Direction attribution likely biases this channel (over, under, neutral) and why.
2. The same table with one alternative reading: what the picture would look like if the channels you flagged as under-credited were credited 20 percent higher. Label this clearly as a sensitivity check, not a result.
3. Which channel comparisons are fair and which are not, given the model and the sales cycle.
4. One test that would settle the biggest disagreement, with the design, the metric, and how long it needs to run.

Rules:
- Do not declare a winning channel. Rank by cost per conversion and then state how far the ranking would have to be wrong to flip.
- Channels with fewer than {{MIN_CONVERSIONS}} conversions are directional only. Label them.
- If the sales cycle is longer than the lookback window, say that before anything else and explain which channels it penalises.
- No em dashes. Plain English.

Fill in before running

Replace each placeholder with your own detail. The more specific you are, the less the model invents.

{{BUSINESS}}
{{CHANNEL_DATA}}
{{COSTS}}
{{MODEL}}
{{SALES_CYCLE}}
{{MIN_CONVERSIONS}}

Getting a better result

Give real spend, including agency fees and production costs, or the ranking is fiction.
The sensitivity check is what to show when someone insists their channel is undervalued.
Take the test in section 4 seriously; it is usually cheaper than another quarter of arguing.

Questions about this prompt

When do I use this rather than the platform reports?

When two channel owners are quoting different numbers at each other. Each platform reports the version most flattering to itself. This puts everything against one model, names the likely direction of bias per channel, and refuses to declare a winner, which is what makes it usable with both owners in the room.

What do I need in front of me?

Channel data, real costs including agency fees and production, the attribution model, the sales cycle length, and a minimum conversion count below which a channel is directional only. Understated costs turn the ranking into fiction, and the sales cycle decides whether the lookback window can measure the channel at all.

What comes back?

A table with cost per conversion and the likely direction of attribution bias per channel, the same table re-read with under-credited channels lifted twenty percent as a labelled sensitivity check, which comparisons are fair, and one test that would settle the argument. The sensitivity table is what you show the owner who feels undervalued.

What is the mistake that costs me here?

Reading the sensitivity check as a result. It is a what if with a number attached and the prompt labels it as such. Strip the label and it becomes a second set of figures people start quoting. If the sales cycle is longer than the lookback, deal with that first, because every row below inherits it.