Deliver a repeatable, data-driven framework for reviewing recent ad-campaign performance — spotlighting spending efficiency, campaign and creative health, and suggested optimizations based on weekly and long-term trends.

The “Media Buying Health Check” prompt outlines a standardized way — using Polar MCP (and a connected LLM like Claude) — to generate a concise, one-page weekly audit of your paid media campaigns. The prompt specifies what data to use (e.g. spend, attributed conversions/revenue from first-party tracking, blended ROAS/CAC, channel and campaign breakdowns, and creative-level metrics) and organizes the output into sections including headline metrics, acquisition efficiency, campaign performance, creative performance, and recommended next steps.

​

1. Ensure you have the Polar MCP installed (see Using Polar MCP).

2. Create a new Project in Claude.

3. In the Instructions box paste the instructions from below.

4. To generate a report, start a new chat in the project with something like: “run the report”.

5. You can then:

   1. ask follow-up questions to go deeper

   2. tell Claude you want to amend the report

   3. customise the instructions yourself for future use

Instructions

Put these in a Claude Project “Instructions” fields.

Produce a brief media buying health check based on the last 7 days data

Focus:

* show the current overall marketing health in contrast to previous 7 days and 365 day average

* highlight any campaigns and creatives with significant spend but concerning performance, as defined by

* use Polar Pixel metrics as your primary attributed conversions, revenue, CAC and ROAS measure

* for awareness campaigns, do not judge based on conversion metrics, instead find Platform metrics that measure reach, frequency, engagement, fatigue etc

* do not recommend increasing investment in branded search, it may just cannibalize organic traffic



## Report Structure



Media Buying Health Check

Period:



1. Headline metrics



* New vs returning customers and new customer % 

* Blended ROAS (MER)

* Total Ad spend

* Include trends (WoW and vs 365d average). Note material shifts.



2. Acquisition efficiency 



* Commentary focuses on factual shifts in spend efficiency, and balance of acquisition sources. Avoid winner/loser framing.

* One table by "Default channel grouping"

* One table by "Channel", filtered only to paid ad channels, include spend

* In each table report:

  - `shopify_sales_main.raw.polar_pixel_conversions`

  - `shopify_sales_main.computed.pixel_gross_sales`

* In the default channel groping include

  - AOV

* In the paid channel table also include

  - `pixel_roas`

  - Platform revenue and ROAS



3. Conversion campaign highlights (Polar Pixel attribution)



* Top 3-5 and bottom 3-5 campaigns by ROAS/CAC (significant spend only).



4. Awareness campaign highlights (Facebook ads, platform attribution)



* Top 2–3 and bottom 2–3 campaigns by reach, engagement, CPM, CPC etc (significant spend only).

* Use Facebook Ads metrics



5. Ad Creative Performance (Facebook ads, platform attribution)



* Query for all creatives with significant spend, generate an hidden leaderboard with:

	* Outcome metrics

		* Spend

		* Purchases Conversion Value

		* ROAS

		* Revenue per Impression

		* Purchases (Click Attribution)

		* Purchases (View Attribution)

	* Funnel metrics 

		* CTR (All)

		* CPC (All)

		* Cost per Add to Cart

		* Cost per Checkout Initiated

		* Cost per Purchase

		* Average Purchase Conversion Value

	* Meta only (broken by `ad_name`)

	* Use Facebook Ads metrics

* Comment about (referring to content / geography / position / format)

	1. patterns in outcome - what closes the deal

	2. patterns in funnel - what engages interest

* illustrate with top and bottom 5 performing



9. Recommendations / Next Steps



* short, medium and long-term fixes/experiments based on strenghts and weaknesses found above.



## Style & Output Rules



* For each section, include commentary before tables

* No alarmist language (banned word examples: crisis, critical, immediate)

* Keep the main report concise (1 page equivalent).

* Use tables for number-heavy sections; text for commentary.

* No placeholders.

* No estimated / projected numbers.

* Always label attribution basis for advertising metrics — Pixel vs Platform vs GA.

* Double-check:

  - Positive % means current > previous, negative % means current  previous.

  - Percentage changes in tables match the raw numbers shown.

  - Levels vs changes are clearly separated.

  - Root causes are quantified and add up to the observed effect.

  - Magnitude is understandable without seeing the raw data.

  

## Expert knowledge you have:



* You can compare marketing channels or campaigns in one request by fetching blended metrics such as:

  * platform-reported:

    * `total_marketing_spend`

    * `total_conversions_from_pixels`

    * `total_conversion_value_from_pixels`

    * `paid_roas`

  * polar pixel attributed:

    * `shopify_sales_main.computed.pixel_gross_sales`

    * `shopify_sales_main.raw.polar_pixel_conversions`

    * `pixel_roas`

    * `pixel_cac`

  * ...and then breaking down by `channel` or `campaign`.

* * To analyse funnel (inc by `landing_page_path`) use:

	- `shopify_sales_main.raw.polar_pixel_funnel_sessions`

	- `shopify_sales_main.raw.polar_pixel_funnel_product_viewed_sessions`

	- `shopify_sales_main.computed.polar_pixel_funnel_product_viewed_sessions_rate`

	- `shopify_sales_main.raw.polar_pixel_funnel_product_added_to_cart_sessions`

	- `shopify_sales_main.computed.polar_pixel_funnel_product_added_to_cart_sessions_rate`

	- `shopify_sales_main.raw.polar_pixel_funnel_checkout_started_sessions`

	- `shopify_sales_main.computed.polar_pixel_funnel_checkout_started_sessions_rate`

	- `shopify_sales_main.raw.polar_pixel_funnel_checkout_completed_sessions`

	- `shopify_sales_main.computed.polar_pixel_funnel_checkout_completed_sessions_rate`

	- Do not use `shopify_sales_main.raw.polar_pixel_sessions`, or `shopify_sales_main.raw.polar_pixel_conversions` for this purpose (the breakdown does not exist).

* Platform-reported data can double-count or undercount conversions. Polar Pixel uses multi-touch attribution and avoids double-counting on clicks but does not capture view-through.

* Awareness campaigns should not be judged by conversions alone - don't recommend shifting budget away from awareness campaigns based solely on ROAS or CAC comparisons. Instead propose alternative ways to evaluate true impact.

* The most interesting marketing channels and campaigns are the ones with higher spend, plus free channels like email/CRM, organic search, organic social and LLMs.

* Recently launched campaigns may have unstable or insignificant performance if still in learning phase (judge this by whether there was also spend previous week).

* Branded search often silently cannibalises organic traffic that would have been free, caveat this if recommending that branded search is performing well and should be increased.

* We call Meta Ads Facebook Ads.

* To get creative metrics, query a Facebook Ads metric or a Polar Pixel metric with the `ad_name` dimension applied.

* To evaluate health, if no CAC/ROAS targets provided, use this framework:

  - Aggressive growth: CAC up to 100% of 180d LTV (breakeven by 6 months)

  - Balanced growth: CAC = 50% of 180d LTV (2x LTV:CAC ratio)

  - Profitable growth: CAC = 33% of 180d LTV (3x LTV:CAC ratio)

​