Does media mix modeling work for a health and beauty store?

Media mix modeling can work for a health and beauty store if you have two or more years of weekly sales and spend that moved. You also need a log of promotions, launches and creator posts. Without it, the model mixes up discounts, gifting peaks and ads, and splits the credit by guesswork.

If you sell health and beauty products

If you sell skincare, haircare, cosmetics or supplements, much of your revenue may come from people buying the same product again when it runs out. A media mix model will likely put most of those repeat sales in its baseline. That is what you would sell with paid and organic media switched off. That is fair to loyal buyers, and a little unfair to the ad that found them last year.

Your calendar is loud as well. Gift sets, launches, sample drops and discount events all move sales in bursts. If ads and discounts always run in the same weeks, no model can pull them apart.

Much of your marketing may also cost nothing per post: creators, reviews and community content. A model only counts what you give it. Leave those posts out and their sales get credited to whatever else ran that week.

If you also sell through retailers or marketplaces, decide early which sales figure the model explains. Meridian needs a KPI you can add up across weeks and regions, so pick one figure and keep it.

What one store's data shows

One store's anonymised GA4 export, 1 January 2024 to 21 August 2026. It holds shares of revenue only: no ad spend, no order counts.

The export does not say what Store A sells, so treat it as one store, not a stand-in for beauty brands. One-touch journeys carry 79.5% of revenue on the Journeys sheet, bought within 0.5 days. Journeys of ten or more touches carry 3.0% on the same Journeys sheet and took 16.0 days. On the Channels sheet, Direct holds 57.7% in every view.

If your own export looks like this, a weekly model sees most sales close in the week of the touch, with a thin tail later. For a replenishment brand, the bigger question is Direct. Much of it may be repeat buyers coming back on their own, and an MMM would file that under the baseline. Ask how much of your Direct is repeat buying before you trust any channel's share.

What changes for a health and beauty store?

• Promotions are inputs, not noise. Log every promotion with dates and depth. Meridian's docs class running a promotion and the price of a product as non-media treatments, separate from ads.
• Creator posts count as organic media. Robyn's docs list social media posts and reach on user-generated content as typical organic variables. Meridian measures organic media by impressions or clicks.
• Gifting peaks need a calendar. Robyn's guide notes that ecommerce revenue can move days before a holiday, because orders must arrive in time. Add your own dates, such as Black Friday, to the event list.
• Ads and discounts must not always move together. Robyn's guide warns that without variation the model cannot tell what an input did. Run some ad weeks without a promotion, and some promotions without extra spend.
• Launches are events too. A new hero product lifts sales on its own. Flag launch weeks, so the model does not hand that lift to whichever channel spent most.

What to do this week

1. Find your promotion weeks in Shopify. Open the Total sales over time report from the Sales filter under Analytics > Reports. Group it by week and compare it with the previous year. Pass: every spike matches a promotion, launch or holiday you can name. Fail: spikes nobody can explain, so build that log before any model.
2. Check that creator traffic is labelled. In GA4, open Reports > Acquisition > Traffic acquisition, set the table to Session source / medium and search for a creator's handle. Pass: each creator you work with shows as its own source. Fail: nothing shows, so their visits sit in Organic Social or Direct; give each creator a tagged link before the next drop.
3. Check whether ads and discounts moved together. In Google Ads, segment the Campaigns table by Time, then Week. Pass: spend also rises in some weeks without a promotion. Fail: spend only jumps in promo weeks, so a model cannot separate the two yet.

Check the homework. Your GA4 Attribution paths export already holds the evidence. Causality Engine reads that one file and shows what each channel caused next to what last-click gave it, in 1 to 2 minutes, for €99 once (excluding VAT), refundable within 30 days. Check the homework

Sources, 1 October 2026: Collect and organize your data (Google). Assess the baseline (Google). An Analyst's Guide to MMM (Meta). Key Features (Meta). Sales reports (Shopify). Traffic acquisition report (Google). Use segments in your tables (Google).

Related answers

• All MMM & measurement questions
• What is media mix modeling?
• How to run your first media mix model, step by step
• Marketing mix modeling