Fashion marketing is fast, promotional, and trend-driven. MMM is the measurement tool that accounts for all three factors at once, so you know whether your campaign drove sales or whether the markdown did.
Fashion and apparel brands face a measurement problem that combines the worst aspects of ecommerce, seasonal businesses, and FMCG all at once. They sell across owned websites, department stores, third-party ecommerce platforms, and physical stores. They run two to four seasonal collections per year, each with its own media campaign and markdown cycle. And they operate in a category where trend cycles, influencer moments, and cultural events can spike demand overnight in ways that have nothing to do with paid media. Standard digital attribution makes little sense in this environment.
The fashion measurement problem
Fashion brands run a predictable cycle: launch campaign at full price, sustain with retargeting and social, mark down to clear end-of-season stock, then go dark before the next collection. Digital attribution inflates the effectiveness of retargeting and paid search during the markdown phase (when conversions spike because prices are cut) and undervalues the brand campaign that launched the season. The result is that brand-building investment looks expensive and promotional digital looks efficient, which pushes brands toward more promotion and less brand over time, damaging long-term brand equity.
High return rates add another layer of complexity. Fashion has some of the highest return rates in ecommerce, often 25 to 40 percent for online orders. A campaign that drives a lot of gross orders may be delivering far less net revenue than it appears. Modelling on gross revenue systematically overvalues channels that attract browsers who order multiple sizes and return most of them.
The fashion and apparel media mix
- Paid social (Meta, TikTok, Pinterest): dominant channel for most fashion brands, strong for product discovery and conversion
- Influencer marketing: high spend, particularly for emerging and luxury brands, creates measurable spikes in traffic and sales
- Paid search and shopping: captures intent at the bottom of the funnel, often benefits from brand campaigns running earlier
- Digital out-of-home: increasingly used by fashion brands for brand positioning in city centres and transport hubs
- Print (fashion press, magazines): still significant for luxury and premium fashion brands targeting a specific reader demographic
- TV and connected TV: used by high street and department store fashion brands for key seasonal moments
- PR and editorial: major driver of cultural relevance, difficult to measure but creates measurable traffic spikes
- Affiliate networks: high volume, particularly during sale periods, with significant attribution overlap
Seasonality and collection cycles
Fashion operates on a retail calendar that includes spring/summer and autumn/winter collection launches, Black Friday and Cyber Monday, the Christmas gifting period, and January and mid-season sales. Each of these events creates a distinct demand and promotional pattern. The model needs a detailed promotional calendar that captures not just which channels were active, but what the promotional mechanic was (10 percent off, free returns, specific product launches) and what the average selling price was during that period.
Unlike FMCG, where seasonality is driven primarily by weather and consumption habits, fashion seasonality is partly driven by the industry's own publishing calendar. Editorial coverage in Vogue, Elle, and online fashion media peaks at collection launch time and drives significant organic traffic. A model without a proxy for editorial coverage will attribute that organic demand spike to paid media, inflating the effectiveness of channels that happened to be running during the launch window.
Model on net revenue after returns, not gross revenue. For fashion brands with 30 percent or higher return rates, the difference is material. A channel that drives high-return customers looks efficient on a gross ROAS basis but may actually be loss-making on a net basis.
Influencer marketing in the model
Influencer spend is often the second or third largest media investment for fashion brands, but it is frequently excluded from MMM because the data is messy. The spend is irregular, post timing varies from the contracted date, and the reach metric differs by creator. Despite these challenges, influencer activity should be included. Build a weekly influencer activity variable that captures the cumulative reach of all influencer posts in that week. Even an imperfect proxy prevents the model from attributing influencer-driven sales spikes to paid search or retargeting.
Price and markdown modelling
Fashion MMM needs a price variable that captures the average selling price or average discount depth in each week. Without it, the model cannot separate the effect of a price cut from the effect of media. If you run your biggest paid social campaign at the same time as your biggest markdown, and you do not include the discount depth as a variable, the model will assign the sales spike entirely to paid social. This is the most common source of inflated media effectiveness estimates in fashion MMM.
Using MMM to improve brand investment decisions
Fashion brands routinely cut brand-building spend (awareness campaigns, editorial, PR investment) under financial pressure, because it looks expensive on a short-term cost-per-sale basis. MMM with properly specified adstock shows the delayed effect of brand investment on full-price sell-through rates and average order values. Brands that have run this analysis consistently find that collection launch campaigns drive not just immediate sales but a sustained lift in full-price demand over the following four to six weeks. Cutting the launch campaign saves media cost but reduces total season revenue.
We sell across our own website, ASOS, and in department stores. Can MMM capture all of those channels?
Yes, provided you have weekly sales data by channel. Model total brand sales (own website plus third-party) as the dependent variable, and include a distribution or availability variable that captures how many third-party retail points are active in any given week. If ASOS adds your brand to a new category or features you in a campaign, that should be flagged as an event variable so the model does not attribute the resulting sales spike to your own paid media.
We have a significant vintage and resale market for our brand. Does that affect the model?
Resale activity affects your brand perception and can even cannibalise new product sales in some categories. You should not include resale revenue in your outcome metric (model only your own direct sales), but you may want to include a resale activity indicator if you have data on it. For most fashion brands, resale is not yet large enough to materially distort a model, but it is worth monitoring.
We are a small independent fashion brand spending around 150,000 pounds a year on paid media. Is MMM appropriate for us?
At that budget level, MMM may produce results with wide uncertainty ranges, particularly if your channel mix has not changed much over the past two years. Consider using controlled geo experiments or a simple channel switching test (pause one channel for four to six weeks and measure the impact on total sales) alongside or instead of MMM. As your budget grows above 300,000 to 400,000 pounds, a full MMM becomes more reliable and more cost-effective relative to the insight it provides.