MASS Analytics built a media mix model for a D2C media company with a subscription service, checked it against an earlier attribution study, and found a modeled revenue gain of more than 11%.
- At a glanceSector
- Media, direct to consumer subscription
- Scope
- One D2C media company, eight media metrics, checked against an earlier attribution study
- Engagement
- Ongoing
Always-ON measurement, delivered on the MassTer platform.
The challenge
For a direct to consumer media company, subscriptions are the business. This client’s subscriptions were shifting. Growth that external influences had driven was fading, and the base the company had built was now carrying more of the weight. The team needed to know which media still earned its budget, which channels had room to grow, and where extra spend would only add cost. MASS Analytics built a media mix model to answer those questions, then tested it against a second method.
Part of the growth had an outside source
The model’s first finding was mixed. External influences had driven almost 30% of the client’s growth, and as they faded, the subscriptions they had brought in began to leave. The good news was the base: it made up about 60% of total revenue, which showed that earlier activity had built a strong foundation. Media activity claimed the other 40%, but at a low return. The question was whether to keep investing.
“The model brought good news and bad news. External influences had driven almost 30% of our growth, yet our base still made up about 60% of revenue.”
A member of the client’s marketing team
Four questions the client could not answer
- Which media still worked. Subscriptions were shifting, and the effect of each channel, direct or indirect, was unclear. Without it, any budget cut or increase was a guess.
- How much room each channel had. Return on ad spend (ROAS) and cost per acquisition (CPA) were known, but not the point where each channel stopped paying back.
- How channels worked together. Channels affect each other, and a channel that looks saturated alone can still pay back when it runs beside another one.
- Whether the first model was right. One method gives one view. Without a second method to check against, hidden effects, such as influencers lifting branded search, stay hidden.
The table below shows what the model found for the three groups of channels that mattered most.
| Channel group | How it was modeled | What it measures | Finding |
|---|---|---|---|
| Twitter, Facebook, Instagram | Response curves with synergy effects between channels | How much extra revenue more spend could still bring | Not yet saturated; moderate investment recommended |
| YouTube | Log linear model with synergy effects | Saturation, and its links to other channels | Saturation looming, but moderate spend still pays when run beside channels with positive synergy |
| Influencers and branded search | Indirect effect in a nested model | Influence that reaches subscriptions through branded search | Missing in the first model; found through the check against the earlier study |
Table 1: Each channel group entered the model with the effects that mattered for it. Findings are modeled results.
The solution
MASS Analytics built the model at national level, using about three years of data and eight media metrics: clicks for branded paid search on Google and Bing, impressions for Google generic paid search, Google display, Instagram, Facebook and Twitter, and views for YouTube. The structure was a log linear regression, which shows how variables interact, combined with a nested model that separates direct, indirect and combined effects of each channel on subscriptions.
Five transformations shaped the media mix model
- Adstock. How long an ad keeps working in memory after it runs.
- Diminishing returns. The point where extra spend adds less, used to model saturation.
- Weighted sum. The different impact channels have when they run at the same time.
- Split. Separate periods, to measure the effect of a change in strategy or ad content.
- Lag. A delayed effect. After a free days promotion, subscriptions rose for a few weeks and then dipped, so the model carried the promotion and a lagged version of it.
Multi touch attribution vs marketing mix modeling: what each saw
Multi touch attribution vs marketing mix modeling is usually framed as a choice. Here, the client used both. The client had an earlier multi touch attribution (MTA) study and agreed to share its results. MMM uses aggregate data to measure online and offline media, plus non-marketing and external factors. MTA uses granular online data. Benchmarking the model against the MTA study showed an indirect effect the first model had missed: influencers lifting branded search. The model was still in calibration, so the team used this integrated measurement to refine it. The result described reality more precisely and predicted more accurately.
Results and Impact
The cross-checked media mix model gave the client a plan it could act on, with a modeled revenue gain of more than 11%. It also showed where the business stood: the base made up about 60% of revenue, and media about 40%.
11%+
modeled revenue gain from an optimized media budget
~60%
of revenue came from the base business
~40%
of revenue came from media activity
~30%
of growth came from external influences
The media mix model found room to grow
The model’s response curve for Twitter showed current revenue sitting far from the point where the curve flattens, so more spend could still raise revenue. Because channels interact, and because the client’s commercial constraints were part of the model, the advice was to invest moderately, not heavily. The same applied to Facebook and Instagram. For the wider budget, the advice was to keep investing in media: it still had to make up for the growth that was fading, and its channels still showed room to improve their return. This is media mix optimization in practice: moderate steps, guided by response curves, not one large shift.
Synergy and ad format kept YouTube worth funding
YouTube showed saturation, which at first suggested cutting spend. The log linear model showed that moderate spend still made sense when it ran beside channels with positive synergy, such as Google generic paid search. Ad format mattered too: a study by Nielsen and Google found that ad format mix can lead to up to twice the ROI on YouTube, and the client used a single format. The advice was to rotate formats and show ads on more devices, to lower cost per thousand impressions (CPM) and raise ROI. Together, these recommendations supported the modeled revenue gain of more than 11%.
From one model to an Always-ON view
A model like this does not have to stop at one build. When it is refreshed on a regular cycle with new data, the same media mix model becomes an Always-ON view: each refresh shows which media still earns its budget, so the plan moves as the market does.
Want to see how to capitalize on change?
MASS Analytics builds models that show which media earns its budget, then checks them against a second method. Talk to our team.
Book a Demo Read the MMM Guide
