How an amusement park brand partnered with MASS Analytics to separate marketing seasonality from media impact, then used the result to reshape its seasonal media plan.
- At a glanceSector
- Entertainment, amusement park
- Scope
- One amusement park brand, seasonal event sales and media across regions, four years of data
- Engagement
- Ongoing
Always-ON measurement, delivered on the MassTer platform.
The challenge
For an amusement park, event sales fall into two months of the year. Outside that window sales are very low, and for one full year the product could not be sold at all. Marketing seasonality of this kind makes media impact hard to read. Seasonal campaigns run only while the event sells, so media and the season rise and fall together. This case study describes how MASS Analytics helped a leading entertainment advertiser pull the two apart, so it could decide which channels deserve budget during the event season.
Media and seasonality moved together
The seasonal campaigns made up less than 15% of total media spend and ran only during the sales window. Because they started and stopped with the event, their execution pattern was almost identical across channels. That created high multicollinearity: a standard model could not tell whether a sale came from the media or from the time of year. The modeled period covered four years of data, one of them a dark year with no sales.
“The campaigns ran only when the event sold, so media and the calendar moved together. The model had to pull them apart.”
A member of the brand’s marketing team
Four things the brand could not see
- The effect of the seasonal campaigns. They ran only in the sales window and held a small share of spend, so their own effect was hidden behind the effect of the season.
- The split of the baseline. Sales sat at a very low level outside the event. No one could say how much of the baseline was seasonal and how much was not.
- The synergy between media and the season. The brand wanted to know whether media and seasonality reinforced each other, and by how much.
- The right channels for the event. The goal was a media plan that put the right channels against the event for the highest ROI. That needed the three answers above.
Each question needed its own treatment in the model. Table 1 summarizes how the rebuild handled them.
| Challenge | Modeling Approach | What It Separates | Finding |
|---|---|---|---|
| Seasonality | Repeating calendar windows that mirror the yearly event | The effect of the season from the effect of media | Sales followed the season closely, so the calendar had to enter the model first |
| Multicollinearity | Weighted aggregation of variables with similar execution | Overlapping effects of near-identical campaigns | Results readable at the grouped level or variable by variable |
| Synergy | Log-linear form, with media and seasonality interacting | The joint effect of media and the season | Online media gained the most from the synergy with the season |
Table 1:How each challenge was handled in the model. Variables sat at sub-channel level, down to creative type.
The solution
MASS Analytics built a pooled regression across regions, with variables at sub-channel and creative-type level. Pooling gave more room to test a large and varied media mix. Two sets of rules guided the search for the best model: statistical constraints to secure significance, and business constraints to keep results close to expectations on spend levels and click-through rates.
Seasonality entered as repeating windows
Sales followed the seasonal calendar closely, so the first step was to reflect the season in the model. The team created repeating windows that mirrored the yearly run of the event and used them as explicit variables. This kept the effect of the calendar apart from the effect of media, and it made it possible to separate the seasonal part of sales from the rest.
Grouped variables handled the overlap
Seasonal campaigns shared almost the same execution pattern, so many media variables looked alike. The team grouped variables with similar execution that belonged to similar media channels into weighted aggregates, which reduced multicollinearity. Results could then be read at the grouped level or variable by variable. A log-linear form let the model measure the synergy between media and seasonality by default.
Results and Impact
The model showed a strong effect of seasonal media on event sales, with an average marketing ROI of 3 over the final modeled period. Sales rose by nearly 50% while seasonal media spend rose by about 60%.
3x
average marketing ROI of seasonal media
~50%
event sales growth on the previous period
~60%
more seasonal media spend than the previous period
<15%
of total media spend held by seasonal media
Seasonal media returned an average ROI of 3
Historical results showed a considerably high impact of seasonal media campaigns on event sales. The average marketing ROI was 3 over the final period of the modeling window. Online media benefited most from the synergy with the season. Execution is concentrated around the two months of the event, and running media in that window boosted consideration further. The model measured this synergy directly, which separated three effects: the season on its own, media on its own, and the extra effect of media scheduled through the hot season.
Sales rose nearly 50% as seasonal spend grew about 60%
Seasonal media spend rose by about 60% on the previous period, and sales rose by nearly 50%. The rise came with a change in media strategy: new sub-channels were added, and more budget went to key channels such as Facebook, Instagram and Google Display. The findings then fed marketing spend optimization for the next season. Online media were timed closer to the event, while offline channels were rescheduled to launch first, ahead of the event, to build reach and awareness.
Want to see what your season is worth?
MASS Analytics builds MMM models that turn seasonal patterns into marketing optimization decisions, and refresh as often as the business changes. Talk to our team.
Book a Demo
