The model measures not just what each channel sells today, but how much it moves your brand and what that brand is worth in sales. It takes your digital and offline media spend, your brand metric and your sales, and tells you how much to invest in each channel.
MEDIA → BRAND → BUSINESS · WITH ITS UNCERTAINTY RANGE
The model shows which channel moves sales and how much it pays to invest in each one. (Video in Spanish.)
A traditional Marketing Mix Model answers how much each dollar in each channel sold. That works for performance, but falls short on brand: brand advertising first changes what people think of your company and, months later, that turns into sales. The model measures that whole chain and answers the two questions a traditional model can’t.
Spend by channel, week by week. Each channel has a role: brand —TV, out-of-home, sponsorships, video— or performance —search, paid social, affiliates—.
The metric you already track —top of mind, first choice, consideration, NPS or branded search— estimated every week, even if the survey is monthly.
The outcome you choose: sales, volume, new customers or orders.
Performance channels hit the business directly; brand channels reach the business through the brand. That separation is what lets you put a sales value on every point of your brand metric.
An ad doesn’t stop working when it goes off air: it fades little by little, like a perfume. Each channel at its own pace: search in days, TV in weeks or months.
Like sugar in coffee: the first spoonfuls change a lot, the last ones barely anything. Every extra dollar in a channel returns less than the one before, and every channel has a ceiling.
In summer people travel more and also think more about airlines. If the model can’t see the calendar, price and capacity, it credits the brand for what summer did. The model separates them.
A brand lift study is like weighing something whose weight you already know: it doesn’t replace the scale, but it corrects it. The model uses those studies to anchor its estimates.
On top of those four ideas, the model learns from your history —spend, brand and sales—, estimates everything with its uncertainty, and lets you simulate scenarios or find the budget that reaches a goal.
How much your spend moves a brand metric and what that lift is worth in sales, on top of what each channel sells right now.
Paid media, SEO, social listening, content, email and automation, web, CRM, competitors, territory and distributor network. Digital and offline, by the same yardstick.
Each channel’s contribution to brand and to sales, and its marginal return: how much every extra dollar you put in adds.
How long each channel’s effect lasts and where it stops paying off: the model shows where each one sits on its curve today.
Seasonality, price, capacity and context go into the model, and it can be calibrated with your brand lift studies.
Tell it the brand goal and the model finds the spend that gets there, with its range. If that mix can’t reach it, it says so.
The technical side: each channel’s memory, saturation, external factors and how it is calibrated with brand lift studies. (Video in Spanish.)
Your brand metric estimated every week —even if the survey is monthly— with its uncertainty band.
A breakdown of your revenue: how much comes from the baseline, the brand, performance and external factors.
Each channel’s saturation curve and where it sits today: where every dollar still pays off and where it no longer does.
How much additional revenue every extra dollar generates in each channel, over a long horizon.
Shift spend between channels and see the estimated impact on brand and sales before you decide.
Results in the language of the board and finance, not marketing jargon.
Several brands or units under one company, each with its own view. If your company has a single brand, you don’t see a selector.
One model for the whole country or split by region. Variation between areas gives the model more information to separate causes, without paying for a bigger survey.
New data is added as it comes in and the model is re-estimated periodically, at the pace your team sets.
Results come as a range, with their uncertainty, never as a magic number.
It’s tested on a simulated market whose true values are known: the model has to find them without being told.
With your data, the model predicts weeks it hasn’t seen, to prove it didn’t overfit.
Your brand lift studies go in as evidence: the model can’t stray from them without a strong reason.
It reliably identifies the best and worst channel. If two channels always run together, it tells you an experiment or regional variation is needed to separate them.
It works on revenue, not margin, and doesn’t count what happens after the horizon: if it’s wrong, it errs low.
| DATA | HOW OFTEN | WHAT IT INCLUDES |
|---|---|---|
| Spend | Weekly, by channel (and region) | Date, channel, region and spend. Ideally impressions or GRPs too. |
| Brand metric | Per survey wave (and region) | Fieldwork dates, sample size and result. Another measurable brand metric, like branded search, also works. |
| Business | Weekly (and by region) | The outcome you choose: sales, volume, new customers or orders. |
| External factors | Weekly | Relative price, capacity, weather, exchange rate, holidays: whatever moves your industry. |
| Brand lift studies | Per study (optional) | Channel, study weeks, measured result and its margin of error. |
On the way in, the model checks the data: missing weeks, channels with no spend, survey waves without fieldwork dates.
Every module can be customized, and its features can be used in part or in full: none of them is mandatory. You can use BOS just to plan tasks and never touch the rest — it stays available, included in the same subscription, for the day you need it.