Marketing measurement and seasonality collide because pricing moves revenue more than advertising, making promotions look like ad wins. This matters because a model that doesn’t flag promo days credits media for revenue a discount drove. Without experiment timing around high-demand periods, fluctuating spend during a holiday makes advertising look more effective than it is. This video explains isolating promotions and seasonality so a model reflects true advertising impact.
Why Marketing Measurement and Seasonality Require Separating Price From Advertising
Marketing measurement and seasonality require separating price from advertising, because pricing changes can move revenue by an order of magnitude more than advertising ever will, and a model that doesn’t isolate that will hand advertising credit it didn’t earn. In this video, Jon Lorenzini, Chief Marketing Science Officer at LiftLab, cites research from Dr. Dominique Hanssens showing that price changes can be roughly 20 times more impactful on revenue than advertising changes, which means a 50%-off promotion, not a sudden leap in ad performance, is almost always what’s actually driving a spike.
Lorenzini describes flagging as one practical fix: marking promotional (or, for some clients, the rarer non-promotional) days so investment decisions about media aren’t accidentally based on P&L swings that had nothing to do with advertising. Seasonality raises a related problem for experiment design specifically. Fluctuating budgets around high-demand periods like Mother’s Day, intentionally or not, can make advertising look far more effective than it actually is if a spend increase happens to land during a natural demand spike. Careful timing and balance in when highs and lows are tested is what protects the resulting diminishing returns curve from that distortion. Lorenzini also warns that overusing promotions carries its own cost beyond measurement accuracy: cannibalizing sales that would have happened anyway erodes brand equity over time, a pattern he’s seen recur enough across clients that many of these safeguards are now built directly into LiftLab’s product, informed in part by the company’s academic advisory board.
In this video, you’ll learn:
How to decouple promotional impacts from advertising performance
The importance of flagging non-promotional and promotional days for modeling
Why seasonality can skew results in marketing measurement models
Best practices for designing experiments that account for market demand
The long-term risks of over-relying on promotions for short-term revenue
TIME STAMPFull Video Transcript: Marketing Measurement, Promotions, and Seasonality Explained
+–
LiftLab’s Jon Lorenzini explains how to keep promotions and seasonality from distorting marketing measurement models.
[0:00 – 0:53] The Challenge of Promotions and Seasonality
A common problem with advertisers, across everyone in the business, is promotions and seasonality. I think the challenge that we have with that is, how do you capture the impact of pricing, promotional changes, or seasonal demand changes in a way that decouples that with your advertising. If you’re running a promotion, and you know, according to Dr. Hanssens’ research, the changes in price are 20 times more impactful than the changes in advertising. So if you’re doing a 20x thing, your advertising doesn’t suddenly get 20 times better on that one day; it’s the promotion, 50% off, 20% off, what have you. So how do we capture that appropriately, and then how do we handle that both in our experimentation stack and our modeling stack, so as to not mess up what is the impact of advertising, because really what you’re measuring is the impact of promotions in your topline revenue.
[0:53 – 1:36] Strategies for Handling Promotional Data
There’s a couple different ways of handling it. I’m going to speak a little bit more generally about it, but one thing is you can flag the day out. Say, hey, this is a day that’s not really reliably estimating the impacts of advertising overall on average, and investment decisions should be made off of that. Your investment decisions might be made off of your P&L, your promotional things, but not your media piece. So one solution is flagging. We’ve had a bunch of different clients with different promotional schedules; some promote every day, and then there’s a couple days that they don’t promote. So it’s almost like the days they don’t promote are the ones that need to be flagged or highlighted, because it’s a negative promotion day. So there is a lot of nuance with how it’s happening.
[1:36 – 2:40] Implications of Seasonality and Experiment Design
And away from the promotional piece and more into the seasonality piece, there’s implications when you’re running experiments because of this. And this is something that, you know, with Mother’s Day coming up and a couple other holidays, there is a challenge to say, if we’re fluctuating budgets, if you happen to fluctuate during a high demand time, or if you happen to fluctuate low during a high demand time, that will impact what comes out of the model. And that’s where we have to make sure we’re very careful in choosing both the time frame and then the balance of when we’re doing those highs and lows, to be able to get that diminishing returns curve. Because if you are pushing hard when there is high demand and pushing low when there’s low demand, it’s going to show that advertising works a lot better than it does. So the experiment design, the handling of the promotional events, handling of the seasonality, and maybe testing at different points in different seasonal places, all give you information that allows you to make a more robust model, and possibly enable you to get closer to what is the impact of just advertising.
[2:40 – 4:05] Avoiding Promotional Cannibalization and Expert Oversight
And that’s really what we want to measure, because a promotional thing, you know, I saw this when I was agency side, we ran so many promotions we cannibalized our existing sales. They were off from their forecast by a large magnitude, and all the assumptions I put in there were correct except for the average order value. So when you’re running a promotion, it’s almost like you can overdose on promotions, because there’s a given number of people in market at a time, there’s a given percentage of that who would buy your product anyway, and by discounting that, you’re making less revenue for yourself. So we see those kinds of instances come up for people who want short-term revenue and high-impact immediate results advertising, but it really does erode your brand equity over a large period of time. So these are the kind of pieces and levers that we consult with our clients on, and that everyone in the industry is kind of talking about as common problems around that seasonality and promotion consideration. And it’s really something that, I’m very lucky, we have a whole advisory board of professors and experts in the field that allow me to bounce ideas off of and make sure we refine things. But after seeing so many different problems across different clients, a lot of those have now been baked into the product, in a way that is like having an expert or an adviser for you but is kind of by default in code.
Key Lessons: Isolating Promotions and Seasonality in Marketing Measurement
Modeling nuances: Flagging specific promotional periods ensures your advertising model is not inaccurately attributing price-driven sales to media spend.
Seasonality awareness: Fluctuating advertising budgets during high-demand holidays like Mother’s Day requires precise tracking to avoid misleading diminishing returns curves.
Strategic restraint: Excessive promotional activity often leads to cannibalization, as it discounts sales that would have occurred without the incentive, ultimately eroding long-term brand equity.
