P1 Commerce - Digital Marketing Firm ads

If your agency’s comp is tied to spend, I want to see what happens when the model says the next dollar is not incremental: do they still recommend spending it? Unless there’s a real system holding spend to marginal impact and incrementality, there is a huge incentive misalignment here. Here’s how we approach this at P1 to solve for this: On the management side, we are incentivized differently because we are held to the marginal impact of that model. We are also running incrementality tests to prove the marginal impact, so we are responsible for both the model and the testing to prove the incremental result to the business. And yes, sometimes, we are buying media, but if the model says the next dollar has a better marginal result somewhere else… the dollars do not always go our way. It means we will make less money in the short term, but we are happy to move them because we are held to the standard of the model and incrementality testing. That is what incentivizes us not to grade our own homework. It also means that we earn our clients’ trust when they see us taking money out of our pocket because it doesn’t make sense for them – and I think that’s why we have such long-term partnerships and referrals. So, if your agency cannot defend marginal impact (and cannot prove it with incrementality testing), you are not really working with a growth partner.

The worst way to find your saturation point is after you've already blown past it and ROI tanked. The era of not caring about your marginal dollar is over. Now that brands care about growing efficiently, you have two options: If you have a model, I recommend starting with the model's interpretation of the saturation point, then stress test up to that point, and watch the marginal ROI as we go. Example: a Meta remarketing campaign saturates quickly because the audience size is only so big. Other tactics can handle more expansion while holding marginal ROI up to a point. Our job is to scale to that point and verify the model's read. If you don't have a model, we personally use blended ROI as the stop gap. I recommend scaling one channel at a time and starting with something that has a more immediate effect. You're more likely to see it in 7–15 days, not 30–50. Then make the next move based on what blended ROI did. If blended ROI holds or improves after the increment, keep scaling that one lever until you see the next dollar stop yielding the same result. If it degrades, pause. Reset spend back to the last level that held, and test the next lever.

You can have the most accurate and robust MMM, but if the billboard vendor spends differently, the agency reallocates mid-month, and nobody tells you, the forecast will miss every time. We get hired to measure performance across the whole business, but we do not always get to execute every channel. So when results miss, the first question is always: did everyone actually run the plan? One of our goals as a measurement partner is to keep everyone accountable. When performance misses, we can go to leadership and quantify what happened: “Here was the plan. Here is what actually ran. Adherence was off by 30%. Forecast missed by X%." This is where senior management has to step in and work with every partner so they stick to the plan. That tightens things up fast, and it changes the culture inside a company – whether it's an internal buyer, a third-party agency, or a vendor – so adherence becomes non-negotiable.

Blended ROI is actually a great metric when it's business as usual. If spend is consistent and tactics are stable, it gives you a clean baseline. It only becomes a problem the moment you expand into new channels aggressively – especially higher funnel tactics with longer time shifts. Spend goes up today, revenue hasn't caught up yet, and blended ROI drops. This is why I don't use blended metrics alone to decide whether to turn off or scale up a channel. Instead, during heavy testing periods, we run incrementality tests in the same window so we can quantify lift while blended ROI is temporarily distorted by the spend ramp. A good rule of thumb I like here is to assume the incremental spend has no effect, and ask if you can live with the low end of blended ROI as the stop gap. If you can’t, you shouldn’t run the test.

Audience overlap is one of the biggest things that's not talked about enough in ecom paid media right now. Let’s say you have CTV and similar campaigns running in a conversion objective, fed off a pixel. So if you're an algorithm trying to drive revenue, what are you gonna do? You're gonna find the signal. What happens is it latches onto the signal of Meta, because Meta is more powerful than TV. Next thing you know, CTV is claiming credit for things that are actually driven by Meta. And audience overlap when you're trying to drive revenue can really back you into a corner. You end up with smaller audiences, higher intent, high frequency, low reach – and then you're just kind of circling around saying, we're spending here, but we're not actually getting more top-line revenue. That's because everything is saturated across all the tactics and visibility. The worst part is when brands see the top line flatline and their first instinct is to spend more. But if everything is saturated and overlapping, more budget just means more frequency on the same people. And now you're just paying more to reach the same audience through different doors. If your media structure doesn't account for where audiences overlap, you can't spend your way out of it. The model will tell you one thing, but the executional side is just as important – and that's the piece most teams skip over.

Sadly, brands get stuck bottom-funnel because it’s the only thing they can prove. When you go up-funnel, the results look “worse.” So you never do it. But what happens when you run out of signal? If you were a machine learning algorithm and you had to show your client revenue, you’d do the same thing: go find behavioral profiles that show intent and are likely to buy. That gets amplified when you have offline channels (catalogs, billboards) driving people to the site – Meta latches onto that and overachieves on conversion campaigns. Now, I understand that you need to prove the economics of your spend to your CMO/CFO. But long-term, limiting yourself to bottom funnel conversions just won’t work. Every time we run an MMM, we see that the higher funnel is a big driver of search intent. It creates the signal you need for the bottom of the funnel conversions to actually exist. So, please, don’t abandon that too soon. Go run an incrementality test – and maybe don’t even measure conversions at first. Measure traffic. Over time, you’ll see the real impact downstream, and you can then adjust spend based on marginal impact.