Is Low CPP Killing Your Growth? The Efficiency Paradox in Media Planning
There is a version of media optimisation that looks excellent on a plan and quietly hollows out a brand. It does not happen through bad decisions. It happens through a metric doing exactly what it is designed to do, and no one asking whether that metric is pointing in the right direction.
That metric is CPP. And the distortions it creates run deeper than most media reviews ever reveal.
What CPP actually optimises for
Cost per point is a reach efficiency measure. It tells you how much it costs to reach one percent of a target audience. Lower CPP means more reach per budget spent. That sounds like a good thing.
But CPP is not audience-neutral. Older demographics are consistently cheaper to reach than younger ones. There is less competition, lower inventory costs, and more available supply. A plan that optimises hard on CPP will therefore drift toward older audiences. This does not happen because anyone explicitly decided it. It happens because the metric rewards it.
This is the first layer of the trap. In a skewed environment, efficiency and audience composition are not independent variables. When one improves, the other can quietly degrade. As budgets migrate toward cheaper inventory, they also migrate toward audiences that are less relevant for future growth. More reach, but not necessarily more of the right reach.
Where it gets structural: the MMM feedback loop
This part rarely appears in efficiency discussions, but it should, because it turns a planning bias into a self-reinforcing system.
Marketing Mix Models are built on what happened. They operate on historical spend, coverage and channel allocation, and the sales and brand outcomes that followed. That history reflects the execution framework, including the effect of CPP optimisation.
If plans have been drifting toward cheaper, older audiences, the internal models have been built on that reality. They reflect a media mix that is already skewed.
The consequence is that future recommendations can inherit this bias. Channels that were under-invested because they were expensive appear weaker in the model. This is not necessarily because they are ineffective, but because they were never given the conditions to perform. Channels that received budget because they were cheap appear stronger, regardless of the quality of demand they created.
The model therefore reinforces the same direction as the CPP metric. The plan optimises further along that path. The next model cycle is built on it. The loop closes.
Where OOH gets lost
Outdoor is a specific casualty of this dynamic, and for a structural reason that might have nothing to do with effectiveness. OOH is rarely deployed nationally. When a brand runs it in key urban markets but measures performance at a national level, the model averages the effect across regions where no activity ran. The channel looks weak. The investment contracts. But what the model read as underperformance was simply absence, and the decision to pull back creates exactly the condition it was measuring.
This is not a modelling error in the technical sense. It is what happens when the measurement architecture does not match the media strategy. The model is doing its job. The result is still incomplete.
What this means at the commercial level
A brand that is growing younger audiences is building future demand. A brand that is drifting older is harvesting existing demand.
Both can look similar in the short term. Sales hold, CPP and efficiency improve, and model outputs appear healthy. The divergence appears later, and often abruptly. Category share among younger audiences weakens. Brand consideration declines in the audiences that matter for future growth. Pricing power erodes as the brand loses relevance.
By the time these signals appear, the internal models have already absorbed the degraded mix as the baseline. Reversing course requires not just a change in strategy, but a willingness to question what the models are recommending.
Breaking the loop
The starting point is to separate efficiency measurement from long term growth effectiveness. CPP should be one input, not the governing metric.
Plans should explicitly track reach among future buyer audiences. This should not sit as a secondary demographic view, but as a primary objective with its own budget logic.
On the modelling side, model outputs need to be pressure-tested for structural bias. The key question is whether historical channel performance reflects true effectiveness or constrained investment, and whether the aggregate buyer group is skewed. This requires a willingness to invest in environments that the models may currently undervalue, supported by incrementality testing designed to generate new signal rather than confirm existing patterns.
The CPP Trap is not inevitable. Escaping it starts with recognising that the metric optimising the plan might not currently be designed to protect the brand's future.