Reach or frequency? Buy reach.
A gross rating point is reach multiplied by frequency, which means a plan can hit its GRP target while buying almost entirely one or the other. Those are very different plans. All else equal, the evidence points to reach.
What a GRP conceals
GRPs are the currency most television and video plans are bought in, and the arithmetic is simple. Reach times average frequency. A hundred GRPs might be fifty percent of the audience reached twice, or ten percent reached ten times. Same number on the plan, completely different campaigns.
Because the metric collapses two variables into one, the tradeoff often never gets discussed. The plan clears its GRP goal and nobody asks which half of the product it bought. That is the question worth asking, because the two do not perform the same.
Some repetition is necessary. Not much.
The case for frequency is real and it is where most planning convention comes from. In 1972 Herbert Krugman argued that three exposures were roughly what it took: the first creates awareness of something new, the second allows evaluation, the third acts as a reminder. That became effective frequency planning, and it is why so many plans still carry a three plus target.
The part that gets lost is what happens after that. The curve flattens quickly. Whatever registration a viewer needs, they largely have it early, and further impressions against the same person buy progressively less. Meanwhile every additional impression against someone already reached is an impression not spent on someone who has seen nothing at all.
The fourth impression on one person competes directly with the first impression on another. Past a low threshold, that trade is a bad one.
What the evidence says
The strongest body of work here comes from the Ehrenberg-Bass Institute, whose position is that reach matters more than frequency of exposure, and continuous advertising beats bursts followed by long gaps. That conclusion rests on decades of category data rather than a single study.
The mechanism behind it is penetration. Andrew Ehrenberg examined 157 brands and found the factor most closely associated with growth or decline was the size of the user base rather than the behavior of existing buyers. The IPA databank points the same direction: across 880 effectiveness papers, 82 percent of reported growth came from penetration and roughly 2 percent from loyalty. Brands grow by being bought by more people. Reach is how you find more people.
A second line of evidence arrives from a different direction. John Philip Jones found that a single exposure close to a purchase decision was often enough to shift choice, which with Erwin Ephron became recency planning: spread weight across more people at lower frequency, and stay present continuously rather than concentrating into flights. Two different research traditions, converging on the same practical advice.
The problem with average frequency
Here is the operational argument, and in practice it is the one that changes plans.
Average frequency is an average over a badly skewed distribution. Ehrenberg-Bass has pointed out that real campaign delivery often reaches heavy viewers more than a hundred times in a year while large parts of the target audience receive one or two opportunities to see, or none. A plan reporting an average frequency of four may have delivered forty to a small group and zero to a much larger one.
Nobody buys that on purpose. It is what happens when you optimize toward a GRP number without looking at the distribution underneath it, and it is why frequency capping and reach curves belong in a plan review rather than average frequency alone.
When frequency deserves more weight
The reach-first position is not unconditional, and the exceptions are worth stating plainly.
- High consideration categories. Jones and Ephron's findings came largely from packaged goods. More repetition is generally needed for considered purchases such as automotive, financial services, and pharmaceutical, where a single exposure is unlikely to carry a complex proposition.
- New brands and new messages. If nobody knows who you are, or the claim is unfamiliar, the early part of the registration curve is steeper and worth paying for.
- Short, dated windows. A sale ending Sunday is not a penetration strategy.
- Very small budgets. Reach spread thin enough becomes noise. Concentration can beat dilution when the money will not cover a market.
None of these overturn the default. They describe when to depart from it, which is a different thing.
What a model can tell you
All of the above is prior knowledge. The useful question is what your own data says, and a media mix model can answer it if the plan is specified to allow it.
Model impressions and reach separately where you can. If the only input is spend, the model cannot distinguish a wide plan from a heavy one. Reach and frequency delivery, where the data exists, should enter as their own variables.
Read the saturation curve as a frequency statement. A channel deep into diminishing returns is usually one where added weight is landing on people already reached. The curve is telling you about distribution, not just about spend.
Use regional variation. Markets with similar spend but different reach and frequency mixes are the closest thing to an experiment most advertisers already own, and a model with regional structure can exploit it.
Test it. A geo holdout comparing a reach weighted plan against a frequency weighted one at equal spend settles the question for your category better than any general principle, this one included.
Where this lands
Buy enough frequency to register, then spend everything else on reach. In most categories that threshold is low, and plans routinely exceed it by a wide margin without anyone deciding to.
The reason is structural. GRPs are easy to buy and frequency is the cheaper way to accumulate them, so plans drift toward depth against a shrinking group of heavy viewers. That drift is rarely a decision. It is what happens when the metric on the plan cannot distinguish between the two things you might be buying.
We build media mix models that run continuously rather than quarterly, for advertisers between five and fifty million in annual media. If that is you, get in touch.