Paid search and paid social are not competing for the same job

The quarterly budget conversation usually goes the same way. Paid search converts at 3 percent, paid social at 0.8 percent, so move the money. It is a defensible reading of the numbers and it is frequently how a company stops growing.

The two channels do different jobs, and only one of them can be scaled by wanting to scale it.

Search is bounded by demand

Paid search reaches people who are already looking. That is why it converts well, and it is also its ceiling. There is a finite number of people searching for what you sell this month. You can take more of that traffic by bidding harder, and you cannot create more of it by spending more.

So a search-only strategy is a market share strategy inside a fixed pool. It works beautifully until you own most of the pool, at which point additional spend buys progressively worse positions on progressively less relevant terms, and the reported conversion rate starts to fall for reasons that have nothing to do with your website.

Social makes the demand that search harvests

Paid social reaches people who were not looking. Most of them do not convert, which is exactly what the conversion rate says. Some of them learn your name, and some fraction of those become the branded searches that convert at 5 percent next quarter and get credited to search.

What each channel is good at. Search: intent captured: high. Search: demand created: low. Social: intent captured: low. Social: demand created: high.

This is the mechanism behind an experience most paid teams have had. Social gets paused to fund search. For four to six weeks nothing bad happens, and the case for the cut looks proven. Then branded search volume softens, search conversion rate drifts down because the traffic mix is shifting toward cold non-brand terms, and nobody attributes it to a decision made last quarter.

How to argue about this with evidence

The measurement that settles it is a holdout, and it costs real money to run. Pause the demand-creation channel in one region, keep it running in a comparable one, and watch branded search volume and direct traffic in both over eight to twelve weeks.

The fourth is the leading indicator worth adding to a dashboard. Demand creation shows up as new visitors before it shows up as anything financial, and a decline there is the earliest honest warning that a cut is working out badly.

One more thing worth resisting. Media mix modelling gets sold as the answer to precisely this question, and for most advertisers it is not. The open source implementations from Google and Meta are genuinely capable and they want a lot more spend history than a small advertiser has. Below roughly $50k a month the model has too little signal to say anything you could not get from a holdout test, at a fraction of the complexity.

Related: your traffic sources do not convert equally covers the wider channel spread, and why your ad platform and your analytics disagree explains why the numbers in this argument never quite reconcile.