Performance Marketing

Meta and Google, end to end. Nine brands across six categories: streaming, film and entertainment, securities markets, finance education, AI tooling, heritage retail, a microdrama app, a restaurant business. Around $2M a year in total.

One budget per account and a target that does not move because a channel had a bad week. The split between the two platforms is a live decision, not a quarterly plan. It changes when the marginal dollar changes, and that can happen inside a week.

Nine accounts in six categories means the same playbook fails in at least two of them at any given time. A jewellery buyer and a streaming signup are not the same purchase and should not be bought the same way.

Most of paid media is not strategy. It is a small number of decisions made repeatedly, with real money on each one. What to fund tomorrow. What to stop funding today. Which creative is still working and which one is being carried by the account around it.

Every asset gets a unique promo ID the moment it is created, so a cut that worked can be found again six weeks later. Attribution at the asset level rather than the campaign level is what makes the repeated decision a decision rather than a guess.

The rest is reading the auction. Prices move because of what other advertisers are doing, not because of anything I did. Telling those two apart is most of the skill.

Around $2M of media returned $6.7M. Blended across every account, 3.3x.

Most of the movement was incrementality work. Pulling budget off placements that were harvesting demand which would have arrived anyway, and buying into the ones where the marginal dollar had not turned yet. No single lever did it, and none of it would have been visible at the campaign level.

3.3x is an average over nine accounts. Averages hide their own spread. Some of those accounts carry the number and some of them drag it, and the useful work is always in the second group.

Around $167K is live every month. That figure is set before the month starts and it does not grow because something is working.

Everything downstream is a consequence of holding it. Creative volume, platform split, bid strategy, which experiments get funded and which ones wait. A budget you can raise mid month teaches you very little. A fixed one forces the actual question, which is what this dollar comes back as and how fast.

How fast matters more than how much. Money that returns inside the month gets redeployed inside the same quarter. That is also what makes defunding something that is working a normal decision rather than an admission of error.

Paid media does not create demand. It finds the people who already want the thing, at a price, and it stops finding them at a certain volume. Past that point every additional dollar comes back as less than the one before it, and spending it anyway is a choice to buy revenue at a worse price.

The ceiling also moves. It moves with what the business ships, with seasonality, and with whatever a competitor is doing to auction prices that week. Finding it again is a weekly exercise, not a one time calculation.

The larger limit is that some of the biggest moves in what the media returns came from something the business did, not from a bid. Reading that honestly is what stops you taking credit for work you had nothing to do with, and it is the only way the next forecast is worth anything.