Growth & Strategy

Store and search work is the part of acquisition that never shows up on the card statement. It still costs. It is paid for in listing work, keyword coverage and catalogue upkeep, and the bill arrives as time instead of spend, which is why it is easy to leave undone.

It gets worked before the auction does. Anything that compounds while spend sits flat is cheaper than buying the same result, and it lowers the number paid media has to carry next month.

None of it replaces paid. It changes what paid is being asked to do.

Nine brands, six categories, and $167K of live budget every month. There are always more places to put a dollar than there is money, so the real decision is what gets starved.

The frame does not change with the category. What does this dollar come back as, and how fast. How fast matters more, because money that returns inside the month gets redeployed inside the same quarter.

The rest of it is being willing to defund something that is working, because something else is working faster. A line does not earn its budget by having earned it last quarter.

Two or three per cycle. Sized so a loss costs a few days of budget and a win is large enough to read against BAU noise. Below that size the result cannot be separated from normal variance, so running it is spend with no information attached.

Winners move into BAU. Losers get written down and do not get re-run under a different name two quarters later. A test that wins and never graduates is an anecdote.

Experiments are funded out of the return the accounts are already producing. A bad cycle costs a few days of that return, not the month.

Blended return is an output of two things that have nothing to do with the ad account. What the product does with someone in their first week, and what is actually on offer that month.

Some of the largest swings came from a catalogue, not a bid. In streaming that is the slate. In heritage retail it is the collection and the season. In a securities product it is the market itself, which nobody buys and nobody controls.

Reading that correctly is what stops you claiming a release you had nothing to do with, and what stops you rebuilding an account that was never the problem.

The allocation frame assumes you can see what came back and roughly when. Where the purchase decision takes months, heritage retail being the clearest case, the return lands outside the window you are measuring in. Judged on speed alone, those lines look worse than they are, and the frame will quietly defund them.

Speed has its own failure. A budget rewritten every week compounds nothing. Some of the decisions that returned the most needed a quarter of being left alone, and the discipline is knowing which ones those are before you start moving money.

Reallocation also only works while the destination can absorb it. A line that returns well at a small size will not hold that return at ten times the money, and where it breaks has to be found again every time the budget grows.