The problem with monthly
Monthly reporting finds problems after the budget that caused them is already spent. By the time a monthly review flags a declining channel or a margin slip, four more weeks of the same spend pattern have already gone out the door, on top of the month the review is actually looking at. A monthly cadence is not wrong because monthly numbers are inaccurate. It is wrong because it is structurally too slow to change a decision before the next round of spend commits.
Three numbers, not a dashboard
A weekly cadence does not need to be exhaustive to work. It needs three numbers, tracked consistently in the same format every week: contribution margin per order, blended ROAS across every channel, and the week-over-week shift in channel mix.
Contribution margin per order catches a promotion or a rising platform fee eating into unit economics before it shows up as a quarter-level surprise. Blended ROAS, built from one total revenue figure divided by one total spend figure rather than an average of per-channel numbers, catches whether the whole system is still profitable even as individual channels move in different directions. Channel mix shift catches something neither of the other two numbers can see on its own: spend quietly drifting toward a channel that reports well but carries thinner margin underneath.
Why these three together, not separately
Each number alone can mislead. A steady blended ROAS can be masking a real decline in one channel if a lagging channel, such as affiliate reporting two to four weeks behind, is temporarily understating its own contribution and dragging the blended figure down for reasons that have nothing to do with performance elsewhere. A stable channel mix can be hiding a margin problem if contribution margin per order is falling evenly across every channel at once. Only the combination catches both failure modes: a system that looks fine at the top line while one part of it quietly deteriorates.
Running all three in the same weekly view also surfaces a fourth thing implicitly: whether a shift is real or noise. A single week’s dip in any one number is common and often meaningless. The same dip repeating for three consecutive weeks, visible because the format never changes, is a pattern worth acting on. Without a consistent weekly baseline, that distinction between noise and pattern is impossible to make, because there is no comparable prior week to check it against.
The discipline is the format, not the sophistication
The value in this cadence comes from consistency, not complexity. The same three numbers, pulled the same way, presented in the same layout, every single week, regardless of whether that week looked eventful or quiet. A team that only runs the numbers when something already looks wrong has already lost the early-warning benefit the cadence exists to provide, since by the time something looks wrong enough to check, the pattern has usually been running for several weeks already.
Run the same three numbers every week: contribution margin per order, blended ROAS, channel mix shift. In the same format, every time, whether or not anything looks unusual that week.