When to use a cost cap

Predictability per result, paid for in reach.

An outcome-shaped strategy without a ceiling will pay above your number when the auction demands it, on the reasoning that the average is what matters. A cap declines those auctions. What you get is a tighter distribution of costs and a smaller number of results; what you lose is every result that was available above the line, including the ones that would have been worth having.

There is a second cost that is easy to miss. A campaign that cannot spend its budget leaves money unspent in a window, and windows do not always come back — a seasonal peak spent at a third of plan is a peak you did not have.

The economics are fixed per unit and the business cannot absorb variance. A known margin per sale, a cash position that will not tolerate a bad fortnight, a client contract that prices per result. Here the average is not what matters, because the bad days have consequences that the good days do not undo.

You are protecting a line rather than growing one. A retargeting line with a hard pool ceiling, an always-on campaign whose job is to be there rather than to scale. Volume above the cap is not what these lines are for, and paying more to get it is buying something you did not order.

You are bounding an experiment. A new audience, a new geography, a new offer. A cap turns an open-ended question into a bounded one: if it can be bought at this price, it is worth exploring further. That is a cleaner experiment than one whose cost is discovered after it ran.

On a campaign that has not yet worked out who to show ads to. A new campaign with a tight cap may never accumulate enough delivery to stabilise, which produces a campaign that is simultaneously efficient and pointless. Get volume first; constrain second.

When the cap is really a target you wish were true. If the number came from what you would like to pay rather than from what the audience currently costs, the cap will do what it is designed to do — refuse — and the diagnosis will be a month of confusion about why delivery is thin. A cap is an instrument for holding a line you have reason to believe is reachable, not for wishing one into existence.

Derive the number the same way as a target: from your margin, backwards, using the CPA calculator on your own figures. The difference is where you place it relative to the distribution — a target sits below the ceiling with room for an average, a cap sits at the point you genuinely will not pay past.

Three readings, all from the daily counts.

Spend against budget. A capped campaign spending well under its budget is telling you the cap is below what the auction is charging. That is information rather than a fault.

Volume trend against price trend. If the cost of the audience is drifting up, a fixed cap becomes progressively tighter without anybody changing it. A cap set once and never revisited is a slow strangulation.

What the uncapped comparison looked like. Not a test, just history. The record holds what the campaign did before the cap and what it did after, so the comparison is at least available. Every applied change is a row with what it moved and from what.

Read the rate beside the volume, always. A cap makes every reported rate look good by construction, which is precisely why the rate alone cannot tell you whether it is working.

A cap is carried on a bid-shaped change request, with money fields as integer minor units of the account's currency. Guardrails are checked when the change request is written, again at approval, and once more immediately before a platform is called, using the same functions each time.

Stated once: your connections read your accounts, a change is written as a draft on the record with the state it would change from and the checks it passed, applying it is gated on approval — by you, or by a rule you wrote — and no platform write reaches a customer account until that platform has granted write access. Applying a cleared change to a platform is in private beta.

A rule you wrote can clear a decrease in a bid-shaped amount under a cap it states in a currency it states. Loosening a cap always waits for a person, and so does any change whose direction cannot be read against the previous value.

Whether a platform offers a true cap at all is that platform's business, and it is stated on the connection: each one carries a capabilities object describing which change operations it supports, where budget attaches, which objectives it accepts and whether a change can be dry-run before it is applied. Read that rather than assuming every platform behaves like the one you know best.

It is not a spend control. It constrains price per result while the budget constrains how much money leaves, and a campaign with a tight cap and a large budget can still spend the whole budget if the audience is cheap enough.

The limits that actually stop spend running away sit underneath and cannot be raised: the account's own daily ceiling, derived from what that account has actually done; the rule that a budget may not move more than a hundred per cent of its current value in one change; and the rolling twenty-four-hour sum of applied increases on a connection, which is the only check that can refuse a change every earlier check allowed.

If the problem you are solving is "spend is running away", those are the instruments, along with the connection's kill switch — pausing a connection stops syncing and blocks every apply written against it, whatever is queued and whatever rule would otherwise have cleared it.

Pace a budget without a month-end scramble is the spend-shaped version of this problem, and Set a target CPA that holds is the instrument to reach for when what you want is an average rather than a line.