What is Maximise conversion value?

Maximise conversion value spends the budget to produce the greatest total conversion value it can. The distinction from its count-based sibling is one word and it changes what the campaign buys.

A count-based strategy ranks a A$40 order and a A$400 order identically, so it prefers whichever is cheaper to win — which is almost always the small one. A value-based strategy ranks them ten to one, so it will pay considerably more to win the large one and still be ahead.

Where every conversion is worth about the same, this difference is immaterial and the two strategies behave alike. Where value varies — a catalogue with a wide price range, a funnel that produces both small trials and large contracts, a business whose margins differ by product line — it is the most consequential setting in the account.

The model maximises the number the account attaches to each conversion. It has no independent view of what a conversion is worth, and it will pursue whatever it is given with complete fidelity.

There are three quality levels and the difference between them is large.

A fixed value on every conversion. Not a value signal. Weighting everything equally leaves the ranking unchanged, so the strategy is maximising a count. Accounts in this state frequently believe they are running value-based bidding and are not.

Revenue as the value. The common implementation, and adequate where margins are even across what is sold. Where they are not, it produces a predictable failure: the strategy finds the highest-revenue orders, which in a mixed catalogue are frequently the thinnest-margin ones. Reports look strong and the money does not follow.

Profit as the value. Revenue less the costs that scale with the order — goods, shipping, payment fees, expected returns — sent with the conversion event. This aligns the model with what the business keeps. It requires product-level cost to reach the conversion payload, usually through a server-side event or an enriched feed, and it is the single highest-value engineering task available to most mixed-margin accounts.

Without a target attached, it spends the budget. The objective is total value, and unspent budget is value not captured, so it will keep buying as long as there is money.

That means efficiency is an output. Return on ad spend will fall as spend rises, because the highest-value opportunities are bought first and the marginal ones are worth less. This is the normal shape and not a fault; the question is whether the marginal conversion is still above the business's break-even, which the strategy does not know unless a target says so.

With a target attached, the arrangement inverts: the bidder declines auctions it does not expect to clear the return, delivery becomes the free variable, and the budget may go unspent. Both configurations are legitimate and they answer different questions. A business constrained by total spend wants the first. A business constrained by required return wants the second.

Stale or wrong values. A value feed that has drifted, a margin model nobody owns, or a currency mismatch all produce a model optimising confidently toward the wrong thing. Because the failure is silent and the reports look internally consistent, it can persist for a long time.

Outliers. A single very large conversion can pull a value-based model toward a pattern that will not repeat. Accounts with occasional enormous orders sometimes cap the value sent for exactly this reason, trading a little accuracy for stability.

First-order value only. The value attached is the value of that transaction. A customer who will return four times is worth far more than their first order, and the model cannot see it. Where repeat purchase is a large share of the business, sending a value that reflects expected lifetime contribution rather than the first order changes what the campaign acquires — and requires a lifetime estimate somebody is willing to defend.

Thin volume. Value-based prediction needs more data than count-based prediction, because it is estimating a distribution rather than a probability. An account short of conversions is short of value-based bidding.

Fix the value pipeline before anything else, and confirm it by comparing the values the platform received against the orders the business actually took.

Decide whether the account is constrained by budget or by required return, and configure accordingly rather than attaching a target out of habit.

Expect the campaign to spend differently within a fortnight of a real value signal arriving. That change is the strategy working, and it is worth telling whoever watches the reports beforehand.

Watch profit rather than revenue when judging the result, because a value strategy fed revenue can raise revenue and lower profit, and the revenue-based report will be delighted about it.

The visible effects are consistent enough to be worth anticipating, because several of them look like problems to anyone not expecting them.

Fewer conversions, more money. The strategy trades small cheap conversions for large expensive ones. A count-based report will show the campaign getting worse.

A higher cost per conversion. Same cause. The figure is no longer the one being optimised and should stop being the one being watched.

A different product mix. Catalogue and shopping campaigns shift toward whatever carries the value being sent. If profit is the value, the mix moves toward margin, which is frequently a visible change on the shop floor as much as in the reports.

A wider spread of outcomes. Value-based bidding is willing to lose several small auctions to win one large one, so week-to-week variance increases. Reading a single week as a verdict is a mistake under any automated strategy and especially under this one.

Telling whoever reads the reports about all four beforehand is worth more than it sounds, because the first fortnight after a value signal lands looks like a regression on every count-based measure in the account.

Send a varying value before switching the strategy, and check it against the orders the business actually took. A value strategy with no value signal is a count strategy, and switching first means the change that matters happens silently later.

Then move to the value-based strategy without a target, and watch what the campaign buys for a few weeks.

Then, if the business has a required return, attach a target derived from contribution margin rather than from what the campaign happened to achieve.

Doing those three in one week produces a result nobody can attribute to any of them.