What is Bid?

A click that converts 3% of the time into a customer contributing A$600 is worth 0.03 × 600 = A$18 at break-even. A bid ceiling of A$18 buys every click the business can afford and leaves nothing for fixed costs; A$12 leaves a third of the contribution behind. A bid of A$1.50, chosen because it looks like a normal click price, declines almost every auction the business would have profited from.

A bid is the maximum an advertiser is prepared to pay for whatever the auction is selling. Depending on the marketplace that is a click, a thousand impressions, a view or a conversion.

The word invites a misreading that costs accounts real money: people treat a bid as a price. On search and social it is not. Those auctions are second-price or a close variant, so the winner pays approximately what was needed to beat the next competitor and the bid is an upper limit that is frequently not reached. On most programmatic display, which moved to first-price, the bid and the price are the same number.

Two practical consequences follow immediately. Raising a bid does not automatically raise costs in auctions already being won comfortably. And an average cost well below the bid cap is evidence of headroom rather than of misconfiguration.

The only defensible source is the value of the outcome and the probability that it follows.

Bid ceiling = value of the outcome × probability it follows

For a click, that means the contribution a customer produces multiplied by the rate at which clicks become customers. A click converting at 3% into a customer worth A$600 of contribution is worth A$18 at break-even. The working bid sits under that ceiling by however much of the contribution has to cover fixed costs and profit.

For a thousand impressions, the chain is one step longer: impressions become clicks at the click-through rate, clicks become customers at the conversion rate, and the ceiling is the product of the three.

What the number should not come from is what a click usually costs, what a competitor appears to pay, or what the account bid last year. Each of those describes somebody's circumstances rather than your economics, and a bid derived from them will be arbitrarily high or arbitrarily low with no way to tell which.

Manual bidding sets a figure per keyword, ad group or placement, and the advertiser owns every one of them.

Its advantages are transparency and control. Nothing is inferred, changes take effect immediately and predictably, and the account behaves the same way today as yesterday. On thin accounts — too few conversions for a model to learn from — it is frequently the better choice, because an automated strategy with nothing to learn from produces erratic delivery rather than intelligence.

Its disadvantages scale with the account. A manual bid is the same for a person likely to convert and a person unlikely to, on a device that performs well and one that does not, at an hour that works and one that does not. Bid adjustments recover some of that granularity and never approach what a per-auction model achieves.

Automated strategies replace the per-auction bid with an instruction: aim for this cost per acquisition, aim for this return, spend this budget to maximise conversions.

The platform then computes a bid for each individual auction from its estimate of that auction's value, which is granularity no manual scheme can match.

The arithmetic behind the number does not change. A target cost per acquisition still has to come from contribution and close rate; a target return still has to come from margin. The difference is where the number is applied, not how it is derived — and a strategy pointed at a target the market cannot supply will not achieve it cheaply, it will simply stop delivering.

Two operational rules matter more here than under manual bidding. Automated strategies need conversion volume — enough optimisation events inside the window that their platform considers the campaign out of learning — and they need stability, because every material change resets the model into a period of unstable delivery.

Auctions rank on expected value, not on money. The bid is multiplied by the platform's estimate of how likely the paid-for event is, and adjusted by quality signals.

The consequence for bidding practice is direct: a higher bid wins more auctions and pays more for them, while a higher predicted action rate wins more auctions and pays less. The second is strictly better and it is reached through ad copy, relevance, structure and landing experience rather than through the bidding interface.

The efficient sequence for an account losing auctions is therefore copy and relevance first, landing page second, structure third, bids last. Reaching for the bid first is the most common order and the most expensive.

A bid increase that produces no additional volume usually means rank was not the constraint.

If the campaign is limited by budget, it was already declining auctions for lack of money, and making it willing to pay more for auctions it cannot afford to enter changes nothing.

If the eligible pool is already mostly won, there is little left to win and the increase buys a higher clearing price on existing wins.

If the losses are driven by quality signals rather than price, a higher bid compensates weakly and expensively for a problem located elsewhere.

The impression share columns separate these three in about a minute, which is why they belong in front of any bidding decision rather than after it.

Changes should be small, spaced and singular.

Small, because a large move is a different campaign rather than an adjustment, particularly under automated strategies where it resets the model. Spaced, because the effect cannot be read until at least a conversion window has passed. Singular, because two changes at once produce a result attributable to neither.

And every change should be judged on cost per outcome against the ceiling, rather than on position, impression share or volume. Those are means. The ceiling is the only number in the account that comes from the business rather than from the auction.