What is Bid strategy?

A bid strategy is the rule a campaign follows when deciding what an auction is worth. Described that way it sounds like a technical setting. In practice it is the most consequential statement of intent in the account, because every strategy pursues what it was told to pursue with complete indifference to whether that was the right thing to want.

A campaign told to get the most clicks for its budget will find the cheapest clicks. A campaign told to get conversions at a target cost will decline anything it does not expect to clear that cost. A campaign told to maximise conversion value will chase the highest values it can find, which in a catalogue with uneven margins is not the same as the most money.

None of those behaviours is a defect. Each is the strategy working. The mistakes in this area are almost never the algorithm failing; they are an objective stated carelessly and then executed faithfully.

Strategies group into three kinds, distinguished by what the advertiser controls.

Manual. The advertiser sets a figure, per keyword, ad group or placement, and it applies to every auction that entity enters. Total control, no per-auction intelligence.

Volume-maximising. The advertiser gives a budget and an objective — clicks, conversions, conversion value — and the platform spends the budget to get as much of that objective as it can. There is no cost target, so the resulting cost per outcome is whatever the auction produced.

Target-based. The advertiser gives a target — a cost per acquisition, a return on ad spend, an impression share — and the platform bids per auction to hit it on average. Delivery becomes a consequence of the target rather than of the budget.

The progression across those three is a trade of control for granularity. A manual bid is one number applied to millions of auctions. A target-based strategy computes a different number for each of them, using signals the advertiser cannot see, and in exchange the advertiser stops being able to say what any individual bid will be.

Maximising clicks suits a campaign whose product is traffic, and almost nothing else. It is also a legitimate first phase for a new campaign that needs to accumulate data before a conversion-based strategy has anything to learn from — used deliberately, with an end date.

Maximising conversions suits a campaign with conversion tracking, enough volume for the model, and a budget that is the real constraint. It has no cost ceiling, so the cost per conversion floats to whatever the budget buys.

Maximising conversion value is the same arrangement pointed at value rather than count, and it suits any account where conversions are worth materially different amounts. It is only as good as the values sent back: given revenue in a catalogue with uneven margins it will find high-revenue, low-profit orders.

Target cost per acquisition suits an account that knows what a conversion can cost and wants delivery bounded by it. Volume becomes the free variable.

Target return on ad spend is the value equivalent, and the same caution about value quality applies with more force.

Target impression share is a visibility strategy rather than a performance one. It buys position, which is occasionally exactly what is wanted — brand defence being the clear case — and is otherwise an expensive way to be seen.

Three things settle which strategy a campaign can actually run, and they are worth checking before the debate about which is best.

Conversion volume. Every model needs examples, and a campaign producing very few gives its strategy almost nothing to predict from. Below whatever threshold the platform applies, a conversion-based strategy is guessing and delivery is erratic — the learning phase is where that constraint is stated in each platform's own terms.

Measurement quality. A strategy optimising toward a mismeasured conversion will pursue the mismeasurement with the same diligence it would bring to a real one. Duplicate tags, a conversion set to an add-to-cart in a business that sells things, or a lead form with no qualification all produce confident, wrong optimisation.

The value signal. Value-based strategies need values. A static value on every conversion is not a value strategy; it is a count strategy in different units.

Where any of the three is missing, the honest answer is to fix it rather than to choose a more sophisticated strategy that depends on it.

A strategy change is not a setting change. The new strategy has no history to predict from, so the campaign enters a period where delivery is unstable and the reported cost is unrepresentative of anything.

Accounts that respond to that instability by changing again never leave it. The discipline is to change one thing, wait at least the conversion window and preferably longer, and only then read the result. Where several campaigns need changing, staggering them keeps at least part of the account stable and makes the comparison possible.

The corollary is that strategy changes should be rare and deliberate. An account that has been through four strategies in a quarter has spent most of the quarter learning.

It cannot fix a target the market will not supply. A target below anything the account has achieved does not lower cost; it lowers delivery, because the bidder declines the auctions it does not expect to clear.

It cannot fix creative. Auctions rank on expected value, and an ad nobody clicks costs more to serve whatever the bid.

It cannot fix the offer, the landing page or the conversion rate, all of which sit between the click and the outcome the strategy is being judged on.

And it cannot decide what an outcome is worth. That number comes from contribution margin and close rate, and it is the input every target-based strategy depends on and none of them supplies.