What a bid strategy optimises
A bid strategy answers one question for every auction a campaign enters: how much is this click worth, given what the advertiser wants? The advertiser does not set bids; the advertiser sets the goal (conversions, conversion value or clicks) and, optionally, a target that constrains it, such as an average cost per acquisition (CPACost of each conversion.) or a return on ad spend (ROASRevenue generated per unit of ad spend.). The strategy then predicts, per auction, the likelihood and value of the outcome and bids accordingly.
In SA360 the strategy is always a portfolio, which has two consequences. The first is that the goal is shared: every campaign in the portfolio is optimised towards the same outcome and the same target, and the strategy moves spend between them towards wherever the target is most achievable. The second is that the portfolio can cross engine accounts, which is what no engine's native bidding can do.
What the strategy learns from is the conversion data it is given. SA360 is unusual in that this can come from outside the engine: Floodlight activities tracked by Campaign Manager 360 and joined to clicks through the engine's click identifier, so a conversion on the confirmation page counts whichever engine delivered the visitor. That one definition, applied across the portfolio, is the main reason to bid in SA360 rather than in each engine. It also means the strategy is only as good as the Floodlight join: a tag that fires without the click identifier produces a conversion the strategy never sees. The Floodlight and reporting guide covers how that join works and how to check it.
Goals and constraints
| Goal | What it maximises | Optional target | Constraints the help centre documents | Typical use |
|---|---|---|---|---|
| Conversions | The number of conversions from the goal's activities | Target CPA | Minimum and maximum bid limits; campaign budgets | Lead generation; sales where each conversion is worth about the same |
| Conversion value | The total value of conversions, from Floodlight revenue or engine values | Target ROAS | Bid limits; campaign budgets | E-commerce; any goal where conversions differ in worth |
| Clicks | The number of clicks within budget | Maximum cost per click | Bid limits; campaign budgets | Traffic goals; new campaigns gathering data before a conversion goal exists |
| Engine-specific goals | Position or impression share objectives where the engine supports them | Varies by engine | Per engine, in the help centre | Brand defence; visibility campaigns on a single engine |
A target is an average the strategy aims for across the portfolio over time, not a cap on any one conversion; some conversions will cost more than the target and some less. Set a target and the strategy may leave budget unspent when it cannot find enough auctions it expects to meet it; leave the target off and the strategy spends the budget for as many conversions, or as much value, as it can. The right starting point for a portfolio with no history is no target, to learn what the campaigns achieve at the budget, then a target set near that figure and moved in steps towards the one the margin arithmetic requires. The bidding hub sets out how a target derives from margin, and the ROAS calculator does the sum for value goals.
Bid limits are the constraint most often misused. A minimum or maximum bid stops the strategy from bidding outside a band, which is reassuring and costly: Google's bidding documentation advises against limits on auction-time bidding because they prevent the strategy from bidding high in the auctions it predicts will convert and low in the ones it predicts will not. They have a place on non-Google engines, where bids are pushed on a schedule and a runaway bid cannot be corrected until the next cycle, and in the first days of a new strategy where the team wants a safety rail; they should be removed once the strategy is trusted.
Portfolios across engines
A cross-engine portfolio assigns Google Ads campaigns and, say, Microsoft Advertising campaigns to one strategy with one goal. The strategy treats them as one pool for the target: it does not promise each engine the same CPA, it promises the portfolio an average, and it will shift spend towards the engine where conversions are cheaper. That is the behaviour most teams want and the behaviour that surprises them when the Microsoft campaigns' spend drifts down.
The mechanics differ by engine, and the difference is documented rather than incidental. For Google Ads campaigns, the strategy sets a bid in each auction using auction-time signals: query, device, location, time, audience membership and the rest. For other engines, SA360 calculates bids and sends them to the engine through its interface on a schedule; the engine then uses those bids in its auctions until the next update. Google does not publish the cadence as a guarantee, and the help centre describes it per engine. The consequence is that the non-Google half of a portfolio responds more slowly, learns from thinner data, and cannot react to a signal that only exists inside the auction. A portfolio that spans engines is therefore two bidding systems sharing a target, not one.
Three design rules follow. Keep campaigns with genuinely different economics (brand and non-brand, markets with different margins) in separate portfolios whatever engine they run on, as the cross-engine management guide recommends for structure generally. Give a cross-engine portfolio enough conversion volume on the smaller engine that its bids are set from data rather than from the portfolio average. And report the portfolio by engine as well as in total, so that a target met overall is not hiding one engine far above it.
Where a non-Google engine's own automated bidding is strong and the engine carries little spend, the alternative is to leave those campaigns on the engine's native bidding and keep them out of the SA360 portfolio, reporting them in SA360 but not bidding them there. The help centre documents which engine-native strategies can coexist with SA360 management.
Conversion goals from Floodlight
A conversion goal is the definition of success the strategy optimises towards. In SA360 it is built from conversion sources: Floodlight activities from the linked Campaign Manager 360 advertiser, conversion actions from the engine, or both, each optionally weighted by a value. A sales activity contributes revenue, so value goals can use it directly; a counter activity contributes a count, and a value can be assigned to it so that leads of different kinds are weighted differently in a value goal.
Four settings decide what the strategy sees. The activities included: a goal that includes both a Floodlight purchase activity and the engine's purchase conversion action counts the same sale twice unless one is excluded. The AttributionAssigning conversion credit to touchpoints. model applied to Floodlight conversions, which decides how a conversion that followed clicks on several engines is credited; the default is last click, and the help centre lists the other models available for Floodlight conversions. The lookback window set in the Floodlight configuration, which decides how long after a click a conversion still counts. And the conversion lagThe time between an ad interaction and the conversion it leads to, which leaves the most recent reporting days incomplete.: a Floodlight conversion is reported some time after it happens, so the strategy's recent window is always incomplete, and the help centre describes the delay to expect.
Floodlight custom variables extend the goal. A variable passed with the tag (order margin, product category, lead quality) can filter which conversions count or supply the value the strategy optimises, so a value goal can bid to margin rather than revenue when the tag passes margin. Setting the tags up to pass those values is Campaign Manager 360 work; the Floodlight tags guide covers it, and the SA360 side is choosing the variable in the goal.
A change to any of these (adding an activity, switching the attribution model, changing the window) is a change to what the strategy is learning from and resets its understanding of the campaigns. Make such changes rarely, deliberately, and one at a time.
Learning, changes and evaluation
A new strategy, or one whose goal, target or campaign set changes significantly, enters a learning period during which the help centre says performance may fluctuate while it adjusts. The strategy's status shows it. The practical rules are the same as for Google Ads Smart Bidding: wait before judging, change targets in small steps rather than one jump, change one thing at a time, and avoid changing a strategy during a period that is already unusual, such as a sale, because the strategy will learn the unusual period as normal.
Evaluation has to respect the lag and the attribution. Compare the delivered CPA or ROAS with the target over a window long enough for Floodlight conversions to have arrived, and on the same attribution basis as the goal; a target set on Floodlight last-click conversions evaluated against the engine's own conversion column will never agree, for the reasons the Floodlight and reporting guide sets out. SA360's bid strategy reporting shows target against delivered over time, and the useful reading is the trend after the learning period, not any single day.
Three signals mean something is wrong rather than merely settling. Spend far below the campaigns' previous level with a status that is not learning usually means a target the strategy cannot meet. A delivered figure that meets the target in total while one engine sits far above it means the portfolio is hiding a problem. And conversions that stop arriving in the goal while the engine still reports them means the Floodlight join has broken, which is a tracking problem, not a bidding one.
A cross-engine target ROAS portfolio over its first quarter
A furniture retailer assigns 22 non-brand Google Ads campaigns and 22 mirrored Microsoft Advertising campaigns to one conversion value strategy with a Floodlight sales activity as the goal. For the first four weeks there is no target; the portfolio spends £96,000 and returns £412,000 in Floodlight revenue, a ROAS of about 430%, with Microsoft contributing 14% of spend at 390%. Gross margin implies a break-even ROAS of 280% and the business wants 420% after a profit allowance, so the team sets 400% and moves to 420% three weeks later. Spend dips to 80% of its previous level for ten days after each change, then recovers. By week twelve the portfolio delivers 428% on £104,000 a month, but segmented by engine Microsoft sits at 350% because its bids are pushed on a schedule and its conversion volume is thinner. The team moves the Microsoft campaigns into their own portfolio at 380%, which they hold, and leaves the Google campaigns at 430%.
SA360 vs engine-native bidding
| Factor | SA360 bid strategy | Google Ads Smart Bidding | Native bidding on other engines |
|---|---|---|---|
| Scope | Campaigns across engine accounts and engines | Campaigns within one Google Ads account or manager account | Campaigns within that engine |
| Bid timing on Google Ads | Auction time | Auction time | Not applicable |
| Bid timing on other engines | Scheduled push through the engine's interface | Not applicable | Auction time, within that engine |
| Conversion source | Floodlight, engine conversions or both | Google Ads conversion actions, including imports | The engine's own conversion tracking |
| Attribution across engines | One model across SA360-tracked clicks | Google Ads credits its own clicks | The engine credits its own clicks |
| Custom values | Floodlight custom variables and engine values | Conversion value rules and per-action values | Per engine |
| Best for | Multi-engine portfolios; one Floodlight definition across channels | Accounts that run on Google Ads only | Low-spend engines left outside SA360 bidding |
The decision is rarely all-or-nothing. A common pattern is SA360 strategies for the cross-engine non-brand portfolios where one Floodlight definition matters, Google Ads Smart Bidding left in place for campaign types SA360 manages less fully, and engine-native bidding for a low-spend engine where the scheduled push would be a handicap. What should not happen is a campaign bid by two systems at once: a campaign in an SA360 portfolio must not also sit in a Google Ads portfolio strategy, and the help centre is specific about which combinations are allowed. The tutorial on moving paid search into Search Ads 360 hands bidding over portfolio by portfolio with that rule in mind.
For an advertiser on Google Ads alone, the Smart Bidding guide describes a system that already does everything an SA360 strategy would do for those campaigns, with fewer moving parts. SA360 bidding earns its place when the portfolio spans engines or when Floodlight is the conversion definition the business has already chosen for every channel.
Common mistakes
- Setting a target before the portfolio has history. A tight target on a new strategy produces little spend and a flattering ROAS on what remains.
- One portfolio for everything. Brand and non-brand, or markets with different margins, sharing a target the strategy averages across.
- Judging the Microsoft half by the Google half. Scheduled bid pushes and thinner data mean the engines in one portfolio behave differently; segment by engine before concluding.
- Double-counting conversions in the goal. A Floodlight purchase activity and the engine's purchase action in the same goal count each sale twice.
- Evaluating on a different basis from the goal. A target set on Floodlight last-click conversions, judged against the engine's conversion column.
- Changing several things at once. A new target, a new activity and a budget change on the same day, with no way to tell which did what.
- Bid limits left on for ever. A safety rail for the first week becomes a permanent brake on auction-time bidding.
- Reacting to a bad week inside the learning period. The status says learning; the fix is patience, not another change.
The SA360 hub places bidding alongside management and measurement; the Floodlight and reporting guide is where the conversions a strategy learns from are defined and checked.
