Search Ads 360 Partners

What the bidding is actually being told to maximize

Before any bid strategy is chosen, somebody has to write down every conversion action the account counts, which of them sit inside the bidding target, and what value each carries. This list is almost always wrong in accounts that grew organically. Typical faults are a newsletter signup and a purchase both counted as conversions with no value differentiation, several actions counting the same underlying event through different tags, and a legacy action nobody remembers creating that still sits in the optimization target. Each bends bidding toward the wrong traffic in a way no keyword work corrects.

The audit is straightforward and belongs in the first fortnight. List every conversion action, its source, counting setting, attribution model, assigned value, and whether it is in the primary set. Then ask, action by action, whether you would spend an incremental pound to obtain one more. Anything you would not should be reclassified as secondary so it still reports without steering bids. Expect this exercise alone to move performance, and expect it to be uncomfortable, because it usually reveals the headline conversion number was counting several things at once.

Getting offline conversions in before the signal decays

For any business where the sale completes days or weeks after the click, the conversion the platform sees is only a proxy. A form submission is not a qualified lead and a qualified lead is not a closed deal, so bidding to the proxy optimizes for whichever sources produce the most of it regardless of quality. Importing the real outcome back against the original click identifier fixes this, and it is the highest-value integration in most lead-generation search accounts. It requires the click identifier captured at form submission, stored in the CRM, and returned later with the outcome and its value.

The constraint that catches people out is timing. Uploaded conversions must arrive inside the platform's accepted window, measured from the original click, and if your sales cycle is longer the true outcome can never be attributed at all. The usual solution is to import a mid-funnel milestone occurring inside the window, carrying a value derived from its historical close rate, then reconcile against the final outcome separately in reporting. Ask how a partner would handle your specific cycle length; it separates people who have done this from people who have read about it.

Bidding to margin rather than to revenue

Return on ad spend targets built on gross revenue treat every pound of sales as equally valuable, which is true in almost no business. A product carrying a forty percent margin and one carrying six percent generate an identical signal, so the algorithm learns to favor whichever is easier to sell, frequently the low-margin one. The fix is to send a value reflecting contribution instead of revenue: revenue less cost of goods, less expected returns, less any fulfilment cost that varies by product. The arithmetic is easy; getting margin data out of the systems holding it is not.

The advanced version substitutes expected lifetime value, which matters most where the first order is unprofitable by design, as in subscription and repeat-purchase businesses. That needs a model of what a new customer is worth over a defined horizon, plus a decision about how much of that future value to declare at acquisition. Be conservative early. Ask how the value is calculated, how often it refreshes, and what happens to bidding when the underlying margin data shifts, because a sudden change in declared values pushes the strategy into a learning period.

Where volume targets quietly cost you money

• maximizing conversions finds the cheapest conversion — which in a mixed account means drifting toward low-value actions and brand queries that would have converted without any bid.
• A target cost per acquisition ignores order size — two customers acquired at the same cost can differ fivefold in what they spend, and bidding cannot see that unless you send it.
• Aggressive targets shrink the account — a target tighter than the account can deliver makes the algorithm stop bidding on anything uncertain, and that volume is rarely recovered by loosening it later.
• Every target change restarts learning — frequent adjustments keep the strategy permanently transitional, which is why a change cadence should be agreed in advance rather than improvised.
• Seasonality needs a declared adjustment — expected short-term conversion-rate shifts around a sale period should be signalled to the platform ahead of time rather than discovered afterwards.

Portfolio bid strategies and when to redraw them

A portfolio bid strategy pools several campaigns under one target so the algorithm can trade performance between them, spending more where returns are strong. That pooling is the point, and it is why the composition of a portfolio is a strategic decision rather than an administrative one. Campaigns grouped together should share a commercial objective and a broadly similar economic profile. Putting a brand campaign into the same portfolio as a prospecting campaign lets the strategy hit its target by leaning on brand traffic, which flatters the reported number while doing nothing incremental.

Portfolios also go stale. They are usually drawn once at setup and inherited for years, long after the goals, product mix and seasonal pattern that justified them have changed. Establish a review point, at minimum quarterly and before any major seasonal period, where the grouping itself is questioned rather than just the targets inside it. Ask when a partner last redrew a portfolio structure and what prompted it. If portfolios get set at launch and tuned thereafter, you are hearing a maintenance posture rather than a management one.

The account structure the automation is standing on

• Match types and negatives still decide what you buy — automation bids on whatever traffic your structure admits, so a missing negative list means optimizing inside a query set you never intended to enter.
• Conversion tracking must be consistent across pooled campaigns — a portfolio whose campaigns are measured with different conversion actions is optimizing toward an incoherent target.
• Budgets constrain strategies more than targets do — a campaign hitting its daily budget every day is not being bid, it is being rationed, and the reported performance reflects the rationing.
• Landing page and feed quality are bidding inputs — conversion rate differences driven by the destination look like traffic quality differences to the algorithm, which bids accordingly and compounds them.
• Legacy geographic and device modifiers interact with automation — manual adjustments left from an earlier era can pull against the strategy, and should be audited rather than assumed harmless.

Brand mix shift reported as performance improvement

The most common false improvement in paid search is a shift in the balance between brand and non-brand traffic. Brand queries convert at high rates and cost little, because the searcher already intends to buy from you. If the proportion of brand traffic rises, either because a television campaign ran or because non-brand bidding was quietly pulled back, blended return on ad spend improves without a single incremental sale. Reported as a headline number, that looks like excellent work. Reported with brand and non-brand separated, it looks like what it is.

Insist on segmented reporting as a standing requirement rather than a diagnostic you request when a number looks odd. Brand and non-brand should be separately identified in the account structure so the split is a reporting dimension instead of a manual filter, and the ratio should appear in every performance summary beside the headline figures. The same logic applies to shopping traffic against text, and to existing customers against new ones wherever you can distinguish them. Any aggregate metric movable by mix needs its mix reported next to it.

Bringing GA4 audiences and CM360 Floodlights into search bidding

SA360 sits between your analytics and your ad server, and the integrations are the reason to use it rather than the native search interfaces. GA4 audiences can be imported and applied to search campaigns, letting you bid differently for people who have viewed a product, abandoned a basket, or purchased recently and are therefore worth less to acquire again. CM360 Floodlight activities can be imported as conversion actions, so the display and search sides of a program are measured against the same definition of a conversion rather than each counting its own.

Both need care. Imported audiences only work at meaningful scale, and a segment too small to matter will not move bidding at all, so ask which are actually large enough to have an effect. Floodlight conversions carry the counting methodology and lookback window configured in the ad server, so importing them without checking means search bidding inherits decisions made by whoever set up the Floodlight, possibly years ago for a different purpose. Ask which conversion source the bidding uses today, because accounts routinely run native and imported actions side by side.

Programs rarely stop at one product. Buyers hiring for Search Ads 360 often pair it with BigQuery partners , Campaign Manager 360 partners or Display & Video 360 partners , or review the whole Google Marketing Platform landscape before committing.