Smart Bidding, attribution, and PMax, explained plainly.

This is the more technical companion to the "then and now" post: less history, more of what actually goes into a well-structured account today.

Smart Bidding, and which kind to use

Target CPA optimizes for conversion volume at or below a set cost per conversion: the right choice when every conversion is worth roughly the same, like a lead form. Target ROAS optimizes for revenue instead, which matters when conversion values actually differ, like an e-commerce cart. Maximize Conversions (or Conversion Value) just spends the budget as effectively as possible without a specific cost or return target, often the right starting point before enough data exists to set a realistic CPA or ROAS goal.

Layering audiences onto Search

Adding audience signals to a standard Search campaign doesn't restrict who sees the ad. Used as an "Observation," it lets bids adjust up or down based on how likely someone is to convert. In-Market audiences are Google's own list of people showing active purchase intent in a category. RLSA (Remarketing Lists for Search Ads) adjusts bids for people who've already visited the site; a common move is bidding meaningfully higher for cart-abandoners specifically. Custom Intent audiences let you build a list based on specific search terms or competitor sites people have shown interest in.

Ad copy that actually gets tested

Responsive Search Ads are the standard format now: up to 15 headlines and 4 descriptions, with Google's system testing combinations to find what performs best for each search. The real skill here is writing meaningfully different headlines, not fifteen versions of the same sentence; the algorithm can't find a winning combination if there's nothing to meaningfully differentiate.

Dynamic Keyword Insertion swaps ad text to match a searcher's exact query. Used well, it makes an ad feel specifically relevant. Used carelessly, without proper fallback text or on messy keyword lists, it produces ads that read as broken or badly grammatical, which looks worse than not using it at all.

Attribution: why "last click" undersells the full picture

The old default model gives 100% of the credit for a conversion to the final click, ignoring everything that led up to it. Data-Driven Attribution uses the account's own conversion history to split credit across the whole path instead, which matters most for keywords and campaigns that tend to introduce someone to a brand rather than close the sale directly. Last-click accounting makes that early-funnel work look worthless when it usually isn't.

Enhanced Conversions sends hashed first-party data (like a hashed email address) back to Google alongside a conversion, which improves tracking accuracy in a world where browser-level tracking keeps getting more restricted.

What Performance Max actually needs to work

Three things determine whether PMax works, and skipping any of them is why some campaigns underperform: clear, accurately tracked conversion goals; a properly varied set of creative assets (text, image, and video, not just what's easiest to reuse); and audience signals to give the algorithm a sensible starting point instead of learning from nothing.