Portfolio Bid Strategies

A portfolio bid strategy is created once at the account level (under Tools and Settings > Bid Strategies) and then attached to any number of campaigns. Once attached, Google no longer treats each campaign as its own optimization problem. It pools the conversion history, search volume, and auction signals from every campaign in the portfolio and sets bids to hit the shared target across the group as a whole, not within each campaign individually. This is the mechanical difference from applying the same Target CPA or Target ROAS strategy separately to five campaigns: separate strategies each learn and bid in isolation, while a portfolio shares one learning pool and one budget-allocation logic.

The setting that trips people up is the optional bid limit paired with the strategy type. A Target CPA portfolio still lets you set a minimum and maximum CPC bid ceiling per sub-strategy, and a Maximize Conversion Value portfolio can carry a target ROAS. Leave those blank and the algorithm has full freedom to move money toward whichever campaign in the group is converting most efficiently at that moment, even if that means a campaign you care about goes quiet for a few days. That reallocation is not a bug, it is the entire point of pooling, but it is also the most common complaint from advertisers who did not expect one campaign to get starved to fund another.

Portfolios matter most when individual campaigns are too small to give Smart Bidding enough conversion volume to learn from. Google's own guidance for exiting the learning phase assumes roughly 30 conversions per campaign in the last 30 days for Target CPA and Target ROAS; a portfolio can blend three campaigns at 12 conversions each into a pool of 36 and start bidding with a usable signal instead of three campaigns individually stuck flat-lining in learning. Portfolios matter less, and can actively hurt, when campaigns serve genuinely different goals - a lead-gen campaign and a branded-search campaign do not belong in the same CPA pool even if both happen to convert, because the algorithm cannot tell you meant to treat them differently.

The most common mistake is building a portfolio around campaigns with meaningfully different conversion values or funnel stages, then being surprised when spend concentrates in the cheapest one. A close second is setting a portfolio-wide CPA that was calculated as a simple average of each campaign's historical CPA rather than a volume-weighted one, which biases the target toward whichever campaign happened to have fewer conversions in the lookback window. A third is forgetting a portfolio exists at all after creating it, then troubleshooting an underperforming campaign for hours without realizing its bids are being set collectively, not individually.

In reporting, portfolio performance is visible in two places: the Bid Strategies report under Tools and Settings shows the aggregate CPA or ROAS the portfolio actually delivered against its target, along with a status column that flags whether it is limited by budget, limited by bid, or learning. At the campaign level, the standard reports still show each campaign's own conversions and cost, but comparing that campaign's individual CPA to the portfolio's blended target is a common error - a single campaign can run above or below the shared target indefinitely as long as the pool average holds, so judging one campaign against the portfolio number in isolation will produce false alarms.

Practically, start a portfolio only after you have decided the campaigns inside it genuinely share one business goal and one acceptable cost. If you want the pooled learning benefit without surrendering all control, set a bid limit range rather than leaving it open, and check the Bid Strategy report weekly rather than assuming Smart Bidding is silently doing the right thing.