Portfolio Bid Strategies: When Sharing a Target Helps (And When It Wrecks Performance)

I see two versions of this mistake constantly. Version one: every campaign in the account gets dumped into a single portfolio tROAS strategy because someone read that Smart Bidding needs volume to work. Version two: the account owner got burned once, panicked, and now every campaign runs on its own standalone target, forever. Both are lazy defaults, not decisions. Portfolio bidding is a tool with a specific mechanism, and once you understand that mechanism, the question of what to group stops being a guess.

A standalone tCPA or tROAS target optimizes one campaign against its own conversion history. A portfolio strategy pools the conversion data from every campaign assigned to it into a single model, then allocates budget and bid adjustments across those campaigns to hit the blended target as efficiently as possible — not to hit the target on each individual campaign equally. That last part is the one people miss. Google isn't trying to make every campaign in the portfolio perform at the target. It's trying to make the portfolio perform at the target, which means some campaigns get starved and others get fed, on purpose.

That's exactly why portfolios are powerful for low-volume campaigns and dangerous for mismatched ones. A campaign generating eight conversions a month is working with a statistically weak signal on its own. Fold it into a portfolio with similar campaigns generating a combined two hundred conversions, and the algorithm suddenly has enough data to make real decisions instead of noisy ones. That's the legitimate use case — pooling thin data into something the bidding model can actually learn from.

A portfolio doesn't make every campaign hit the target — it makes the blend hit the target, sometimes by quietly sacrificing one campaign for another.

The damage happens when you group campaigns that look similar on paper but behave differently in the funnel or the P&L. I've audited accounts where a prospecting campaign and a remarketing campaign shared one tROAS target. The remarketing campaign converts at three or four times the rate, so the portfolio logic quietly shifted budget toward it over weeks, starving prospecting of spend it needed to keep feeding the funnel. Three months later, remarketing volume collapses too, because there's no new traffic left to remarket to. Nobody touched a setting. The portfolio did exactly what it was told — hit the blended target — and broke the account doing it.

Margin differences cause the same failure in a different shape. If one product line carries 60% margin and another carries 15%, and both sit under one shared tROAS, the algorithm treats a dollar of revenue the same regardless of what it costs you to deliver. It will happily push spend toward the low-margin line if it converts more cheaply, hit your target ROAS on paper, and shrink your actual profit. This is the same root issue covered in Value-Based Bidding: Why Every Lead Isn't Worth the Same to Google — a shared target can't account for differences a human needs to flag first.

Before grouping any campaigns under one portfolio, I check three things, in order.

• Conversion volume: is each campaign individually generating enough conversions (generally 30+ a month) for Smart Bidding to have a stable signal on its own? If yes, there's less need to pool.
• Funnel stage: are the campaigns targeting roughly the same point in the buying journey — cold prospecting vs. warm remarketing vs. branded search all behave too differently to share a target safely.
• Margin and deal value: do the products, services, or lead types across these campaigns carry comparable value to the business? If one converts at a fraction of the profit of another, a shared ROAS target will optimize toward volume, not profit.

If a campaign fails any one of those three checks relative to the others in the proposed group, I run it standalone. If a cluster of campaigns passes all three — say, three similar prospecting campaigns for different product categories at comparable margin, each individually too thin on conversions to bid well alone — that's a legitimate portfolio candidate.

Worth noting: grouping decisions aren't permanent. I revisit them whenever a campaign's volume crosses the threshold where it no longer needs pooled data, or when margin structure changes on a product line. A campaign that needed a portfolio at launch often outgrows it within a few months and performs better split out on its own target once it has enough history.