Smart BiddingTeaching the machine what matters

What smart bidding does#

In manual bidding, you set a maximum price per click for each keyword and adjust it from time to time. In smart bidding, the system sets a different bid for every single auction, based on its prediction of how likely that particular search, by that particular person, in that particular context, is to lead to a conversion and how valuable that conversion is likely to be.

It considers signals such as device, location, time, language, browser, the query itself, audience membership and many more, in combinations no human could manage. That is its genuine advantage. Its genuine weakness is that it optimises only toward what it can see: the conversions you track and the values you attach to them.

An algorithm will faithfully pursue the wrong goal; the strategist's job is to give it the right one.

The main strategies#

Google groups its automated strategies around two questions: are you optimising for the number of conversions or for their value, and do you want to set a target or simply maximise within budget? Strategy names and options evolve, so check current documentation, but the logic is stable.

There are also click-focused strategies, which aim for traffic rather than conversions, and impression share strategies, which aim for visibility on specific searches. They have niche uses, such as defending brand terms, but they do not optimise toward business outcomes.

Choosing between them#

The deciding factor is usually whether your conversions are worth roughly the same amount. If they are, cost-based strategies are simpler. If they vary a lot, as with product baskets of different sizes or leads of different quality, value-based strategies let the system favour the valuable ones. Our ROAS versus CPA guide explores this choice.

What the algorithm needs from you#

• A conversion that reflects business value. If you count newsletter sign-ups as equal to sales, the system will buy newsletter sign-ups.
• Values where outcomes differ. Even estimated values are better than none.
• Enough volume. The system learns from conversions. Low-volume campaigns learn slowly; consolidating campaigns or optimising for a more frequent event can help.
• Timely data. For lead businesses, importing qualified leads or sales, as described in offline conversion tracking, turns lead volume into lead quality.
• Stability. Big changes to targets, budgets or conversion settings restart learning.

Setting targets without strangling the campaign#

A target is a constraint. Set it too tight and the system bids only on the auctions it is most certain about, which can collapse volume. Set it too loose and it will happily spend on marginal traffic. The practical approach is to start near what the campaign has actually achieved recently, then move the target in small steps, allowing time between changes for the system to adjust.

Work out in advance what you can afford. A target CPA should come from the economics of the business: the margin on a sale, the lifetime value of a customer, the close rate of a lead. It should not come from a number someone saw in a report about another company.

The learning period#

When you launch a smart bidding strategy or change it significantly, Google shows a learning status while the system calibrates. Performance can be uneven during this time. Judge nothing in the first days, avoid stacking changes, and wait for at least one full conversion cycle, longer if customers typically take weeks to convert after clicking.

Seasonal events, such as festive sales, can confuse the system because conversion rates jump temporarily. Google offers tools to tell the system about expected short-term changes; check current documentation for seasonality adjustments and data exclusions, which are also useful when tracking breaks.

When smart bidding goes wrong#

The most common failure is broken or inflated tracking. A tag that fires twice, a thank-you page that loads on refresh, or a micro-conversion set as primary can all teach the system to chase nothing. When results change suddenly, check tracking before blaming the algorithm.

Smart bidding and the rest of the account#

Automated bidding works best alongside broader keyword match types, consolidated structures and plenty of creative variety, because all of these give the system more room to find value. It is also the engine behind Performance Max. The human role shifts from setting bids to setting goals, protecting data quality and judging whether the business, not just the dashboard, is better off.

Portfolio strategies and shared budgets#

Google lets you apply a bid strategy across several campaigns as a portfolio, pooling their conversion data so the system can learn faster and balance results across them. This can help low-volume campaigns that would struggle alone. It also means individual campaigns may run above or below the target while the portfolio as a whole meets it, so judge portfolios on their combined results. Check current documentation for how portfolios interact with shared budgets.

A practical sequence#

1. 01Fix conversion tracking so the primary conversion reflects real value.
2. 02Add conversion values, even estimated ones, where outcomes differ.
3. 03Start with a maximise strategy to gather data, if history is thin.
4. 04Move to a target once performance is stable, set near actual results.
5. 05Adjust targets gradually and allow time between changes.
6. 06Import offline outcomes to teach the system what quality looks like.
7. 07Review search terms and placements regularly; automation does not remove the need.

Published by Fabulous.Media, a network of specialist marketing agencies. Updated 9 October 2026. Platform features change often; check current official documentation before acting on platform-specific detail.