Smart bidding learning period - what to expect and what not to do
The learning period is the phase after you make a significant change to a campaign where Google's bidding algorithm re-establishes its performance model. It is a frustrating time for most advertisers because performance can look worse before it improves.
When you launch a new campaign with smart bidding, switch bidding strategies, or make major structural changes, the campaign enters a learning period. Google flags this in the status column. During this time, the algorithm is testing different bid adjustments and building its model for your specific campaign. Performance can be inconsistent and often looks worse than usual.
How long the learning period lasts
Typically one to two weeks, though it can extend to up to six weeks for campaigns with low conversion volume. The more conversions your campaign generates per week, the faster the algorithm has data to refine its model and exit the learning phase. Campaigns getting fewer than 5 conversions a week can stay in learning for weeks. If that is your situation, the solution is to either broaden targeting to generate more signal or to use a less conversion-dependent bidding strategy until volume builds.
What triggers the learning period
Not just launching a new campaign. Changing bidding strategy, significantly adjusting your Target CPA or ROAS, adding or removing a large number of keywords, changing your ad rotation settings, or making major budget changes can all trigger a new learning period. This is important to understand because it means frequent changes actively hurt performance. Every time you reset the learning period, you delay the point at which the algorithm is working at full effectiveness.
What not to do during learning
The temptation is to intervene when performance looks poor during learning. Resist it. Making further changes mid-learning resets the clock and prevents the algorithm from completing its optimisation cycle. The exception is if something has genuinely gone wrong - a tracking error, a significant budget problem, or a landing page that is down. Fix real problems. Leave the bidding algorithm alone to work.
Setting realistic expectations
Tell your clients and stakeholders that the first two to four weeks of a new smart bidding campaign should not be judged against steady-state performance benchmarks. Set expectations upfront. The learning period is not a failure - it is a necessary phase. Agencies and advertisers who do not explain this end up making changes during learning that make things worse and then blame the strategy rather than their own impatience.
Found this useful?
Start a conversation - no pitch, no pressure.