Definition
The learning phase is the period at the start of a new or significantly edited automated ad campaign during which the platform's optimisation algorithm gathers data and calibrates delivery before it can perform stably. During this phase, results are volatile and often worse than the campaign will eventually achieve, and cost per outcome swings widely.
Platforms typically define exit from the learning phase as reaching a threshold number of optimisation events (for example a set number of conversions) within a recent window; until then the campaign is "learning".
The practical implications are that a campaign should not be judged or heavily changed during the learning phase, that edits reset it, and that a campaign optimised toward a rare event may never gather enough of them to exit — which is a signal to optimise toward an earlier, more frequent event instead.
In context
For iGaming media buyers, the learning phase interacts badly with the vertical's long conversion lag. A campaign optimised toward first deposits may generate FTDs too infrequently to exit the learning phase in a reasonable time, so delivery stays unstable and expensive.
The common solution is to optimise toward an earlier, higher-frequency event — registration, or a qualified landing-page action — that correlates with deposits, gather enough of those to stabilise, and use value rules or a conversions API feeding deposit value to steer quality. Buyers also budget for the learning-phase cost as a known upfront expense rather than reacting to it as a failure.
The discipline points are: do not make edits during the learning phase (each meaningful change restarts it and wastes the spend already invested in learning); give the campaign enough budget to reach the event threshold in the platform's window rather than starving it into permanent learning; consolidate rather than fragment (many tiny ad sets each never exit learning); and judge the campaign only after it has stabilised. Compliance still applies from the first impression — the learning phase is not a grace period for creative or targeting rules.
For affiliates buying their own media, understanding the learning phase prevents the common mistake of killing or overhauling a campaign in its first days based on volatile data that does not reflect its steady-state performance.
Worked example
A buyer's casino campaign optimised to FTD cannot exit the learning phase because deposits are too infrequent. It switches optimisation to registration (which correlates with deposits) plus a conversions API feeding deposit value, gives each ad set enough budget to hit the event threshold, and stops editing.
The campaign stabilises and cost per FTD settles below target.
Related terms
Frequently asked questions
Browse the full iGaming & affiliate glossary — hundreds of EN/RU terms with examples.
← Back to glossary