AI / ML
Fine-Tuning
Training a pre-existing LLM on domain-specific data to improve its performance for particular tasks. In a marketing context that domain data is typically historical performance, attribution patterns and creative metadata. The intended gain is discrimination a general model lacks: telling creative fatigue apart from audience saturation when both present as the same ROAS decline. Fine-tuning is not free, and it is often the wrong tool where retrieval over current data would answer the question better.
Why it matters
It buys domain fluency at the cost of a training cycle, and it is frequently the wrong tool where retrieval over current data would answer the question better.
In practice
Reach for it when you need consistent form or vocabulary. Reach for retrieval when you need current facts.
Related terms