Traditionally, search marketers base their search results on a “last click wins” basis. This means that the last click a consumer makes always gets attributed the sales revenue or conversion, regardless of how many other searches were made prior. The result is that brand terms often appear hugely profitable and costly generic terms appear to offer an extremely low ROI, if any at all. This makes it difficult to correctly classify “head” vs “tail” terms.
To combat this discrepancy, whenever we estimate performance for a keyword, we also calculate a confidence interval related to that prediction. When the confidence interval is too large, it means the prediction is useless (typical for keywords with very low traffic). We then need to aggregate in a relevant way (which is usually different from the way keywords are structured in ad groups) to get a critical mass of stats.
For this reason, we offer two different algorithms for automated bid management:
- Long-tail (tail): analyzes a bucket of clicks from a given set of terms and makes bid changes based on statistical significance.
- Core (head): analyzes past performance by week, day, and hour, making bid decisions based on a variety of statistically significant externalities.
Our automated bid management analyzes traffic patterns observed at similar times throughout an account’s history and makes preemptive bid decisions to effectively anticipate consumer behavior and minimize CPCs which effectively maximizes ROI. Automated strategies base decisions upon the smallest possible set of terms. If a given KW has enough data, then the decision will be based strictly on that term’s data set. The data set is expanded until enough data is available to make a statistically significant decision. The first expansion is to the ad group level, and then to the sub-category level in terms of the KWs portfolio group.
Jacqueline Brown
