Calculating Optimal Bids with Irregular Data and Conversion Latency

In a previous post, we explored the effects of seasonality and cyclical trends on revenue-per-click (RPC) and conversion rate. Today, we’ll take a look at how identifying and excluding irregular or outlying data, and accounting for conversion latency, are critical to calculating optimal bids and maintaining control over revenue outcomes.

Seasonality and Outlying Performance

During the holiday shopping season, RPC and conversion rates can double in the months leading up to mid-December, and drop dramatically thereafter. Leveraging data during these periods of irregular paid search performance can result in suboptimal bid calculations. In order to factor these types of performance shifts into their bidding strategy, search marketers must first identify outlying and irregular data and subsequently exclude those dates or date ranges from bid calculations. Using advanced filters and alerts, search marketers can manage their data by exception and quickly identify the extent to which seasonality or cyclical behavior has impacted paid search performance.

For example, you might create an alert to notify you when RPC or conversion rate has increased by more than 50% of the average over the last three days. It’s possible that a new promotion or period of seasonality is causing a significant shift in performance. Excluding dates with outlying RPC or conversion rates will prevent calculations from inflating bids even as performance returns to normal.

Conversion Latency

Looking past seasonality and cyclical trends, date exclusions are also critical in accounting for conversion latency—the time between an initial ad click and an eventual conversion or revenue. Conversion latency varies across industries and product lines, ranging from same session to several months. For certain businesses with high consideration products or services, conversions and revenue can go unattributed to click and cost data for extended periods of time. As a result, bid calculations that leverage these periods of incomplete data fail to maximize performance. To address conversion latency, search marketers need the ability to exclude the most recent days from bid calculations.

For instance, if it typically takes two days for a customer to complete a purchase after clicking on a paid search ad, a sound bidding strategy would exclude the last two days from bid calculations. This would ensure that bids aren't being calculated using click and cost data that would otherwise have revenue attributed to it after two days. Dynamically extending or shortening a rolling exclusion window (in the example above, it would be a two day rolling date exclusion), depending on business needs, enables search marketers to calculate optimal bids based on a complete picture of paid search performance.

Informed vs. Reactive Bidding

For some businesses, high conversion latency can often warrant leveraging a lengthy rolling date exclusion. However, to remain competitive and respond quickly to shifts in the bidding landscape, carefully consider how long of an exclusion window is used. For example, let’s pretend that a business needs to wait sixty days until 98% of their revenue is attributed back to their paid search clicks. With a sixty day rolling date exclusion, it would require them to wait nearly two months before making informed bid calculations. Due to this length of time, they would undoubtedly fail to capitalize on immediate revenue opportunities. On the other hand, let’s assume that the same business can attribute 80% of their revenue after seven days. Using a seven day rolling date exclusion, they could still calculate informed bids while remaining reactive to the current bidding landscape.