Every day we feed data into our bid management system to give it the information it needs to calculate the right product and keyword bids. If a campaign’s recent conversion numbers deviate considerably from its long-term results, using the data could lead to suboptimal bid calculations. But data exclusion is not always the way to go when dealing with irregularities in campaign performance – so being able to differentiate between the two scenarios is critical: When is it time to clean out your data, and when are you better advised to leave temporary performance changes to your tool?
Extraordinary performance shifts can have positive or negative causes. Whether the performance shift is favorable for your business or not – as long as it’s not representative of the campaign’s long-term performance, this should not affect your decision to exclude data.Positive outliers: Discount campaigns
Allowing significant sales discounts on specific days, i.e. Black Friday or similar, may cause a sudden rush for your webshop and lead to crazy conversion rates. Even though such a high number of sales is great for your business, your data for the day is most probably not representative of the rest of the year – reason enough to exclude this date or date range from the keyword history.
Negative outliers: Tracking errors or lack of tracking
Tracking errors that cannot be restored at a later point: Some changes to the website can cause tracking pixels trouble and stop it from tracking conversions. As a result, the data would tell the bid management system that previously successful keywords don’t generate any conversions at the moment. The system would then start to lower bids for these keywords continually.
If it is not possible to get decent data for the time period in retrospect to make up for tracking errors, it is better to remove the faulty period from the record altogether.
Seasonal shifts: No need for excluding data
You might have observed similar highs and lows in campaign performance if your business is affected by weather conditions and seasonal shifts throughout the year. These changes tend to take effect over longer periods of time. November and December are among the most seasonal months in SEA, for instance. Looking at the graph for December below, it shows a steady increase in the conversion rate and a relatively sudden drop just before Christmas, but within a couple of days the tool was able to react. Data exclusions were not necessary: Bids went back to normal quickly, because performance data from the latest week influences daily bid adjustments the most. Giving the most weight to the most recent data allows our tool to react to shifts that take place over longer periods of time very quickly.
How can I exclude data from the feed?
If you are planning to offer a discount special on a particular day of the year that you know will lead to extraordinary conversion rates that are considerably higher than your usual numbers, simply inform us before it happens. We are able to exclude date ranges beforehand. Also, if you notice any erroneous data that you’d like us to exclude, please contact us. Together we will have a look at your campaigns and discuss the next steps.