Every year, PPC managers are faced with the same situation: as soon as the weather warms up, consumers tend to immediately stop buying winter clothes. As a result, bid management inevitably hits a wall for certain periods of time.
This is because many bid management algorithms are derived from retrospective performance data. Calculating optimal bids based on recent performance is a powerful approach for most bid management scenarios. However, retrospective bidding will lead to a dead end when consumer demand changes quickly. In the end, past performance drivers may remain in hibernation and advertisers are doomed to miss out on potential profits once seasonal conditions change – unless these keywords see a significant increase in their Max CPCs.
Against this background, I will describe to you two simple procedures that have already helped one of our clients in the fashion industry to escape this seasonal trap after updating their product range to spring/summer. These steps will give you a clear idea of how to test maximum bids (Max CPCs) based on last year’s performance data and how to do the necessary fine-tuning of the bids during the testing period.
Keyword Karma: Past Performance Data for Present-Day Profits
Analysing campaigns across different markets and industries, we’ve discovered that seasonal changes cause significant differences in performance levels across different brands and product categories.
The good news, though, is that if your account has gone through seasonal shifts in recent years, looking at that data will allow you to anticipate potential returns within the product categories that you advertise. Using these powerful insights, you will have the chance to boost your performance drivers and get a head start. Collecting data for retrospective bidding, in contrast, may turn into a lengthy process that would leave you with undesirable opportunity costs.
Step 1: Keyword Reincarnation – Inject Predictive Data into your Bids
In order to adjust the overall bid configuration of our account to these seasonal shifts, we aggregated historical performance data for keywords by their semantic attributes. By semantic attributes, we mean clusters of keywords that advertise similar products and contain shared attributes such as brand, category, gender, or buy words.
Examples of semantic clusters:
Once you have identified clusters with a promising earnings per click (EPC = margin / clicks), you can apply these insights to price those keywords. The following example for “Generic # skinny jeans # [No Gender] # [No Buy Words]” illustrates the procedure:
In the example above, we looked at historical performance data for the next eight weeks as of today. We then calculated the Cluster EPC for the semantic cluster of “Generic # skinny jeans # [No Gender] # [No Buy Words]”. This allowed us to anticipate the potential EPCs for the coming eight weeks (0.93€).
To play it safe, we decided to deduct a certain percentage from each match type. As an additional safe guard, we advise you to stay within an anticipated ROI range that is suitable for your project. In our case, the targeted ROI (= (EPC – CPC) / CPC) for testing seasonal bids was 50%.
Finally, the remaining adjusted EPC helped us decide how much to increase our Max CPCs for each keyword.
Step 2: Managing a Seasonal Turnaround by Fine-Tuning Your Bids
After increasing your Max CPCs, you may need to wait some 48 to 72 hours before starting to fine-tune your bids. Keep in mind that the bid of some of your keywords may still be too low to trigger ads. For this reason, we repeatedly revised our targeted keywords and every 24 to 48 hours we gradually increased the bids by +10 to +15% for those that did not generate impressions and clicks.
Once your keyword testing gathers momentum, you will need to direct the oomph. At this point, the goal is to single out the most extreme outliers and to reduce their bids in order to avoid unnecessary losses. This can be achieved by running your bid management under softer conditions than usual. We “softened” our bid management by running it at a cancellation rate of only 15%(compared to the usual account average of -57%). Then, we chose to down-price the following keywords:
- Keywords with a last average CPC during the testing period
- AND keywords identified for down-pricing even under “softer” bid management conditions, for instance, working with fractions of the usual conversion rate.
While this procedure ensures that the most extreme outliers, for example those keywords that are underperforming under softer bid management criteria, are phased out, it gives keywords that are still being tested the benefit of the doubt. We recommend repeating this procedure every 24 to 48 hours while gradually moving towards your standard bid management configuration, for example, raising your applied cancellation rate by 10% with each repetition.
Having done both steps, you will see the numbers improve:
I summed up the approach again in a more chronological order:
- Day 1: After shortlisting promising semantic clusters based on overall performance data, we increased the Max CPCs of those keywords that proved relevant in reference to the anticipated EPCs for the coming eight weeks. To play it safe, we buffered keywords by their match type at -15% (exact match), -20% (phrase match), -30% (broad match modified), and -50% (broad match). We increased the bids towards an anticipated ROI of 50% as an additional safeguard.
- Day 2 to 3: After 48 hours, we had to boost some Max CPCs by a further +15% in order to generate additional traffic for our targeted keywords.
- Day 4 to 9: We started fine-tuning the bids by down-pricing keywords with a last average CPC after calculating the bids based on a cancellation rate of -15%. Similarly, we applied increased cancellation rates of -35% on day seven and -45% on day eight.
The trend looks like the following:
Seasonal changes across different markets and industries cause significant differences in performance levels across different brands and product categories. Therefore, we recommend that you analyse and visualise keyword clusters and their performance levels over the year and apply this knowledge to your bids when retrospective bidding doesn’t result in optimal Max CPCs.
The test results above show that it pays off to consider all of the areas in which you expect positive seasonal trends. In fact, predictive bidding reactivated keywords that “hibernated” throughout the winter season and allowed us to reallocate our campaign budget to keyword clusters that more accurately represent shifting seasonal consumer demands. This is indicated by increases in conversion rates, ROI, and overall click volume that begin with the bids’ fine-tuning and gather momentum with late conversions.
Waiting to collect retrospective bid management data would have delayed this process considerably and would have caused a decline in AdWords Quality Scores compared to those of competitors targeting the same keywords. Thus, being pro-active in times of quickly changing consumer demand allows you to stay ahead of the competition.
What are your experiences with predictive bid management? Looking forward to discuss with you in the comments section.