Analyzing Google Shopping data and making the right decisions based on this data is a typical PPC Manager’s challenge. Everyone wants to improve the Key Performance Indicators (KPI) like the Return on Advertising Spend (ROAS). In this blog post, we will discuss how to avoid the most common ROAS optimization pitfalls.
- Avoid two common ROAS optimization pitfalls when analyzing your Google Shopping data
- Pitfall 1: The short-term biased view
- Pitfall 2: Cutting of the long tail
Let us introduce the following example: We start with a short-term view. Your 14 day’s report states $15 000 of costs, $105 000 of revenue and thus a ROAS of 7 as depicted below. As you can see, we segmented the “Total” row into separate rows for Converters (blue) and Non-converters (orange).
The “Total” row is the information you already had before. Now you can see that the costs for Non-converters are twice as high than the cost of Converters. It looks like you could increase your ROAS to a maximum of 21 by an elimination of costs of non-converting products – which is very much better than 7. But there are two pitfalls: you forgot about the long term and the long tail.
Pitfall 1: The short-term-only biased view
Let’s try to change our point of view and look at the non-converters as our “Opportunity budget” – meaning that we invest that amount to realize unpredictable conversions in the future. Still, the cost ratio of your Opportunity budget to your Converters is 2:1 in the short-term.
Now switch to a more long-term view, e.g. a 6 months’ period. Do you see the decreasing ROAS in the Converters’ row? The Converter’s ROAS is now down to 10.5. That is, because in the long-term different products of your long tail usually will also convert. Thus, the new cost ratio is now 2:1 in favor of the converters. You see, the cost ratio depends on the length of the observed time window!
The longer the period gets, the closer the ROAS of 21 gets to the Total ROAS of 7, in our example the long-term ROAS of the Converters is 10.5. Of course, the Converters’ ROAS can only reach the Total ROAS if you sold every item of your inventory at least once in the obtained period. Usually, that is not the case.
Pitfall 2: Cutting off the long tail
What would happen to the revenue if we cut the opportunity budget down to zero? As different products – apart from top-sellers – convert in each period, focusing on only one period is of course selection biased.
Now we look at the process in more detail: Every period, you have converters and non-converters. The converters can be top-sellers, which will bring in conversions in every period. But most of your items (usually thousands of products) will belong to the long tail. That means, they do not convert in every period but sometimes they convert. Yesterdays non-performers can convert today or tomorrow!
The influence of Opportunity Budget over several periods is depicted below. You can see moving items of different shapes from left to right and from right to left. Only top-sellers (the stars) stay in the converters part, because they sell every period. Circles and diamonds indicate the state of conversion in the previous period of each item. E.g. diamonds converted also in the previous period while circles didn’t. Top-sellers always convert. Usually, every period has items which don’t perform anymore and from the previous non-converters some items will convert. This exchange is indicated by the arrows. As you can see, the converting items vary but the revenue is constant.
If you simply cut down your Opportunity budget you eliminate the long tail as depicted below. When you do not invest in last periods’ Non-converters anymore, they don’t get a chance to convert this period, next period or anytime in the future. Thus, after a few periods, only your top-sellers will be left. Your overall ROAS could improve dramatically but your revenue will decrease in the same manner! Of course, this is only true if you have products apart from top-sellers, which is the case for most online shops. A typical long tail is 80% or more of the inventory while real top-sellers are less than 1%.
Optimization of the ROAS over the whole product assortment is a very important task and is more complex as it looks at first sight. There are top-sellers which convert in every period. Apart from that, there are items which don’t convert each period, e.g. converts just once or twice a year. We call these items the long tail.
Adjusted bids will optimize the long tail instead of eliminating it and still save money and improve the ROAS. The clue is to set your bids item wise and to include historical performance to calculate the amount you’re willing to pay for each item. Of course, the more products you have, the more this becomes a task for an automated bidding software rather than a human being.
Stay tuned for more interesting news from our “Whoop!” Data Science department!