For many Ecommerce businesses, AdWords shopping campaigns can make up a large share of PPC traffic, and often a significant proportion of overall revenue. The importance of these campaigns can sometimes provoke paralysis on the part of digital marketing managers; it’s easy to adopt the attitude of “if it’s not broke, don’t fix it” when the prospect of even a small dip in performance can make a big difference in the bottom line. While such a thought-process is certainly understandable it can foreclose on growth and optimization opportunities. This is especially significant in a PPC landscape that is seeing more and more Ecommerce spend weighted towards shopping campaigns. In this blog post, I’ll make the case for testing Target ROAS bidding for AdWords shopping campaigns, and share some tips on how to successfully set up such a test.

What Is Target ROAS Bidding?

First, let’s establish what Target ROAS bidding is and how it works. AdWords describes Target ROAS bidding as “setting an average conversion value you’d like to get for each dollar you spend on your ads. With Target ROAS bidding, AdWords automatically sets bids to help get as much conversion value as possible at the target ROAS you set.” In other words, if you set a Target ROAS bid of 500% for a shopping campaign, AdWords will aim to produce $5.00 in revenue for every $1.00 spent, and maximize conversions within those parameters. In general, a higher Target ROAS setting would lead to lower volume, and a lower Target ROAS goal would lead to relatively more volume.

Note that ROAS is calculated simply as revenue/cost – if you report revenue as (revenue – cost) / cost, you’ll have to convert your ROAS goals to the simpler formula in order to set the target for the campaign.

A Brief Case Study

One of Hanapin’s clients, an Ecommerce business specializing in industrial supplies, had been seeing satisfactory results from their shopping campaign that was utilizing an Enhanced CPC bid strategy. The client wanted to increase volume and grow overall revenue, and after seeing success with automated strategies with other campaigns in their account, was amenable to testing Target ROAS bidding for the shopping campaign. They had previously used Enhanced CPC, a strategy that already makes use of limited automated bidding within a range of the manually set bids. I’ll explain more in depth how we set up the test later in this post, but wanted to preface the case for Target ROAS bidding in shopping campaigns with the results below:

The switch to Target ROAS bidding delivered outstanding results for the client. Not only did conversion volume and revenue grow by nearly 50%, the growth occurred without losing anything in the way of efficiency. In fact, ROAS actually improved by ~7%.

Granted, this is just one example of improved performance, and one could easily make the case that the results could be anomalous for a shopping campaign. I hope, though, that it does illustrate that there are significant gains possible when switching bid strategy to Target ROAS. At the very least, I hope it supports the case that rather than assuming Target ROAS will harm the performance of any one shopping campaign, testing the bid management switch may be warranted. Below, I’ll speak more generally as to why those managing shopping campaigns should give this strategy a shot.

The Case For Target ROAS Bidding

There are many, many factors that can affect the predicted revenue generated from a click on a shopping ad, among which are:

  • The product being advertised.
  • The location of the user.
  • The audience(s) that the user belongs to.
  • The device the user is searching on.
  • The time and/or day that the search is occurring.

Good digital marketers that are manually adjusting bids will evaluate each of these variables, and set bids and bid adjustments accordingly. The truth of the matter is, though, that while an advertiser can manually account for many of these variables, with enough historical data a machine learning algorithm is likely better equipped to evaluate them in their full complexity. This is especially true for companies with large shopping feeds. For example, manual bidding may be a reasonable time investment for a company with 250 SKUs, but that time investment can become much more burdensome for feeds that have 250,000 SKUs.

Perhaps you’ve had a bad experience with automation in the past, or are a skeptic of machine learning strategies in digital marketing. If you fall in that camp, consider the following:

  • Google’s machine learning algorithms change and improve over time, which suggests that even if you’ve seen bad performance in the past, it may be worth giving the machines another chance.
  • Even if the performance gains are marginal or non-existent, Target ROAS bidding frees the digital marketer to spend more time on other aspects of the campaign to improve performance, such as feed management, search query optimization, and maximizing profit through restructuring.
  • When you’re thinking back on failed tests, consider if you gave the experiment enough time to succeed. Testing takes time, especially when testing an automated strategy. More often than not, the algorithm can’t make decisions right away because there is no data to back it.
  • If the tests you had previously run were limited by budget (and that is no longer the case) it may be worth revisiting them – limited budgets mean limited data, and the test may not have had enough volume to arrive at a truly statistically significant result.

That said: Target ROAS bidding strategies are never guaranteed to be more effective than manual ones, and digital marketers should always view a bid strategy shift as a provisional test, not a set-it-and-forget-it switch. Additionally, tests such as these require some tolerance for risk, and should not necessarily be attempted for campaigns that are consistently exceeding goal if the business could not suffer even a small performance hit. One specific pitfall to keep in mind: if the target is set higher than historical performance, growth may be limited. Often in this situation marketers will initially see promising performance, but have difficulty growing the campaign.

Setting Up The Test

First, some bad news: AdWords does not allow for the creation of experimental shopping campaigns. There’s no denying that this is a major bummer, as it means that you won’t be able to split traffic evenly between a Target ROAS shopping campaign experiment and the status quo bid strategy. Instead, you’ll have to run the test sequentially rather than simultaneously. Keeping that in mind, here are a few tips on setting up your experiment successfully:

  • Set a reasonable target ROAS bid: The rule of thumb for campaigns that don’t have strict ROAS goals is to set your target ROAS bid at or just above the historical ROAS of the campaign. If, however, you seek to increase volume and the business has a tolerance for lower efficiency, you should set the ROAS lower. In the opposite case, set the target ROAS higher.
  • Run the test during a period of low seasonality: Because the test will have to be sequential, it is unavoidable that seasonality will pollute the results to some degree. Still, you can minimize this issue by choosing a period where seasonality is relatively low. For example, if your shopping campaign sees its most static performance during the summer months, this might be an ideal time to experiment with switching bid type. Also, keep in mind that what works in the off-season may not be ideal for the busy season. For example, Target-ROAS could be good for off-season/lower budget/efficiency, but during peak times it may be about site traffic and being aggressive, in which case optimize for clicks or conversions may be more ideal.
  • Decide on the test parameters and metrics beforehand: As with any type of experiment, it is important to know what a win and a loss looks like beforehand. Decide how long you’ll run the test and what constitutes a statistically-significant winning result prior to initiating the test. Otherwise, you could be left with ambiguous results.
  • Understand that the “learning” period may not be representative of long-term performance: When a campaign is switched to a Target ROAS bid management strategy, it will undergo a period of “learning” as it gathers data and sets bids. In my experience, this period tends to last ~1 week, although the length of this period will vary with volume.You may want to exclude the learning period from the experiment results when evaluating performance. Note: you can view whether or not a campaign is in the “learning” state in the settings tab.

Conclusion

If you are manually adjusting bids for a shopping campaign, I hope that this post at least has you considering whether or not a Target ROAS bidding strategy is worth testing. Not convinced? Or have you seen poor performance from applying a Target ROAS strategy to your shopping campaigns in the past? Chime in on Twitter @ppchero!