A few months ago, AdWords quietly rolled out a new flexible bid strategy for Shopping that some of us users geeked out over: Optimize by ROAS.
Previously, the closest anything had come to that was Optimize by CPA. While that was great for lead gen, it left a lot to be desired in ecommerce. Typically in ecommerce, you’re dealing with a lot of different products that ultimately carry a different monetary value or profit for the retailer, and that has a trickle-down effect. Some keywords lead to higher dollar-value purchases, so there’s a certain tolerance for higher CPAs. Some other keywords tend to be a “one and done” purchase, so they’re less valuable and thus a lower CPA is wanted.
For example, let’s say you are an electronics retailer. You sell TVs, but you know that your average order value is higher than just the TV, because users buy a bracket to hang it, fancy cables, and things like that. In other words, they tend to buy add-ons that increase the basket value beyond just the initial TV model they searched for. (And hopefully you have a great check-out process that encourages that type of behavior.)
On the other hand, let’s say you’re a retailer that sells t-shirts with movie slogans on them. Users search, purchase the one t-shirt, or maybe add on one other. There’s less cushion there in the margins to absorb a high CPA.
This creates a complex world if you try and live by CPA, because if you carry thousands of different products and they all behave differently, it’s difficult to manage all those different flexible bids. Ultimately the retailer cares about the return on each media dollar they spend, so return on ad spend is the guiding metric.
What is “Optimize by ROAS”?
Optimize by ROAS was first released in beta, and then rolled out to all accounts shortly thereafter. It’s considered a shared bid type, so here’s how you set it up:
In the lower left-hand margin of AdWords, click on “Shared Library”:
On the next screen, choose “Bid Strategies”:
Click the red button for adding a strategy, and you’ll see the one for targeting a return on ad spend:
The next screen will prompt you to choose the Campaign(s) you are wanting to apply this bid type to. If they have a conversion history, choosing a Campaign will populate a dialog at the bottom that recommends the ROAS you should shoot for, but the box to enter it is still free-form.
Once you choose that, there will be a “learning” phase that lasts a few days in AdWords. After that, you have basically relinquished your control of the bidding, and are letting Google do its thing to get the ROAS you asked for.
How Does It Do?
I tested this on two very different accounts to see how it would behave.
One of the things I was very conscious of was the transaction volume. In previous stints with the CPA Optimization option, I found Google would hit the CPA I asked for, but at the expense of volume. Naturally, I was curious if that would happen here, so the accounts I picked had different volumes: one had steady, high volume, and one was more sporadic but still with a long history.
Account #1: High Volume
This client’s highest-volume month is always October, which fell during part of the test period. To keep things on an even keel, I removed that and just looked at the weeks before and after, which have historically been the same in regards to user behavior. Here are the main metrics, noted as whether it was before I implemented Optimize by ROAS vs. After:
- ROAS before: 261.5%
- ROAS after: 279.8%
- Conversion rate before: 4.3%
- Conversion rate after: 4.4%
- Transactions per day before: 6
- Transactions per day after: 4
The targeted ROAS (which was suggested by AdWords) was 360%, so we obviously didn’t hit that mark. When I included their busy month of October, they were closer to it, but still never hit it.
So all in all, not a failure. Google’s way did squeeze some more ROAS out of the deal, though I did notice a small drop in volume.
Account #2: Lower Volume
This account had miffed me for months, because it behaved erratically. I was interested to see if choosing to Optimize by ROAS would help smooth out the hills and valleys I kept experiencing. There would be days with 1-2 transactions, and then some that were 5 or higher.
Unlike the other account, this was more of a “one and done” environment. It’s a practical medical product; people know what they’re searching for, they buy it, and they move on. It’s very transactional in nature, and there is nothing special or different about what this retailer carries vs. other retailers who carry the exact same brands. Therefore, the ROAS has always been low, around 120% when you averaged out several months, so there isn’t a lot of wiggle room on what it costs per transaction.
I don’t think I actually even need to list the stats on this one to show what happened:
Obviously, it didn’t behave any less erratically. I implemented the bid strategy towards the end of August, and while it randomly hit some home runs on ROAS, I think that was a stroke of luck as opposed to any strategic gains based on Google’s algorithms. Transactions took a completely nosedive, and we abandoned the tactic for this client.
What’s the Verdict?
I do think there’s merit in this flexible bid strategy, but two things need to be a reality for it to work:
- There’s steady volume and a long history. Account #1 had both of those, and Account #2 had neither. Account #2’s longer history wasn’t enough to make up for the erratic manner in which the conversions happened, despite having a long period of time to average them out.
- You have to gauge if there’s a volume concern. The drop in 2 a day for Account #1 may have a different impact on retailers. If it’s a small volume drop but their ROAS soars, it might be less of a concern. While this strategy did do better for the retailer on ROAS, it wasn’t really enough that it offset losing some volume. Your mileage will vary.
It’s certainly worth testing, but remember this is set at a Campaign level and not an Ad Group one, so choose carefully. I had reorganized Account #1’s Shopping setup previously and grouped the consistent High ROAS products together, so there was some tolerance in fluctuation. You might not want to test this on a Campaign that contributes massively to the bottom line if you rely on it as a steady performer.