SMX Overtime: When to use automation to improve account performance
PPC expert Duane Brown answers questions from SMX Advanced about what works, and what doesn’t, in smart shopping campaigns, automated bidding and third-party automation tools.
Although a scheduled presenter for “Automation – The Next Generation,” PPC expert Duane Brown of Take Some Risk was ill and with much regret missed this session at SMX Advanced, he was happy to answer some questions from attendees as well as share a helpful resource for Google Ad scripts.
How long should it take before a third-party platform should be helping your account performance improve? When should we cut?
This is nothing different than running an A/B test on your ad account. You need to know what success looks like before you go into the process of picking a third-party tool. You also need to understand why you are using the tool (is it to help with your CPA or grow revenue). If you don’t know what success looks like, then you can not determine if you should cut a tool or not. To start, I would test a new tool. The first month shouldn’t be a huge change but as we get into months two and three… if I don’t see a performance boost in our account, I will start to ask why. Odds are it is time to cut. Any tool should ideally save you time and money while also growing your revenue.
If you work in an environment where frequent testing is the norm, how do you get over the fluctuations you’ll see due to the learning period of an automated bidding strategy and have stable performance quicker?
You could test less. Sometimes running six tests in a month is better than trying to run ten tests. The other option would be to figure out how you get more data. Most bidding strategies do better with more data to help Google understand what is going on. With any testing, you have to deal with fluctuations and accept that as part of doing business. Even when something fails, you would have learned something from the experiment.
What are your thoughts on smart shopping campaigns? And how to make it perform better?
Smart shopping is great when you have tons of data to back the algorithm. The challenge we face is that you grow revenue and maybe lower your CPA but then plateau in performance. Limiting your performance and growth potential is not always worth doing smart shopping. Also, the black box of data and ads showing outside your target region does not always make it the best option for an e-commerce brand.
It’s the end of the quarter and my boss says: “We need to increase conversions during this last week.” Now that I’m using Target CPA, I find I can’t as quickly ramp conversions. Any suggestions for getting faster results when using machine learning bid strategies?
If you are asking this question, does this mean it happens a lot? If so, it sounds like a communication and education issue. You are not always going to be able to ramp up performance as you see fit. Your boss should know they can’t just come to you and ask in the last week to make it rain for them, especially if another area of the business is hurting performance wise. I would be finding out sooner than the last week of the month if we need to see a higher performance.
How do you know the right third-party automation tool to pick? Are there any studies doing a side-by-side of the top services?
Most studies are either going to be biased and/or take into account nuances that may not apply to you when it comes to using any tool or even picking a rule or scripts. I always start with why am I doing this job and what am I hoping to get out of it. Then I look for tools that can help do the job and maintain the quality of work that we want. It’s also ok to only automate 70% of your work and still have a few manual steps. Automation does not need to be 100%.
What should we consider when determining the right attribution model? DDA does not provide any data.
The right attribution model should move your business in the right direction. There is no right attribution model and it’s a test-and-learn approach as you understand how your attribution model and using that data affects your business. This is not a set and forget it process either.
