During day-to-day work with Google Ads, it’s easy to forget we work with the biggest database in human history powered by the state of the art machine learning algorithms. Considering that, I think it’s pretty important to make sure how this system works and what you can do to make the most of it.
If you’re an advanced PPC specialist with millions of ad spends and years of experience, you may think you know everything you need to know.
But I suggest to at least skim through this article because the system changes all the time and I dare to say that all of the experiences we gathered 6 years ago is now almost useless.
- Feed the algorithms with data.
- Use the right attribution model.
- Remember about cross-device paths.
- Consolidate everything you can.
- Exchange data with other channels.
- Do not underestimate Optimization Score.
- Forget about CTR-focused ads analysis.
- Be patient.
- Consider the profit margin.
- Consider seasonality.
What is Smart Bidding?
Just to be sure we are all on the same page – Google Ads is an auction system. The more you pay and the more relevant your ad is – the more often your ad will be visible.
You can set the cost per click manually for every keyword. And for example, adjust the price so you’d pay 30% less during the weekends.
However, a few years ago Google introduced automatic bid strategies that use machine learning to optimize the ads for your goal. Usually – conversions or conversion value.
These strategies became super powerful and are still being improved. That’s why I decided to share a few insights and hints on how to make sure your advertising strategy is coherent with smart bidding. I hope it will be valuable to you.
1. Feed the algorithms with data
I can’t stress how important this is for the end-goal.
I think we can all agree that even if you’re a leader in your industry, it’s usually a waste of money to get a 100% impression share for all of your keywords. You want to maximize your conversions, so you’d like to reach the perfect customers with the highest chance to convert among all of the users who use particular keywords.
To find your perfect customers, Google analyzes more than 71 million signals about the user and your business.
Signals may be: time of the day, time of the month, device, browser, user’s search history, age, gender, interests, etc., so there’s no way you can beat it manually. But you can feed it with additional data to make it even more powerful.
Make sure you have your conversion settings measured properly.
Make sure you have enough conversions per day and per campaign for algorithms to learn.
If you don’t have enough conversions – take a step back and instead of tracking sales, track leads.
If you don’t have enough leads, maybe you need to take another step back and optimize for engagement and so on.
You also want to add as many remarketing lists as possible – website visitors, app users, e-mail addresses, youtube video viewers, etc. These audience lists will also help to optimize your campaign. Even if you’re not using them for your targeting, keep them on the account. It will help!
2. Use the right attribution model
Users’ paths may be complicated and to analyze them properly and get insights you should use some advanced analysis models like the Markov chain or similar. But the good news is, even if you have no idea how to do it – Google can do it for you if you have enough data.
There are a lot of articles written on this topic already, so I must simplify it and focus on the essence here.
Long story short, a data-driven attribution model will be the best solution usually. If you can’t use it – anything besides the last click makes sense.
Why? Because all of the other models analyze the whole user path. And the last click ignores everything besides, well, the last click. So this model won’t use most of the data we have. And using the data is crucial.
As you can see the last-click model is really mean to most of your data, so don’t use it.
3. Remember about cross-device paths
As we established – the more data you have, the better the algorithms will work.
Unfortunately for data analysts and luckily for the rest of us, connecting all of the users’ activity is not so simple. The well-known scenario is: people often research on mobile and buy on desktop. It’s probably not so common today compared to a few years ago, but still, it messes things up.
If your campaign goal is to get as many conversions as possible and your mobile conversions are lower because people buy from you on a desktop, then the mobile will be underrated and the smart bidding may not work properly. How to deal with this problem?
- Find differences between the CR between devices. Why is the conversion rate lower? Can you improve it? Can you make the path simpler for your mobile users?
- Find the real meaning of mobile on your conversion path. Try Google Signals or Facebook Attribution. It’s not the time to go into details now but these tools can help you see the real mobile conversions.
- If, based on your analysis, you’ve decided that your mobile campaigns need more attention – create a separate campaign for them. If not – you can check the box and go to the next point.
4. Consolidate everything you can
SKAGs (Single Keyword AdGroups) account structure is very popular, but I dare to say it’s not the best solution today.
The most important job we have is to give our users relevant and consistent experience for every keyword in the campaign. So, if you have one keyword, one ad, and one landing page in the ad group, you can easily make sure they match. That’s the clear advantage of SKAGs.
But there’s a more profitable approach – whenever there’s a potential to merge ad groups and still provide a great experience – do it.
The fewer ad groups, campaigns, and accounts you have – the better. The more conversions per campaign you have – the faster they will learn and make a profit.
These are the official Google requirements. You can start with 0 but don’t expect good results. Even 30 per month is not enough in my experience. That’s why I suggest consolidation.
5. Exchange data with other channels
I hope by now you get the idea which direction I suggest to follow.
The more high-quality data you have – the better your campaign will be.
Think about all of the data in your company’s every department you can use to optimize your Google campaigns. If you have plenty of customers every month and your goal is a direct sale – that’s easy, just focus on customers. But usually, it’s more complicated and every lead is different.
- Maybe you can upload the data from your email provider to focus on the people who engaged in your newsletters.
- Maybe you can use a lead scoring to let Google know what is your perfect audience.
- Maybe create a lookalike audience from the top 20% of your customers based on LTV.
The possibilities are endless and different for every business. Using machine learning saved you some time on bid management – use it for data management.
This is the data management landscape in a typical company. Data silos. Try to connect them, even if it’s not your job initially.
6. Do not underestimate Optimization Score
Lots of Google Ads managers tend to disregard the recommendation called Optimization Score. (If you haven’t seen it just click on “Recommendations” in a left-hand panel). And I’m not surprised. A lot of these automated suggestions sometimes just don’t make sense. But there’s a good reason you should keep an eye on them.
A smart-bidding algorithm is more and more like a black-box. It will do the work but you won’t be 100% sure how. Optimization Score gives you hints on how Google analyzes your campaign and when it thinks there’s a potential to grow. Even if it’s suggestion is not good enough for implementation – look around it and think about what it means.
For example, if Google suggests adding a new keyword – there’s a fair chance you already knew that. But you don’t have to use it right away. Instead, think about your account structure, match types, and keywords cannibalization and maybe it will make sense to add some variation of this keyword.
Think about the users’ intent behind this keyword. Why did the algorithm suggest it for your campaign? Maybe there are other keywords for this segment? Maybe you already have similar keywords in your search terms report? Think about these suggestions even if it’s not an instant win.
Besides, for now, the optimization score is just a hint for the advertisers. But based on the recent changes Google made, I wouldn’t be surprised if it will be directly impacting results and costs in the future.
7. Forget about CTR-focused ads analysis
Don’t get me wrong. Of course, you want people to click on your ads, but it’s not the most important factor right now.
The old-school approach was to create 3-4 ads, let them work for a few weeks, and then pause the worst-performing ad based on the CTR. It won’t work in 2020.
You may have two identical search queries, but they are truly different. Why? Because there are two different people behind them and they can be on a different part of the sales funnel.
One of your ads may have a great CTR, but it works only in 10% of cases and the other one may have the lower CTR, but it’s relevant for 90% of queries. It’s a very common scenario.
So based on the CTR, we should pause the second ad. Our CTR will go high and we can all shake hands and high-fives. But in reality, when the less-relevant ad will get more impressions, the average CTR will go down.
The solution is to keep as many different ad texts as possible (all of them have to be relevant for the user of course) and let Google handle it.
You can use responsive ads for this purpose, but the regular ads will also do the trick. After a while, you can analyze them, look for insights, and add more texts if you’ll have more ideas. But there’s no point in excluding ads as long as they make sense. They won’t hurt you. They can only help.
8. Be patient
I remember when the maximum conversion strategy was introduced. My Google account manager strongly recommended to use it. So I did. When my costs and CPA increased by 100%, I asked what’s going on and I’ve heard I needed to wait 4 to 6 weeks until it would work.
As you can assume, I refused politely, got back to the manual CPC strategy, and forgot about it.
But today’s algorithms are much better and if you have enough data, they can learn faster. Upload as many conversions as possible – even historical data using Offline Conversion Tracking. This way, you’ll reduce the learning period but even then, you should wait at least a week before making drastic changes.
It’s the marathon, not the sprint. Patience will pay off in the long term.
9. Consider the profit margin
If you have more than one product or pricing plan, make sure you include this information in your conversion tracking. Let’s take a look at the following example.
We have two products: a mountain bike and a bike helmet. The bike is obviously more expensive, but the profit margin is higher. The helmet is cheaper and the margin is lower.
If we’ll ask Google to maximize our ROAS, we may end up selling tons of helmets and get $0 profit. Where a product with a lower ROAS (a bike) would bring us a higher profit. It’s a trap!
Right now Google doesn’t have a “Maximize profit” bidding strategy, but you can manually manage campaigns to make sure your most important products are in the loop. You can also try to send the profit-margin data to Google instead of your full price.
And the same goes for SaaS products with different pricing and different lifetime value. If you won’t include this data – you may end up with more leads but lower LTV instead of potentially more-profitable fewer leads with skyrocketing LTV.
10. Consider seasonality
The other thing you have to consider using smart bidding is seasonality.
If you plan to do some big changes like the Christmas sale, black Friday, Cyber Monday, etc. – algorithms won’t be prepared for that. And they are very sensitive to extreme changes (who isn’t).
Luckily there’s a feature created especially for this purpose so you don’t have to look for walkarounds. Look under “Shared Library”, find “Bid Strategies” and then click “Advanced Controls”.
Here you’ll be able to create a seasonality adjustment to inform the algorithms i.e. you expect the 50% higher conversion rates for the next 3 days. But don’t worry if you don’t deliver it. You’ll spend a little more, but they won’t judge you. They are robots.
The conclusion is: we can all focus more on the creative part and deliver a great experience for users. AI will help us with the rest. Actually, most of the above points are also true for Facebook Ads but the devil is in the detail and it’s a topic for another time.
If you liked it, you can find the shareable summary below. Thanks for reading and I wish you high conversion rates!