So … your client’s campaign has launched on time and it’s been running for a couple weeks. Now your client wants to know what optimizations you have made. This can be challenging for junior buyers that have never been schooled on performance optimization.
Let’s take a look at some of the first places to look and what changes to make when it’s time to start optimizing your campaign. But before we go through some of the optimizations, I want to encourage you NOT to make all these changes as once. In fact, it’s much better to do one at a time and only one per week. That is not a hard and fast rule but it’s a good rule of thumb so you don’t “over-optimize” your campaign. DV360 is particularly sensitive to making too many changes and you can often see performance suffer as a result of doing too much.
Without further ado, let’s dive in:
A. Double-Check Your Conversion Pixel
This is a slightly odd list because it starts with ‘A’ instead of ‘1’ but that’s because all the other actions on this list are pointless if your conversion pixel is not set up correctly. Before you proceed to the rest of the list, confirm that:
- Your pixel is firing (you can use the Floodlight activity report to get a good view of all of your floodlight activities and conversions at once).
- Make sure that the correct floodlight activity is added to ALL of your line items.
- Make sure that your look back windows make sense (my recommendation: 30 day click / 1 day view through)
Lastly, be careful with adding more than one conversion to your line item. Remember that the bidding algo does not distinguish between one conversion from another … a conversion is a conversion is a conversion, whether it’s a page view, a scroll, or a sale. So if you have (for example) a floodlight activity created for an ‘add-to-cart’ action and a sale action, DV360 will value these two the same. If you want to score these actions differently, you should create a custom bidding algorithm.
1. Bidding Algorithm
It’s hard to pick one metric that has the most impact on your performance but I think bidding algorithm is the most natural place to start. Needless to say if you want to drive as many conversions as possible, you don’t want to choose an algorithm like ‘maximize viewable impressions’.
It may seem like there are a lot of bidding algorithms available but I think you’ll quickly see that there are only 2-3 that you are likely to use (Note: recent changes to the UI alter the bidding strategies you have available depending on the objective you have set in the Insertion Order).
If you have a performance campaign I recommend using Maximize Conversions 9 of 10 times. There are exceptions but you can be fairly certain that maximize conversions is the right bidding strategy. Even if you have a ROAS goal, I STILL recommend determining your LTV of a new acquired customer and backing into your CPA goal. We promise to have a more detailed deep-dive into the topic of bidding algos on a separate post.
But for now (assuming your goal is to drive some type of action, e.g. sales, downloads, leads, etc.), then use the maximize conversions algorithm. It is also preferable to use the setting ‘while prioritizing target CPA’. However, there are times when you will not have a choice and will have to use the recommended setting ‘while prioritzing my budget in full’.
IMPORTANT: I highly recommend when using maximize conversions (or any other ‘maximize’ bidding strategy), to put in a CPM ceiling so the system does not bid too high. You will see a check mark box marked ‘Do not exceed average CPM of $0.00 USD’. It’s entirely possible that the system could bid way over an acceptable amount so you want to put on a CPM ceiling. I recommend no higher than $6 for display and $20 for video.
These CPM limits are only recommended for performance campaigns; obviously don’t use them when you are trying to win impressions on premium sites. Another exception would be retargeting lines or lines with very small audiences. You might have to bid higher if your line items is targeting only a very small audience.
2. Inventory: What Exchanges Are You Buying On?
Often times, teams already have a prescribed set of exchanges or PMPs that they are targeting for a particular client, campaign or initiative. Adding or removing exchanges (or PMPs) is not an optimization lever I use very often, but it’s a good idea to come back to this section of your targeting and make sure you are bidding in the right auctions.
When reviewing other buyers’ campaigns, I often find some surprises here. Make sure that your buying in the exchanges that you intend to be in and that you’re don’t have any obscure inventory sources.
It’s easier to start by going to the reporting dashboard, and running a quick report of your media spend by exchange (the dimension is called ‘exchange’).
8. Performance by Device: Desktop vs Mobile
This is a personal preference but I like to start looking at my performance by dividing the media spend into mobile vs desktop. These two devices are completely different and the performance will be always be different (with few exceptions). In most cases, I like to separate desktop vs mobile on different line items (or maybe insertion orders), so I will know directly from the platform which one is performing best before pulling a single report.
If you do not have your line items broken out this way, then run a report of your campaign with the media broken out by device (the dimension is called ‘device’ in DV360 reporting UI). From there you can decide how you want to allocate your money if one platform is outperforming another. I usually don’t turn one completely off, but I will often limit the spend of the device that doesn’t perform well compared to the other device.
5. Performance by Audience
Depending on your style and the directive from the client, you may or may not be using audiences within your targeting. I have run many successful campaigns where I did not use any audience targeting and simply let google’s maximize conversions bidding algorithm do the work. Assuming you ARE using audiences, then this is a good place to start drilling into your media performance by audience.
Again, if I am using audiences to target, I like to separate out the insertion order into line items by audience. If you have already done that, then you won’t need to pull a report or refer to a dashboard to see performance by audience. Not only that, breaking out your line items by audience makes it much easier to turn down or off audiences that are not performing or allocate more spend towards those that are.
One note on audience reports: you might be disappointed to see that you cannot view your performance cleanly by audience because often times conversions come from people that fall into more than one audience that you are targeting. If that is the case, then both audiences get credit for the conversion. This isn’t very convenient because you don’t want to double or triple-count some of your conversion actions. Yet another reason to break out your line items by audience because you can measure performance by line item rather than by the audience report.
Performance by Creative
Media buyers often forget how crucial creatives are towards campaign success. If I could only pick one (strong creative or strong campaign management), I would choose good creative every day and twice on Sundays. For this reason, don’t overlook these crucial optimization lever. You know the drill: pull a report with creative as a dimension and have a look at which creatives or creative themes are performance.
It’s important to remember the context that your creatives are running in. If you have the same creatives running across all of your targeting lines, then it won’t make a difference. But if you have some creatives running on lines that have completely different targeting compared to others, then you might be getting some false signals by measuring performance by creative. For this reason, it’s best to pull as much of this data as possible in one report that you can pivot and drill down into 2nd and 3rd dimensions.
Performance by Creative Size
Assessing performance by creative size really goes hand in hand with the creative itself. For this reason, I would always pull these numbers together so you can drill into different combinations of dimensions and really see what’s going on. One common example of this is comparing 300x600 to 320x50. These are very different creatives sizes and it’s no surprise that they would perform differently. However, one is an exclusively desktop creative while the other exclusively mobile. When you compare these two, are you comparing the performance of these two creative sizes or really the performance of desktop vs mobile? It’s impossible to say, but again, pull as much of this data as you can out in one report so you can mix and match these dimensions and drill into where your media is performing and where your campaigns are leaking money.
Performance by Publisher (app/URL)
Speaking of pulling data out in one report, performance by publisher (e.g. website or mobile app), is one report that you will have to limit to pulling out only by itself or one combined with one additional dimension. The reason for this is because a URL/app report (unless you have a relatively small inclusion list) with have thousands of publishers on it each and every day. If you combine publisher report with another dimension like date (for example), you will get a report this is thousands of rows long multiplied by however many days your campaign has been running.
Nevertheless, managing a campaign without keeping a close eye on your publisher list is like driving blind-folded. Whenever possible, it’s better to run with an inclusion list (or a basket of PMPs if that’s not too unmanageable) so you never have to worry about showing up on poor-performing sites, or worse, sites that will get you in trouble with the client.
Even if you are running with an inclusion list, performance will vary drastically from from publisher to the next. It’s pretty simple once you have the report pulled: remove any publishers that have a higher CPA than your target. Don’t be afraid of removing pubs that have at least 30 clicks but zero conversions. On the same token, if you have a big publisher that is scaling AND performing well, I would reach out to them and see if you can set up a bundle of deal IDs based on THEIR first party data. Buyers don’t do this enough. These larger publishers will be able to break out their inventory into different deal IDs based on data that only they have about their site/app users. In the world of no 3rd party cookies, this becomes even more valuable.
With that said, it can be time-consuming working with individual publishers and manually setting up Deal IDs. So only choose publishers that are large enough for it to make sense. Yahoo and Microsoft are two good examples that might be top-performers for you and have a lot of inventory spread across many different web properties.
Performance by Environment: App vs Web
Mobile web vs mobile app (DV360 refers to this as ‘environment’) are two completely different animals and you need to analyze how your performance is by environment. Often times, you will find that mobile apps do not perform as well as mobile web. I won’t go into measurement issues on mobile app compared to mobile web (that’s a post for another day), but just remember that you need to check your performance by environment and allocate budgets away towards one or the other depending on performance.
11. Performance by Browser
Of all the levels we have discussed so far, there is a reason why this one is coming last. With that said, browser is becoming more and more important as chrome vs Safari & Firefox handle privacy, cookies and tracking differently. When running on DV360, you will always see an advantage with chrome. And with 70 % market share (last I checked), chrome is not a bad place to be. After a few weeks into your campaign, run a report by browser to see how performance varies. Adjust your budgets accordingly. NOTE: browser is one of the levers of optimization that is available in DV360 but NOT in Google Display Network (as of the time of this writing).
Conclusion
Needless to say, there are many more techniques for you to use in optimizing your campaigns for better performance, but these will keep you busy for quite a while. Remember: don’t “over tweak” your campaign. After making changes to campaign targeting, DV360’s bidding algos will take at least a few days to re-calibrate. You need to give each optimization change time to digest before you move on to the next one. If you feel you need to change many things at once, I recommend creating a new line item with your desired targeting and slowly phase out the line items that don’t have optimal targeting.