|
Hey, I took a mini-break last week for Easter and spent the day cooking some lamb shanks instead; you didn’t miss anything. Email-wise, I mean. The lamb shanks were divine. But I shall make up for my gourmet slacking with an appropriately meaty missive this week. When not gorging myself, I did some testing on Facebook’s new ad image formats and thought it would be a great time to break down:
- how ad images are evolving
- why Facebook is making these changes
- where you can safely test this stuff
- what you need to do about it right now.
Well, let’s answer the final one immediately: you don’t need to make any changes, so don’t panic if you have ads running, or even ads planned and images already designed or whatever. The first big takeaway is you don't have to use the new image format. Doing nothing is a totally valid option and indeed I did little or nothing about this myself in the last couple of months since I heard about this change. Don’t worry on that front at all. But maybe don’t do absolutely nothing when it comes to images either. I certainly don’t recommend using Facebook’s fancy new ad image generator. Over on my Facebook Page this week, we have been, eh, enjoying what the AI system suggested I use to advertise a gritty crime novel.
A gritty crime novel with a very human detective, I shouldn’t have to add! Maybe we shouldn’t be so worried about robots taking our jobs, eh? I recommend being a little more hands-on with your own ad images – either designing them yourself or getting them designed for you, perhaps. Whatever your approach, there might be some reasons that you will want to play with these changes as there are potential benefits for you. And with the margins for success and failure (and profit) often being so tight on Facebook… every little performance boost helps. So, let’s break it all down today in advance of a new Image Workshop episode that will basically be a screen-sharing design tutorial covering these changes in audiovisual detail so you can follow along. (Make sure you subscribe on YouTube to get notified when that drops!) new reccosI first heard about this change from Jon Loomer a couple of months ago, but Facebook hadn’t updated much of the official guidance and documentation yet so I wanted to be sure this would be an official change and wasn’t just some testing or whatever, as Facebook is wont to do. But I noticed this week that Facebook is now officially recommending the new image size and has updated its guidance on old image sizes too. You can read the new language here, if you are (also) that kind of nerd, but just note that Facebook hasn’t updated its guidance everywhere yet and you will still see the old recommendations in various places. The new Design Recommendations are as follows:
File type: JPG or PNG
Ratio: 1.91:1 to 4:5
Resolution:
-1:1 ratio: 1440 x 1440 pixels
-4:5 ratio: 1440 x 1800 pixels
And these are the new Technical Recommendations:
Maximum file size: 30 MB
Minimum width: 600 pixels
Minimum height:
-1:1 ratio: 600 pixels
-4:5 ratio: 750 pixels
Aspect ratio tolerance: 3%
What you really need to pay attention to are the two bolded items. First, the pixel size of the standard square image that you will be familiar with from Instagram (historically) and Facebook (a little more recently) has changed somewhat. Before, Facebook was recommending 1080 x 1080 pixels and that has updated to 1440 x 1440 pixels. A small change on the face of it, you are still designing a square image (or what Facebook refers to as a 1:1 ratio image) but I would recommend changing the pixel size of your square images going forward. I wouldn’t be so concerned about continuing to use an older 1080 x 1080 image you already have designed, but I would just take Facebook’s guidance here going forward to prevent any possible blurring or loss of resolution in the future. Although I should note I couldn’t see any issues in my testing from using 1080 x 1080 images this week. (Side note: if you want to convert exisiting 1080 x 1080 images I’d recommend using a tool like Canva and its Magic Resize feature (I think this is only available on Canva Pro now) rather than just changing the image size in something more basic like MS Paint, as that could introduce the very blurriness that you are trying to avoid! If you can’t resize the image properly yourself… then just upload it as 1080 x 1080 and you’ll have fewer potential issues.) I have started designing all new square images as 1440 x 1440 now, because it makes the transition to the new format much easier in the design process. That new format – the second thing you should pay attention to – is a slightly taller, almost portrait image, of 1440 x 1800 pixels. And this is what Facebook is recommending we use now over the square image you might be used to. This is a soft push towards a new ad image format. You can still use square images (indeed, you can still use the much older letterbox images) but let’s take a look at why Facebook is doing this, and how it could potentially benefit you. why facebook, why?The new image format is slightly more like the shape of a book, you might have noticed, although that’s not the reason Facebook is recommending this change. The real reason is similar to why we moved away from that old letterbox shape several years ago – phone screens are portrait and desktops are landscape, and Facebook’s users are increasingly using the platform on mobile devices rather than on desktop computers. Just like when Facebook introduced that Insta-influenced square ad image, its testing has shown that a taller ad image performs better on mobile - understandable as it dominates the screen. As most traffic is mobile these days, Facebook is recommending we switch to those more portrait-shaped images going forward. But as with any Facebook changes, I recommend a little caution here. Facebook’s generic advice is tailored for its typical advertisers; authors and readers are far from typical. Before embracing these changes wholesale, I would suggest doing some testing – as I have been doing this week. Or just letting others test it and report back. There really is no great rush to adopt these changes – not least because the changes aren’t quite as dramatic as the shift from letterbox to square a few years back. And the results don’t seem dramatically different either, quite frankly, if my initial testing is anything to go by. The point of this newsletter is simply to let you know what is changing and why, rather than some kind of PSA where you must take immediate action. But if you do want to test, here’s how I recommend doing so in a way that won’t impact your operations so much. testing timesIf you follow my recommended approach with Facebook Ads – covered in this free hour-long tutorial on YouTube – then you probably have a very streamlined campaign structure. My default is to have one ad, inside one ad set, inside one campaign. And then I will have separate campaigns for each retailer and territory. So, one for Amazon USA, a second for Amazon UK, and so on. That structure might be replicated to push a new release, and again to push a free Book 1, and again to push an entire series at once with a series page ad – depending on what I have going on (with a lot more going on when I’m launching a book, for example). But the basic structure is one campaign, containing one ad set, containing one ad. However, when I have a campaign that is expected to reach a large audience – either because I want it to run it during launch week at a higher spend, or because I want to run it for a longer time at a lower spend – then I like to combat ad fatigue by putting an alternate image in the campaign. While Facebook has a few different ways to do that automatically these days, my preferred method is a bit more old school – i.e. to make a second ad inside the ad set and letting the system show both and then defer to the winner. The advantage of doing it this way is that you can (a) monitor performance separately and compare performance easily on the fly and (b) manually intervene if desired. And that might be necessary because, like all AI I guess, the system is over-confident and picks a winner really early. Like after 100 impressions or less, which any (puny human) statistician will tell you is an insufficient sample size. I guess we shouldn’t be overly surprised about the next part, given its artistic efforts above, but its winner-picking can seem a little suspect sometimes too, especially at first blush. For example, these were the initial results on my test. I actually included three image variants in my test, for kicks. The standard square image, the new, taller image, and the older letterbox image. And this is how results looked after about $40 in spend across a few hours. Letterbox – CPM 4.67, Spend 1.36, Impressions 291, Clicks 12, CTR 4.12%, CPC 0.12 Square – CPM 5.10, Spend 4.02, Impressions 789, Clicks 17, CTR 2.15%, CPC 0.24 Portrait – CPM 4.76, Spend 25.60, Impressions 5,374, Clicks 113, CTR 2.10%, CPC 0.23 If you are wondering why Facebook has chosen the Portrait as a winner here, you are not alone because it’s not winning on any metric we can see – CTR, CPC, CPM, take yer pick. You might think Facebook is putting its finger on the scale here, but sometimes it’s simply just a reporting delay – especially when serving the ad this quickly (I’m testing during a launch so the budgets are high). When I checked back a little later, results had settled down as the ad was dialled in. CPMs halved, CPC dropped, and the apparent performance gap between all three ads closed significantly once the alternates got some more impressions, and Facebook’s choice of winner seemed a little more logical. That’s not to say that Facebook isn’t preferring its new format, more that doing so seemed less of an outlier once reporting had caught up and all three ads reached a certain level of serving. I don’t know for sure – this is just a guess – but I suspect that clicks are reported a little faster than impressions which makes early results seem a little more favorable, and when more impressions come in on the less-served ads, the apparent gap in results closes quite a bit. An alternative possibility is that the system serves to the most likely clickers first so initial results can be more favorable. (And both could be true to a certain extent.) When you want even more control – and testing is one of those scenarios where it’s often useful to have as much control as possible – then you can go a step further and put each image format in its own separate campaign, so you can guarantee that a certain amount of money will be devoted to testing each image. (Indeed, I might have done that myself if I wasn’t testing during a launch where I had other considerations.) Or you can simply switch the Portrait and Letterbox ads off manually and force the system to devote all the budget to the Square ad – for example. And I often simply do this rather than going to the trouble of creating a separate campaign. But it’s up to you. Outside of testing, the other reason I often set things up like this is to combat ad fatigue. During a launch of your own, if you are spending a lot of money and/or if your audience is more limited in size, you might see your Frequency creeping up. This is a measure of how often your audience – on average – as seen your ad. If that goes above 2.0 in a week, I will often look to freshen things up. And a very basic way to do that is to simply switch off the ad format that has over-served, and force the system to deliver the under-served one instead. It might have a slightly lower performance – assuming the system did indeed correctly select the winner – but that might perform better than serving the exact same ad repeatedly to the same people. I get good results from this approach at least – it’s something you can test for yourself, if you wish. Whatever your approach here, don’t be too worried about these changes. I’m sure some people will try to dress them up as seismic, and proceed to sell you the solution, as is the way of our people. However, I don’t personally see it as being a big change, or something you need to jump on immediately or anything like that. But at least you know what it’s all about now. And my results above did settle down, as I said, with the winner-picking seeming a little more logical, but maybe I’ll dive into the test results in a more substantive way next time, when I should have more than a day’s data to share and we can see if there is any variance in different territories. Dave P.S. Noodling music this week is Haruomi Hosono with The Last Paradise. |
![[New post] Level Up Your Author HQ With GoCreate.Me](https://lh3.googleusercontent.com/blogger_img_proxy/AEn0k_vlPLoVWsmzP6FHeo4_qT5uBfkDBChbZNzoH7BkdDI6ar3S45WTjGyswqx8iP1b3_J2JARGJbrpAnCPMPS0Tdg72aESwSIFwf_dbQrA6hBVmNs6tXccQnBfzZWuMa1gV18bLK9hGpXvEuZlLr0CIhy1bzPkcUXcZhenhTZwsMqlTbzIEVKMDXTX7py61vGGBg=w72-h72-p-k-no-nu)
No comments:
Post a Comment
Note: Only a member of this blog may post a comment.