Our Experience with App Store Search Ads
November 17, 2016 ∞
Apple started rolling out Search Ads on the U.S. App Store last month, offering developers and marketers an opportunity to promote their apps on the search screen of the iOS App Store.
Upon signing up for Search Ads developers receive a $100 credit. We were a bit sceptical as to the potential efficacy of advertising Castro in this way but since we could try it out for free we figured it was worth a shot.
In this post I’ll outline how that worked out for us and try to tease out some conclusions we can draw from the data.
Initial Scepticism
Castro is a $5 paid app. After Apple’s cut, we receive about $3.50 for each sale. It seemed unlikely to us that we could afford to buy ad spots at a rate that would make financial sense for a product with such a low lifetime value per customer. We experimented a little in the past with Facebook and podcast ads, and our experiences lead us to conclude that advertising could only really work for more expensive products, or services with recurring revenue.
But App Store ads were new and perhaps things would work out differently there. Apple had mentioned they’d take search term relevance into account when deciding what ad to show so perhaps that could play to our advantage.
Setting up the campaign
Search Ads operate on a pay-per-tap basis. You designate which search terms you’d like to advertise against and how much you’re willing to pay if the user taps through from the search result screen to your product page.
Lukáš Petr wrote a great blog post about experimenting with App Store search ads and we referenced this when calculating what we could realistically afford to pay for a tap.
Based on App Store analytics we saw that it was taking an average of about 45 product page views at the time to get one paid sale.
Working back from there we figured that if we paid $0.08 for a tap then we would at least break even, so we decided to start there. Reading Allen Pike’s Socking Simians blog post we realised there was a pretty good chance Apple wouldn’t even display an ad at that price but we figured it was a sensible place to start the experiment.
We bid on a bunch of generic podcast related terms like “podcast player”, “podcatcher”, and “podcasts”. These are terms that are relevant to Castro but for which our natural ranking is low enough that I don’t believe it leads to a meaningful amount of sales.
We refrained from advertising against competing app names, and explicitly added our main competitors to the campaign negative keywords field to ensure we didn’t end up doing so even unintentionally.
There was some debate on Twitter about the acceptability of targeting ads against competitor’s apps. The general consensus seemed to be that it was not only acceptable but advisable to do so. It felt wrong to us though, so we refrained. I can envisage a future where indie developers are essentially forced to pay if we want to be the first search result for our own app names. I don’t believe this one abstention can stop that, but I also don’t want to be in any way responsible for pushing things in that direction.
For those unfamiliar with Search Ads I should note that all ads are of a standard format, designed by Apple, and without any additional custom copy from marketers. They’re basically an alternate style search result derived from the meta-data that populates your App Store product page.
Early Results
Impressions (a user seeing an ad) trickled in over the first few days of the campaign and we had a handful of taps but no conversions (a user actually buying the app) yet. We were struggling to even spend the first dollar of our budget so we started doubling our bid each day.
$0.16… $0.32…
Impressions weren’t improving much and we began wondering what it would take before we managed to get a conversion. Then two came along in a burst and the limited dataset made them look promising:
- We had spent only $0.84 so far so Cost Per Acquisition (CPA) was only $0.42
- One conversion had come from a search for “podcasts” and the tap only cost us $0.05
- The other was from a search for “podcast manager”, three taps at $0.15 each had lead to one conversion
This was a week into the experiment. Two extra sales at this CPA was some extra profit but not significant enough for us to settle at this rate. We kept doubling.
$0.64… $1.28…
Still only two conversions.
$2.56
Once our bid got above $2 impressions went up considerably and our Cost Per Tap (CPT) was still below $2 so we figured we were “winning” the bulk of the auctions. (The price you pay is determined by what the second highest bid.)
We stopped adjusting the price and let it sit there to run its course, which took about three weeks.
Final Results
- We spent all of the $100 credit we received from Apple.
- Our ad received 5,117 impressions.
- 109 users tapped on the ad. That’s a Tap Through Rate of 2.13% and an average CPT of $0.92, for those of you comparing against your own stats.
- Of the users that tapped through to our product page 14 went on to buy Castro, that’s a Conversion Rate of 12.84% and a CPA of $7.14.
Conclusions
One interesting observation from this data is that the conversion rate from a user viewing our product page to buying Castro was much higher than what we were seeing organically through App Store analytics.
- Conversion rate on ads was 12.84% versus about 2.22% organically at the time we started the campaign
- In recent weeks the organic conversion rate has improved to around 5%.
- I believe the lower rate before was due to heavy traffic we received while featured on the App Store during launch. Hundreds of thousands of potential customers tapped through to our page from featured banners knowing nothing about the app in advance, and these users would seem to convert to customers at a lower rate than others who seek the app out.
At $7.14, the CPA we saw here is more than twice the revenue we generate from a sale. We might be able to improve this some bit, but it’s difficult at the moment to see it changing enough to become a viable marketing opportunity for us.
We have some ideas that might help. Since the search ads don’t have any additional advertising copy these changes mostly focus on improving our App Store presence itself. There could therefore reasonably be benefits beyond improving ad performance:
- Our full app name is currently “Castro — Play and Share Podcasts” which is slightly too long to fit in the ad format and so the “Podcasts” gets truncated. We will rename the app to “Castro: Podcast Player” in our next update. This could potentially lead to more taps since the function of the app should be clearer. I don’t imagine it will lead to more conversions though.
- Our product page could be improved. We’re working on a new set of screenshots at the moment that should tell a more compelling story about the app. Hopefully these will help us increase the conversion rate.
- Perhaps the app description could be improved too although I imagine its impact would be minimal.
- We could experiment with app pricing. Perhaps $5 is too much, or too little.
- We could start advertising against competing app names, but it’s unlikely that we’d get a better return on those terms than the generic ones, so we’d be selling our souls for nothing.
And of course there’s some known unknowns:
- Search Ads are still new and people are all figuring it out at the same time (including Apple). The rates for each keyword may be varying a lot. At the moment, when we run at a low rate, it’s profitable but the volume is so small that it’s hardly worth the attention. Perhaps we should keep our toe in the water though in case prices do settle down.
- It’s not clear yet if Apple has a cut off CPT bid below which they’ll refuse to place an ad. It seems not based on the small number of impressions we did get while at $0.08.
- Apple’s criteria for determining relevance and how they apply that criteria is opaque. Would Apple accept a lower bid for a more relevant app for example? How is relevance determined? Do we face the same obstacles with getting good ad rates as we do with improving our organic search ranking?
Am I missing anything else? If you’ve further thoughts on this, or other data that contradicts my conclusions I’d love to continue this discussion on Twitter or by email.
The silver lining is we only used the free $100 credit, so we did get about $49 from the experiment, and this blog post!