CaddieOpen in CaddieDownload for Mac

The Simplest Facebook Ads Strategy for App Installs

Shared by Caddie Library · Updated
App Masters11:50Open on YouTube
Chapters show their main points

The bottom line

App installs scale on Facebook when you start at the install, move down the funnel step by step, and steer the algorithm with broad targeting, selective creative and high-quality feedback signals. This lets the algorithm find the right users instead of just the cheapest ones.

Key takeaways

  • Start by optimizing for the install, then move down the funnel. 0:00
    • Optimizing for level 10 of a game from day one starves the campaign of data.
    • The path is install, then a level 2 or trial start, then subscription.
  • Broad targeting is the only way to reach scale. 1:00
    • Interest targeting works, but it never gave scale.
    • Seed lists from other apps can be tested at 1%, 5% and 10%. Audience expansion should stay within 50% above your estimated audience.
  • Creative screens who responds to the ad. 1:55
    • The Abound app first drew in people outside its audience. Adding Indian faces and a send-money-to-India feature made viewers self-select.
    • The result is an 80% match rate, where only 1% of the demographic is Indian.
  • Feedback signals must be recent, frequent and high quality. 2:39
    • Facebook needs about 50 events in a rolling seven days. Google needs about 10 per campaign per day.
    • At Elevate Labs, firing the trial-start event 12 hours late, only for users who had not cancelled, improved user quality by almost a third.
  • Early small-budget wins do not prove product-market fit. 10:50
    • At a few hundred dollars a day, the algorithm easily finds the low-hanging fruit.
    • At $10,000 a day, costs often go out of control.

Don't be a hero. Do what works, especially in your beginning.

Vivek Girotra, VP of Marketing at Times Internet

Chapter summaries

0:00 Steering the algorithm from the side

The algorithm can't be controlled directly, only nudged with targeting, creative and feedback signals.

  • Three side levers guide the algorithm 0:32
    • It works like sweeping a curling stone: you guide it but never touch it.
    • The three levers are targeting, creative and feedback signals.
    • Each lever corrects the algorithm's course after it starts.
  • Broad targeting needs experiments to find the audience 1:15
    • Audience lists from sister apps seed the campaigns for each locale.
    • Expansion test: 4 million Indians in the US against a 15 million audience is too wide.
    • An audience member asked about Advantage Plus, and the answer was the same 50% cap.

1:55 Creative as a screen

An ad's content decides who bothers to respond.

  • Creative filters out people outside the audience 2:00
    • Abound launched as a cash back and rewards app, and shopping is such a big category that it drew the wrong people.
    • Cameos from B-level but recognizable celebrities saying something nice about the product have had some success.

2:39 Feedback signals: recency, frequency, quality

The event sent back to the platform must arrive soon, arrive often and predict revenue.

  • Recency is how soon after install the event fires 2:45
    • Google's 10 events a day is why it recommends a budget of 10 times your target CPA.
  • Custom events can raise quality beyond the standard pipeline 3:25
    • Engineering built "HQ sign up" and "HQ bank link" events for users judged high quality.
    • At Elevate Labs, trials were 7 days and auto-renewed into a subscription.
    • A bar chart showed 30-odd percent of cancellations came in the first six hours.
    • Tools like RevenueCat might offer something similar, but the speaker isn't sure.

5:18 Broad versus interest targeting

Interest targeting can win on cost, but the right choice depends on the scale you want.

  • Interest targeting can cut cost per install 5:18
    • For a sports betting client, targeting people who liked all 32 NFL teams beat a single NFL interest.
    • Broad has run expensive for this marketer, who focuses on cheap campaigns rather than scale.
    • Without first-party data, it all depends on the scale you're trying to achieve.

5:55 Broad targeting, B2B and Advantage Plus

Strong creative can replace an audience list, and Advantage Plus needs a tighter seed to avoid over-expanding.

  • Without an audience list, broad targeting plus creative is enough. 5:56
    • The machine learning finds the best audience from a small sample.
    • For B2B, LinkedIn is the first place to look on a limited budget.
    • LinkedIn targets by company, designation and years of experience.
    • If you use interests, the more granular, the more useful.
  • Advantage Plus beats manual targeting but can stretch a lookalike too far. 7:25
    • A 10% lookalike seed may expand beyond your real core audience.
    • Dial back with a 5% seed list, then run Advantage Plus on it.
    • Expansion grows your audience base, so watch how far it goes.

7:52 Lookalikes, retargeting and first-party data

Results timing depends on the category, and retargeting is hard to read on iOS.

  • How fast a campaign pays off depends on the category. 7:56
    • Hyper casual games get a CPI of about $3 to $6.
    • Midcore games run $20 to $25.
    • A 4X builder game cost almost $50 per install.
    • Audience and targeting shift these numbers further.
  • Retargeting on iOS is murky, while Android still works. 8:26
    • With no device ID there is no suppression list, so installed users see ads.
    • SKAN can now tell redownloads apart.
    • Use lookalikes for prospecting and custom audiences for retargeting.
  • First-party data powers both retargeting and lookalikes. 9:14
    • RFM analysis (recency, frequency, monetization) splits users into power, casual and churned.
    • Messaging those segments on Facebook is retargeting.
    • A custom audience of power users seeds the lookalike.

10:00 Starting out and placements

Beginners should borrow experience, copy proven ads and leave placements to the algorithm.

  • A newcomer should bring in an experienced person and copy what works. 10:02
    • Budget for learning costs.
    • Copy winning ads from the Facebook or TikTok ad library.
    • Skip expensive famous influencers at the start.
    • Track retention and when it flattens to see how sticky the product is.
  • Dynamic placements let the algorithm pick where ads run. 11:26
    • It chooses between Instagram, Messenger and Facebook.
    • He once manually excluded Audience Network, but no longer does.
    • Instagram ads are effective for app downloads.

In the growth library

Vivek Girotra, VP of Marketing at Times Internet. Teaches how to run Facebook app install campaigns by walking down the funnel, going broad on targeting, and steering the algorithm with creative and feedback signals. The result worth hearing is an 80% audience match rate on a niche app and a trial-event trick that lifted user quality by almost a third.

Lessons that draw on this:

More from App Masters · The whole library, and ask it anything

Transcript312 lines
0:00

anytime you're starting off Facebook Google whatever you always want to walk down the fun like you don't want to start off a campaign that is optimizing for a level 10 of your game so you want to start off with the install get some data get your learnings make some modifications the product or your onboarding as you need then maybe start a level two optimized Campa and key or a trial start and then go to subscription so you always want to be slowly walking down the funnel to reach your sweet spot of scale and and Revenue you seen that sport where you have this disc of ice and you're kind of just using this brush to guide it into the right direction you can't actually touch it you don't have control of it I think that is really what you're doing to the algo where you're trying to Gast correct it through ancillary levs and there are three main levels that I found for that one is targeting second is creative and

0:57

the third is uh your feedback signals so the first one is targeting right we've generally found that interest targeting it works but I have never been able to get scale for it if you're looking to build a venture back company at scale then really the only way out is Broad targeting and then running different experiments to figure out how how to best find that audience one way we have done that is use first party data right we have so many different products within the times portfolio we use a lot of audience lists from other apps and use them as seed list for different local IES you can keep running experiment with 1% 5% 10% try with audience expansion I think someone asked about the Advantage Plus as well audience expansion works but I would recommend that in general whatever you think your estimated T is your estimated audience from as a result of the expansion should not be more than 50%

1:48

above that so for example if I know there are 4 million Indians in the US and my audience is 15 million then it it's it expanded the second is creative it's a screening mechanism right so the product when it's called abound by The Times of India when we launched it in the first phase it was cash back and rewards application right shopping cat is just such a big category it started attracting a lot of people who were just not part of our audience set so we started adding more faces that were Brown and we added this feature of sending money to India now that itself acts as a self- selecting procedure and we've seen some level of success with Cameo getting B level celebrities but recognizable faces and getting them to run ads uh say something nice about the product third one which is you know feedback signals and here three things matter recency frequency quality recency is how soon after the install does this

2:49

event occur second is frequency Facebook requires at least last check a rolling day seven 50 events as feedback signal for campaign Google requires 10 events per campaign per day so that's that's why they have the recommendation of your budget should be 10 times your tcpa and the third is the quality you know what is it that you are actually trying to optimize for right so that event has to ultimately be or have a good correlation to their ultimate goal which is I'm assuming Revenue what I found is is not just depending on your regular event flow in your pipeline you know your ETL that is normally there but also thinking a little bit deeper and creating synthetic events so so we worked really closely with our our engineering team and created what what we call a HQ sign up and HQ bank link so these are users that we've identified as high quality and they and we send that back as a feedback signal to Facebook Google as an

3:47

optimization event something similar that I had done at a previous company with Elevate labs and that was subscription at brain training games so what used to happen is you would sign up for a 7-Day trial which would then Auto renew du a subscription and once we created a bar chart of how much time after the trial sign up do people cancel the trial we found that a good chunk I say 30 odd person occurred in the first six hours so what we did is we created a synthetic event that instead of firing the trial start at the time of trial start it was an adjusted trial start and we waited for 12 hours after the trial and we discounted anyone who had already cancelled their trial so basically what I did was I introduced a mild DeLay So my recency is still taken care of but I improved the quality by almost oneir and that I think was clutch in terms of us being able to find the right users and you know get the high quality users on

4:45

board was that something that you had to build internally Vivic or is that like something you can do through okay I would reckon at this point some of this tooling might be available if you are running some of the like a revenue cat or something but I'm not sure what we done at you you know times right is as a result of this for bound we are now running at an 80% basically match rate for our specifically specific audience I call an ATX magnification right you have 1% of the demographic audience but 80% of the audience in our app is now an Indian person everybody talks about broad but every time I've run broad it gets super expensive and you know I like to focus on like really affordable cheap campaigns you know what I mean I'm not talking about scale and so one client we had that was in the sports like betting we were interest based targeting with NFL when I changed the interest based targeting to all the people who liked

5:38

all the 32 teams I dropped the cper install so I personally like to do interest based targeting but I'm not an expert like you are so tell talk to me like what's the best approach if you didn't if you did not have that first- party data that you do have with the look likes it all depends on the scale you're trying to achieve yeah in most cases you are better off even even if you don't have an audience list available keeping the targeting Broad and letting your creative do the heavy lifting so at this point the ml has gotten smart enough that it can with a very quick sample size determine what kind of an audience is the best F if you do want to interest targeting uh I think you want to try and just understand a little bit more about your audience and NFL is just so broad so I think more granular you're able to get is going to be always more useful let's back to Mary's question real quick I have a

6:31

hypothesis that maybe salespeople are more interested she should she do more interest based targeting like this or should she create creatives that say Hey salespeople you know we we got an app that's going to help you remember is that like is creative more important then and you could just keep it broad from like you know 35 to 40 you know 55 whatever that's an interesting one I haven't really a B2B campaign here before but you know my if I had a limited budget my first inclination to be would be to use LinkedIn and Target by designations instead of trying to go broad on Facebook where so so it's not just audience but also context that is important right I would say it's not the first place I would go if I were targeting sales people right the targeting levels on on LinkedIn have become pretty sophisticated you can Target people by companies you can Target them by their designation number

7:22

of us of experience Benjamin said any words on the Advantage Plus targeting from Facebook I'll perform all my manual targeting attempts you know it will increase your audience base but you have to make sure it doesn't expand it too much so sometimes what might happen is if you do a 10% lookalike this on a seed list with Advantage Plus it might expand your audience a little beyond what actually is your Cod so you might want to dial it back by taking a a 5% seed list and running an Advantage Plus selection with that peram asks how long does a a lookalike a retargeting campaign take to really produce results completely depends on your product your category what kind of audience you're targeting let's go back to the gaming analogy if you're running a hyper casual game you you'll get a few dollar CPI like maybe three to6 dollars if you're running a midcore so 20 to 25 machine Zone where we had you know 4X Builder

8:19

game we would get you almost $50 for install so I think it really depends on your audience your Target and and retargeting now is especially in iOS it's it's a very dicey again because by default you are retargeting anyway right now if you're using because there's no device ID so there's no suppression list anymore so your ads are in effect being seen by people who may already have the app installed and you can actually find that I think I think scan is now able to differentiate that as well as the redownload but retargeting it's it's really tough to get a a read on retargeting right now on iOS but on Android obviously you still have those Avenue available so do you is it something you should still try to attempt or is it like hey there's other ways think about other AD channels if you're trying to scale up Facebook again look likes I would use lookalikes for prospecting I don't I don't know if how

9:09

retargeting is when you're a custom audience so just to differentiate that right right retargeting is when you have a custom audience say you know this is something we do at our end as well right we we collect first party data which is your email and phone number we obviously built out a robust CRM and you know data pipeline we conduct rfm analysis at RN right recency frequency monetization analysis to sort diversify our audience into different segments some are power user some are kind of casual users some of them are turned out and then we can take that first by data create customer audiences and then send those people messages through avenues like Facebook so that that would be more of a retargeting maneuver if I had to do lookalikes I would take out power users get enough data to put that in a customer audience and build a look like off that customer audience to Prospect and find users who are more like a power

9:59

users where would you start if you have no experience with Facebook if you have no experience with Facebook I would find someone who has experience with Facebook assuming that you know you have expertise or at least the consultation the ear of someone who has done this before right I I would one anticipate some amount of learning cost right you have to be willing to spend money to make money two I say don't be a hero right do what works especially in your beginning don't be fancy and try to spend money on some famous influencer literally just go to your Facebook ad library or Tik Tok ad Library see what works in this category replicate that and run your ad to see what kind of response you get what is the your retention what do you know how how when does it flatten out so to understand how sticky your your product is for this audience and one important thing I really want to highlight the fallacy of

10:52

the early stage product you know you'll meet a company that creates an MVP runs a few hundred in in Facebook ad they find this audience and they love the product and they're like holy we found product Market fit let's get more money from investors and once they start spending you know $10,000 a day the cost go out of control right and that is because if you are operating at a scale of you know a few hundred a day any algorithm based Network they can generally really pinpoint the low hanging fruit pretty easily all right nurx says Instagram ads really effective on app downloads in general you want to give the algo freedom to select between Instagram messenger Facebook fan Etc there was a time when I would kind of be a little bit more manual and select hey I don't want my ads to run on fan or something but I think at this point you just want to have Dynamic placements