What are best practices for mobile re-targeting?

Retargetting campaign defined as: Getting back valuable users you can no longer reach through a local notification on a push notification in the "target" app. 

I have found the best practice to be to first identify why someone should do retargeting

Good reasons to do retargeting:

• App has a GREAT New feature (a tournament or a gamification system or "an event")
• App has genuinely NEW content (A new story in a magazine, or an article in a magazine)
• App has cornered the market, low future acquisition potential
• A bandaid: because you've got some marketing dollars but not dev or design bandwidth to build out proper investment and return features in your core-loop. Meaning: You live in the twilight just before real, real, global launch... And you just don't have local notifications for your first two sprints (My brothers and sisters, I feel your pain like it was yesterday) 

How to do mobile retargeting in terms of effectiveness : 

These are somewhat anonymized but actual first-hand real-world experiences from a dozen app publishers over the last 10 years mostly games with >1 million downloads which were around for more than a couple of years. Your experience may have been different, but if they're the same let me know in this thread :)... heck, if they've been different I'm even more curious.

1. Your company's Catalog/Club/VIP membership app which weighs less than 20 MB. Because you, OR your total-long-term-thinking Superstar CEO  had/have amazing foresight to nudge your power users there when they still loved you 
2. Facebook/Instagram/FAN paid campaigns because you, as a mobile UA professional,  of course, used an MMP from day1 And >90% of users have still given you their IDFAs and advertising Ids  (stunned at how many folks still forego MMPs integration at phase1)
3. Google UAC retargeting campaigns: if you can get into the beta and have the ids from above MMP...or your dev teams are Firebase fans AND your monetization folks like Firebase/Admob and you're solid on China iOS
4. Snap if you've got younger users and an MMP
5. Email if you have had the foresight to collect them from your power users because you and your PM didn't listen to HiPPO which said three years ago:
   • No one reads email. Netflix uses it, and Supercell but that's them we're a cool mobile start-up
   • It's not 1999
   • It has only 2% open rate
   • We'll never have a situation where 50% of our revenue comes from 0.25% of our downloads
   • Besides no one will want to use our app for so long as to switch to a new phone
   • If you really want a budget for re-targeting ads I'll simply allocate it when we need it because we'll never have a cash problem in this company
6. DSPS because everyone else says they kinda work sometimes. And you have no fear about moving around PII because you don't live in Europe or California

Where I've found retargeting campaigns the MOST useful in terms of pure yield is you've got say 200 million downloads in a 250-300 million download market. And of course, the worst thing to do in a retargeting campaign is to launch it and have nothing meaningful to say in the product. THAT'S been a great way to waste scarce capital.

My sense is that re-targeting is something that most -- maybe 75-80%? -- mobile advertisers have not experimented with. When I ask clients why they aren't currently running re-targeting campaigns or have never run re-targeting campaigns, the most common response is: it doesn't make intuitive sense that a user that churned should be advertised to again.

This is sensible logic, but I've also seen enough successful re-targeting campaigns to know that it isn't absolutely true. For one, an app can have substantially changed since a user churned, offering new functionality and purpose that is appealing. And second, many apps are intent-driven (eg. travel, dating), and "churn" can be temporary for these apps.

I've seen a broad range of re-targeting utilization, from situations in which most of the advertiser's spend is on re-targeting campaigns to (more commonly) situations where the advertiser isn't running re-targeting campaigns at all. Here are a few thoughts:

First, it's important to distinguish between re-targeting and re-engagement. Re-targeting campaigns are generally targeted to users that have either churned or who expressed some interest in the app (eg. clicked an ad) but didn't ultimately download. The purpose of these campaigns is to engage people who aren't currently users of the app.

Re-engagement campaigns are targeted to existing users of the app for the purposes of increasing their participation in the app; for instance, a re-engagement campaign might be targeted to existing users of a mobile game around a weekly event in an attempt to get them involved in it. Both of these campaigns should be designed to create incremental increases in engagement / monetization, meaning new revenue is generated that wouldn't exist in the absense of these campaigns.

What does that incremental engagement / monetization look like? For re-engagement campaigns, it means the existing, active user sees an ad and performs some action as a result that they otherwise would not have been expected to perform. For re-targeting campaigns, it means a non-user sees an ad, downloads the app, and engages or monetizes in a way that they otherwise wouldn't have.

Since the idea behind incrementality is that only new revenue is attributed to the source campaign, the most common form of measurement is through holdout groups: taking some subset of the targetable audience, setting it aside, and not exposing ads to it so as to establish a baseline of performance. How the holdout group behaves is meant to represent the negative scenario -- what would have happened had the ad campaign never been run. By comparing the holdout group's performance to the targeted group's, the advertiser can assess the impact of its campaign and the degree to which it delivered incremental revenue. A typical size of a holdout group is 10% of the overall targetable population, although this might be lower if the targetable population is very large.

The companies that I've seen rely on re-targeting and re-engagement campaigns as an important / critical source of new revenue generally run them constantly, using rolling churn definitions (eg. "users that haven't logged in for the last 30 days") to continuously create "lists" of users that they can target on advertising platforms. All of these campaigns utilize holdout groups for measurement and then the holdout groups are targeted in the next "round" of campaigns so as to not reduce the impact of re-targeting / re-engagement by 10%. Since this measurement work is so consistent, these advertisers usually have dedicated analytics infrastructure and tools to help with analysis.

The most common platform that I've seen used for re-targeting is Facebook, since Facebook makes uploading audience lists via API very easy. I've also seen re-targeting DSPs, like Remerge and Adikteev, be used to facilitate re-targeting and re-engagement campaigns. These companies take lists of users from advertisers and bid on inventory as they "see" those users in auctions on programmatic exchanges.