Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Mobile & App Marketing › Mobile Messaging

Re-engaging Lapsed App Users

How to bring lapsed app users back: defining a lapse for your app, who each channel can still reach, the reason to return, paid re-engagement, and holdouts.
Guideintermediate11 min readUpdated

When an app user has lapsed

A lapse is a gap in use after which a person rarely comes back on their own. The right gap depends entirely on how the app is used. A messaging or fitness habit is broken by days without a session; a travel or tax app may be used a few times a year and is not lapsed after months. A borrowed figure catches the wrong people: too short, and the campaign spends its effort on people who were simply between sessions; too long, and it reaches people whose habit has already gone.

Take the gap from the app's own data. The cohort analysis tutorial builds the tables that show, for each group of users, how many are still active after each period, and the churn and retention metrics guide sets out how to define churn and retention so the numbers mean the same thing every month. From those, plot the chance that a user returns after each length of gap since their last session. The point where that chance falls away is the lapse threshold, and users approaching it are the ones where a message is most likely to matter. The tutorial on setting an app's lapse threshold builds the table of return chances by length of silence from event data, step by step, and checks it by season and by how people use the app.

Define the session that counts, too. Opening the app to dismiss a notification is not the behaviour the business needs; completing the core action is. A lapse measured against core actions, the same action the app treats as activation, finds people whose habit is breaking even if they still open the app. The win-back glossary entry makes the same point for purchases: set the threshold relative to the customer's own rhythm, not as one number for everyone.

Not every lapsed user deserves the same effort. RFM segmentation sorts people by how recently they were active, how often and how much value they brought, and its treatments separate those at risk, who need a reason to return before the habit breaks, from those long lapsed, who get a time-limited attempt and then suppression. Recency and frequency also decide who is worth a paid touch, which costs money per person reached.

Who each channel can still reach

The channel decision for a lapsed user is mostly made by what they have left switched on. The comparison of SMS and push sets out the difference: push reaches an app installation and needs the operating system's permission; SMS reaches a phone number and needs marketing consent under direct-marketing rules. Map every lapsed user to the channels that can still reach them before writing a single message.

Lapsed user's statePushIn-appEmailSMSPaid media
App installed, notifications onYesOn return onlyIf they consentedIf they consentedPossible, rarely needed
App installed, notifications offNoOn return onlyIf they consentedIf they consentedYes, through app audiences
App uninstalledNoNoIf they consentedIf they consented, occasionallyYes, where audiences allow
Never gave email or phone consentOnly if installed with notifications onOn return onlyNoNoYes, where audiences allow

Three rules follow. Push is the cheapest channel for people who can receive it, but its reach shrinks with every disabled notification, which is why the push opt-in rate and the enabled rate matter long after install. Email and SMS reach people push cannot, but only where they consented to that channel; the SMS and WhatsApp automation guide sets out, from the ICO's direct marketing guidance and the CTIA's messaging principles, that consent should be specific to the channel, that every message should identify the sender, and that an opt-out must work and be honoured. That is educational guidance rather than legal advice; check the rules for each market. And the routing table in the SMS vs push comparison sends win-back for uninstalled users by email first, with SMS only occasionally and with consent, never by push, which cannot reach them.

A lapsed user needs one well-chosen reason to come back, as the push notifications and in-app messaging guide puts it, not a daily reminder that the app exists. The strongest reasons are specific to the person: something new since they left that relates to what they used (new levels in a game they played, a new route on a commute they saved, a reply they never read), or something they left unfinished (a saved basket, a half-completed course, a booking draft). The customer retention strategy guide adds the deeper point: win-back works best when it addresses the reason people left. If many lapsed because a feature was missing or broken, the message to send is that it is fixed.

The words matter as much as for any push. The tutorial on writing push notifications leads with what changed for this person, names the specific thing and says what a tap will show; a re-engagement push written as "We miss you!" fails all three.

The tap then has to deliver the reason. A re-engagement message should carry a deep link to the exact screen it promised, with the item loaded, as the push guide insists for every message. For uninstalled users the link has an extra job: a person who taps an email link without the app goes through the store, and the deferred deep link described in the deep links and app analytics guide carries the destination through the reinstall, subject to the platform limits the guide describes. Design the fallback for when it does not, and for a person who returns signed out, because a returning user who lands on a login screen and then the home page has lost the reason they came back.

Firebase's documentation describes Cloud Messaging for push and In-App Messaging with event-based triggers, as the push guide states, and most engagement tools work the same way, which is what lets a re-engagement message follow behaviour, such as no core action for the length of the lapse threshold, rather than a calendar.

Paid media can reach lapsed users that owned channels cannot, at a cost per person that owned channels do not carry. The app install campaigns guide lists the campaign types the networks offer for apps, and two of them are aimed at existing users: Google Ads App campaigns include an engagement type alongside installs and pre-registration, and TikTok's app promotion includes retargeting alongside installs, as the guide's table states from each network's help centre. This page claims nothing further about either network's features, audience minimums or matching rules; their own documentation covers those.

The audiences behind such campaigns come from the app's events. The deep links guide describes audiences built from events and properties, such as people who activated but have not returned in a set period, feeding messaging tools and ad networks alike. Use the same lapse definition for paid as for owned channels, so the two do not chase different groups, and exclude people that owned channels can reach cheaply, such as users with notifications on who have not yet been sent the push. Recency and frequency from the RFM segments decide who justifies the cost: a recently lapsed, previously frequent user is a better paid prospect than someone who opened the app once a long time ago.

Paid re-engagement ads need the same deep link discipline as owned messages, landing an existing user on the screen the ad promised rather than the store listing or the home screen, and the same measurement, because an ad network counts the returns it believes it caused under its own rules.

Measuring returns against a holdout

Counting every lapsed user who came back after a message as recovered overstates the campaign, because some were coming back anyway. The push guide's answer applies directly: keep a random share of eligible users who receive nothing, a holdout group, and compare returns between the groups. The difference is what the campaign caused. The win-back glossary entry gives the same method for purchases and shows how much a report can overstate recovery when the trigger catches people who were between sessions.

Run the holdout per channel where the volumes allow, because channels differ in cost and in reach. A push to users with notifications on costs almost nothing; an email to uninstallers costs little; SMS and paid media cost per message or per impression. Each should earn its place against its own holdout. Where an offer is in question, the win-back offer test runs three arms, an offer, a message without one and a holdout, so the value of the contact and the value of the discount are priced separately.

Count the costs as well as the returns. A re-engagement push that brings a few people back and makes more of them turn off notifications has shrunk the channel for every later message; an SMS that brings a few back and draws opt-outs has spent consent that took effort to collect. Track disables, topic opt-outs, unsubscribes and uninstalls by group, as the push guide advises, and read returns by cohort over a window long enough to show whether they stay or lapse again, which shows up later in churn.

Knowing when to stop

A re-engagement sequence should be short and have an end, as the retention strategy guide and the win-back entry both say. After a few well-spaced attempts across the channels a person can still be reached on, someone who has not returned has made a choice. Move them to a low-frequency list or suppress them from re-engagement entirely, and stop paying to reach them.

Stop a campaign, not just a person, when the holdout says returns would have happened anyway. A re-engagement push whose holdout returns at the same rate is noise with a cost in disables; an email whose holdout matches it can be cut without loss. Stop paid re-engagement when the incremental returns it buys cost more than they are worth, judged on what returners go on to do rather than on the return itself. And when many lapsed users share one reason for leaving, stop messaging and fix the reason, because the retention strategy guide's order holds: diagnose, change the experience, then tell the people it affected.

A puzzle studio brings back lapsed players (illustrative)

Illustrative example — not a real result

Wibbleknock Puzzles defines its lapse and tests its channels

Wibbleknock Puzzles, an invented puzzle game, used to send "We miss you!" to every player inactive for 7 days. Its own return curves show that most players who skip 4 days come back by themselves, while fewer than 1 in 10 of those gone 14 days ever return unprompted. It moves the threshold to 10 days without completing a puzzle. Players past it with notifications on, about 31,000 a month, get one push naming the new level pack in the series they played most, deep-linked to its first puzzle. Uninstallers who agreed to email, about 12,000 a month, get one email with the same news and a link that reopens the pack after reinstall. Each channel keeps a 15% holdout. After 8 weeks, 14.2% of push recipients complete a puzzle within 14 days against 9.1% of their holdout, while notification disables run at 0.4% against 0.3%. The email recovers 2.6% of uninstallers against 2.2% in its holdout, a difference within noise at this volume. Wibbleknock keeps the push, cuts the email to a single message when a large pack launches, and suppresses players who do not return after that.

The mobile messaging hub sets re-engagement beside the other message types, and the push guide covers caps and quiet hours that apply to re-engagement as much as to anything else.

Common mistakes

  • A borrowed lapse threshold. A fixed number of days catches people who were between sessions or misses people whose habit is already gone. Take the gap from the app's own return curves.
  • Push for people push cannot reach. Users who uninstalled or turned notifications off cannot receive a push. Map each lapsed user to the channels still open to them.
  • Treating app permission as channel consent. Email and SMS need their own consent, specific to the channel, with sender identification and a working opt-out.
  • "We miss you" instead of a reason. A lapsed user needs something new or unfinished that matters to them, not a reminder that the app exists.
  • A tap that lands on the home screen. Deep link every re-engagement message to the screen it promised, through the reinstall where needed.
  • Paying to reach people a push would have reached. Exclude users owned channels can reach cheaply from paid re-engagement audiences.
  • Counting every return as a recovery. Without a holdout, the campaign takes credit for people who were coming back anyway.
  • Never stopping. A sequence with no end spends goodwill and money on people who have decided. Suppress after a few attempts.

Frequently asked questions

How long before an app user counts as lapsed?

It depends on how often people normally use the app, so take the number from your own data rather than from a rule of thumb. Plot how likely users are to return after each length of gap since their last session, and set the lapse at the point where the chance of a return falls away. A daily habit app and an app used for an occasional trip will land on very different gaps, and a borrowed number catches people who were simply between sessions.

Can you send push notifications to people who uninstalled the app?

No. Push reaches an app installation, so once the app is removed there is nothing to deliver to, and the same is true in practice for people who turned notifications off. Those users can be reached only through channels they separately agreed to, usually email and, where they consented, SMS, or through paid media. The SMS vs push comparison routes win-back for uninstalled users to email first and SMS occasionally, with consent.

Should re-engagement campaigns include a discount?

Not by default. A discount sent to every lapsed user pays for some returns that would have happened anyway and teaches people that leaving pays. Start with a reason to return that has nothing to do with money, such as something new for that person or something they left unfinished, and test whether an offer adds anything with the win-back offer test, which compares an offer, a message without one and a holdout.

Terms used on this page

Frameworks

Where this page is used

Also related

Sources

  1. Firebase documentation — Google
  2. Google Ads Help — Google Ads Help
  3. TikTok Ads Help Center — TikTok for Business
  4. Direct marketing guidance — Information Commissioner's Office (UK)
  5. CTIA messaging principles and best practices — CTIA

Related