Ilja Torski

Someone who has stopped using an app may still be reachable; behaviors that suggest a decision to leave deserve a separate, faster response. They also caution that winning back deleted users may not pay off equally across app categories.

In a fitness app, four workouts last month and none this month indicate inactivity.

• The same user opening settings twice, then disappearing for ten days, may be showing a pre-deletion signal.
• Torski recommends treating those patterns differently.

Read original

Torski argues that teams should compare messaged users with a group that receives no message, then report the difference rather than claiming every treatment conversion.

In a gaming LiveOps campaign, Torski reports 5.60% conversion among 80,000 messaged users and 4.20% among 8,500 controls: a reported 33.3% uplift. They say messaging drove 1,120 purchases, rather than all 4,480 treatment purchases.

Read original

That makes post-install behavior useful before the next user installs.

Torski suggests sharing which value propositions prompt re-engagement, which feature users try first, and which acquisition sources still produce active users at Day 30. Those signals can shape store-page copy and screenshots.

Read original

Their examples connect web capture, email, push, and delivery updates to different moments of user intent.

An e-commerce visitor might submit an email for a first-visit offer, then receive push messages after installing. For an order, Torski recommends timely tracking updates, proactive delay notices, and a post-delivery follow-up.

Read original · Original 2 · Original 3 · Original 4

Their argument is that relevance depends on who is affected by a platform feature, and that messages to AI-mediated shoppers may become less frequent but more consequential. They frame the tension as keeping users informed and in control without recreating notification fatigue.

Torski says language support could leave some global app users outside iOS notification prioritization, making audience segmentation important. In agentic commerce, they imagine messages confirming a purchase within a user’s price limit or asking them to approve a better option.

Read original · Original 2

Start here to understand the framework behind the other posts.

Torski separates two states that retention dashboards often merge.

• Inactivity means the app is still installed, so push notifications or behavioral triggers can still reach the user.
• Deletion is a decision: afterward, only email or SMS remains if the app collected that information.
• They cite 62% of installed apps going unused in a given month and over half of downloads being uninstalled within 30 days.
• In a fitness app, someone who logged four workouts last month and none this month is inactive; someone who logged four workouts, opened settings twice, then stayed away for 10 days is showing a pre-deletion signal.
• Torski recommends segmenting inactivity by depth and pattern, treating pre-deletion signals as a separate high-priority trigger, and being realistic about win-back economics.

For brands planning a campaign

Use this as the opening framework for a webinar or article series on managing user inactivity, with the fitness example as a diagnostic exercise.

Torski explains why A/B tests alone can overstate messaging impact.

• A control group withholds messaging from a segment to show what would have happened anyway.
• In the example, a gaming client sent a LiveOps event push to 80,000 users in the treatment group and 8,500 in the control group, then measured in-app purchases over the event window.
• Treatment conversion was 5.60%, versus 4.20% in the control group: a +33.3% uplift at 95% confidence, with 1,120 purchases directly driven by messaging.
• Without the control group, all 4,480 treatment conversions might have looked like campaign results, though 3,360 would have happened anyway.
• Torski also lists subscription re-engagement, e-commerce cart recovery, and fintech feature adoption as control-group use cases.

For brands planning a campaign

Use this for content aimed at growth and CRM teams defending messaging budgets. The gaming example provides concrete figures for an article or webinar slide.

Read it to see how they frame the next platform shift.

Torski argues that agentic commerce changes the purpose of messaging.

• Push notifications once triggered decisions; if AI agents make purchases on users’ behalf, decisions are already made, and messaging becomes the layer where people audit, approve, or override them.
• Nudge tactics such as low-stock warnings and limited-time discounts rely on human hesitation and distraction, which agents don’t share.
• They predict that messaging will shift from persuasion to protocol, in-app messages will become an explanation layer, frequency will collapse as importance increases, and the central tension will be autonomy versus control.

For brands planning a campaign

Use this in a future-facing webinar segment that moves beyond campaign mechanics to how messaging earns attention when agents mediate purchases.

Torski notes that teams often use ASO data to optimize CRM but rarely use CRM results to adapt ASO.

• CRM data shows which messages retain users, which features drive activation, and which acquisition sources produce loyal cohorts, yet it usually stays in the retention dashboard.
• They identify three data points to share upstream: the value propositions that resonate after install, the features users engage with first, and the acquisition sources that produce users still active at Day 30.
• Torski notes that a customer engagement platform such as Pushwoosh already has push performance by segment, first-session behavioral events, and onboarding funnel drop-offs.

For brands planning a campaign

Use this as the outline for an article or webinar on connecting retention and acquisition teams, with the three data points as a viewer checklist.

It’s a useful starting point for teams that focus on engagement only after install.

Torski presents web popups as a first-party data capture point that feeds the messaging stack before a user installs an app.

• A web visit often signals high intent: someone searches for a product, lands on the site, and reads about it.
• They offer examples by category: media can offer free articles in exchange for a subscription; e-commerce can offer a discount for an email address, then follow up with push; fintech can offer a rate estimate and start email nurture immediately; and games can offer a launch bonus before release.
• Torski recommends syncing the popup with messaging channels so each channel plays its part in sequence.

For brands planning a campaign

Use this as the pre-install section of a guide or article on owned-channel capture, with the four category examples as ready-made scenarios.

Torski highlights a practical constraint in iOS 27’s AI-powered notification prioritization: it sorts notifications by urgency and relevance only for users whose language is among 16 supported languages.

• Arabic, Hindi, Polish, Thai, Indonesian, Russian, Ukrainian, and Bengali aren’t on the list.
• For global apps in fintech, gaming, e-commerce, and delivery, this creates two realities within the iOS audience.
• For users of supported languages, generic broadcast messages are more likely to be deprioritized, making behavioral triggers and relevant copy more important.
• For users of unsupported languages, nothing changes on the OS side, so the messaging platform must deliver relevance through segmentation, send-time optimization, behavioral triggers, and personalization.

For brands planning a campaign

Use this in a webinar or article on global push deliverability, especially for teams that haven’t checked their audience by country and language.

Torski reacts to AppsFlyer's 2026 State of e-commerce report, which shows that 92% of iOS ad spend now goes to remarketing.

• They argue that brands spend heavily on paid re-engagement while underusing owned messaging.
• Most e-commerce messaging is transactional, such as order confirmations and shipping updates, but they point to upsell sequences, behavior-based cross-sell, and win-back campaigns.
• Behavior-triggered messages convert 3-5x better than broadcast campaigns, and owned channels avoid CPMs.
• They also warn against waiting until Q4 to re-engage dormant users: competitors who messaged consistently through the year will convert them first.

For brands planning a campaign

Use this as the market-context opener for an e-commerce retention campaign, then move into segmentation, triggers, and testing as roadmap priorities.

It is the strongest concrete case study in Ilja Torski's posts.

Torski describes a B2B2C case in which Hussle Technology sells a white-label sales enablement platform to direct and field sales brands, which roll the app out to their own distributor networks.

• The product only works if hundreds of thousands of individual sellers use it, so the company built lifecycle journeys like those in consumer apps.
• The example includes a 60-day onboarding sequence that branches on behavioral signals such as syncing contacts, making a first share, or getting a click back.
• A separate re-engagement track fires for distributors who go quiet.
• One template architecture is cloned across hundreds of brand deployments, making the system scalable.

For brands planning a campaign

A B2B2C article or webinar can use this as a full case study, discussing the 60-day sequence and cloned template architecture as operational details.

An inactive user still has the app installed, so push and behavioural triggers can still reach them; a deleted user has made a decision, leaving only email or SMS — and only if that contact data was collected.

• Torski illustrates the distinction with a fitness app: someone who logged four workouts last month and none this month is inactive, while someone who logged four workouts, then opened the settings page twice and disappeared for ten days is showing a pre-deletion signal.
• They recommend segmenting inactivity by depth and pattern rather than recency alone, and treating signals such as accessing account settings, viewing cancellation pages or filing an unresolved support ticket as a separate trigger category for immediate, high-priority intervention instead of a standard re-engagement flow.

Segment inactivity by depth and pattern, not just recency. Route pre-deletion signals into a separate, immediate intervention track rather than a routine re-engagement sequence.

Torski says email and SMS win-back works best in high-intent categories such as fintech and fitness, where users had a specific goal they did not complete. For entertainment or utility apps, prevention is the play because win-back returns rarely justify the effort.

For brands planning a campaign

A webinar or article series built on this distinction has a clear spine: how to define and detect pre-deletion behaviour, what intervention looks like at each stage, and how to make the economic case for prevention over win-back. A partner is buying a named framework with industry-specific examples, not general retention advice; the topic suits a session for CRM and lifecycle teams.

Examples: Source 1 · Source 2

On iOS 27's AI-powered notification prioritisation, Torski notes that the feature depends on the user's language, lists 16 supported languages and points out that Arabic, Hindi, Polish, Thai, Indonesian, Russian, Ukrainian and Bengali are not among them.

• For apps with audiences in the Middle East, South or Southeast Asia, or Eastern Europe, whether a notification is noticed or buried still depends on the messaging platform's own segmentation, send-time optimisation and behavioural triggers.
• Torski applies similar reasoning to agentic commerce: when AI agents make purchases, urgency-based nudges stop working, messaging shifts from persuasion to protocol (a purchase executed within a set threshold, or a request to approve a new option), frequency collapses, and almost every notification that remains matters.
• They frame the central tension as autonomy versus control: too little messaging and users feel they have lost control; too much and notification fatigue returns.

Check language and country coverage before assuming a platform feature applies to an audience. Treat unsupported segments as a separate strategy, with the messaging platform responsible for relevance.

For brands planning a campaign

This offers a current, checkable angle for collaboration: audit how much of an app's audience will actually be affected by an operating-system feature before building strategy around it, and design for the segment that is not. The agentic-commerce post provides a forward-looking hook for an article, panel or conference session without over-claiming what has already happened.

Examples: Source 1 · Source 2 · Source 3

In a gaming example, they set out 80,000 users in treatment and 8,500 in control during a LiveOps event: treatment conversion was 5.60% against 4.20% in control, a 33.3% uplift statistically significant at 95% confidence, with 1,120 purchases attributed to messaging rather than all 4,480 treatment conversions.

• They name further use cases for control groups in subscription re-engagement, e-commerce cart and post-browse triggers, and fintech feature adoption, and boil the question down to whether sending the message made any difference.
• In a separate post, Torski recommends testing messaging the way ads are tested, including the ability to segment by purchase behaviour and trigger campaigns from app activity.

Hold back a control segment and measure uplift against it over the same window, rather than counting all treatment conversions as campaign results.

For brands planning a campaign

Strong material for a practical, numbers-first article or webinar segment on incrementality: how to size a holdout, read uplift and report messaging return internally. It also shows that Torski can present a specific client result, not just a conceptual argument — useful if the collaboration promises demonstrated results rather than opinion.

Examples: Source 1 · Source 2

CRM knows which messages retain users, which features drive activation and which acquisition sources produce cohorts still active at day 30, yet that data often stays in the retention dashboard instead of reaching the app-store optimisation team.

• They name three data points worth sharing upstream: value propositions that resonate after install, the feature users engage with first in their opening session, and retention curves by acquisition source.
• Push performance by segment, first-session behavioural events and onboarding funnel drop-offs already sit inside an engagement platform, Torski notes, so the gap is usually process rather than tooling.

Feed post-install behavioural and retention signals into store-page and acquisition decisions. Treat retention by acquisition source as a measure of quality, not volume.

For brands planning a campaign

Useful for a collaboration aimed at growth or product teams, not just CRM specialists: a session on closing the loop between retention data and store-page or acquisition decisions, with concrete handoff points between teams. It broadens the appeal beyond lifecycle marketers.

Examples: Source 1

Most e-commerce messaging remains transactional (order confirmation, shipping, delivery), they argue, while behavioural triggers for upsell, cross-sell and win-back convert several times better than broadcast campaigns and carry no media cost.

• They extend the same logic to web popups as a first-party data capture point before install, with examples from media, e-commerce, fintech and games where a single email submitted on the website activates email and then push in sequence.
• Torski also addresses teams that already run their own CRM or marketing automation stack and want to replace an unreliable push provider without reworking the rest of their setup.
• They list what such teams typically look for: reliable delivery, GDPR and CCPA compliance out of the box, and SDKs that integrate with iOS, Android, Flutter and React Native.

Assign each channel a function and a stage of the user journey instead of repeating the same message everywhere. Sync web capture directly with the messaging stack.

For brands planning a campaign

A ready-made theme for an article series on owned-channel strategy: where messaging fits against paid remarketing, how to sequence web capture into push and email, and what infrastructure a growth roadmap needs for behavioural triggers and testing. It also reaches a second audience: technically minded teams evaluating infrastructure on price, deliverability and compliance.

Examples: Source 1 · Source 2 · Source 3

Torski describes brands investing in next-day or same-day logistics while order confirmations, shipping updates and delay notices lag behind, turning a customer's excitement into uncertainty.

• They recommend real-time communication: push and SMS the moment the parcel moves, proactive notices before support tickets accumulate, and post-delivery follow-up that turns a completed order into a review, cross-sell or repeat purchase.
• Torski frames the goal as turning logistics moments into engagement moments, linking faster communication to repeat purchase behaviour and lifetime value, and asking retail teams whether shipping anxiety is eating into customer value.

Communicate in real time across push, SMS and email at each stage of the delivery journey. Send proactive notices when a delay happens, not after a complaint.

For brands planning a campaign

A concrete seasonal angle for retail and delivery audiences: holiday shipping anxiety as a revenue issue, not a service issue. It supports a co-authored piece or webinar timed to peak season, and brands can localise the framing with their own logistics data.

Examples: Source 1 · Source 2

The product only works if hundreds of thousands of individual sellers use it, so the company built lifecycle journeys the way consumer apps do: a 60-day onboarding sequence that branches on real behavioural signals such as syncing contacts, making a first share or getting a click back, plus a separate re-engagement track for distributors who go quiet. It all runs from one template architecture cloned across hundreds of brand deployments, which they identify as the key to scaling.

Branch onboarding on real behavioural signals and clone a single template architecture across deployments, giving each brand its own journeys without rebuilding them.

Torski frames distributor engagement as a harder version of the same consumer problem: more stakeholders and less patience for guesswork.

For brands planning a campaign

This is the clearest example of a topic that spans B2B and consumer engagement, useful for collaborations targeting platform and channel teams rather than only consumer app marketers. It also shows Torski explaining a partner's architecture and results in a way that holds up to technical scrutiny.

Examples: Source 1

Posts are concrete and often quantified.

• Several end by naming a specific capability, plan or pricing detail of the platform they work with, with a case-study link where one exists.
• The writing is direct and example-led rather than abstract.

For brands planning a campaign

A collaboration should use a structured, example-led format with a direct question to the audience, which transfers well to article series, newsletter contributions and webinar segments. Campaign requests that supply a real data point or client situation are more likely to match Torski's existing writing than requests for general commentary.

Examples: Source 1 · Source 2 · Source 3 · Source 4 · Source 5 · Source 6 · Source 7 · Source 8 · Source 9 · Source 10 · Source 11

They tell readers about the push-only plan at $7 per 1,000 MAU and link the live pricing calculator, noting that if the number works "we can talk"; explain how web popups sync with messaging channels in Pushwoosh; and recommend routing Pushwoosh CRM retention data to ASO teams. The posts do not specify whether the tie is employment, agency work or another arrangement.

View example

The post showcases the company's lifecycle work.

View example

View example

Yuwen’s example of a subscription tier where a ten-day login gap in month four predicts an 85% chance of not renewing in month six reflects the same instinct as Torski’s pre-deletion signals: find the behaviour that precedes the loss, then intervene while the relationship can still be saved rather than sending a win-back after the fact.

• Both also distinguish reported numbers from real ones.
• Yuwen rebuilds attribution across Shopify, ad and email data because each platform claims the same sale; Torski uses holdout groups for the same reason.

Yuwen approaches the problem through data infrastructure: predictive models, data warehouses, deduplication logic and attribution models built for brand operators.

• Torski stays in the messaging layer: which channel can still reach a user, what the notification should say, and how the operating system or an AI agent changes what gets through.
• Read Yuwen to understand why your channel reports are wrong; read Torski to learn what to send and when.

Read original · Original 2

They get specific about where the first contact happens: a home-page-only popup with no extra discount, pre-purchase flows built on winning ad angles, and a deliberate refusal to trade bounce rate for subscriber count. Torski makes the same structural argument from the app side, treating web popups as a first-party capture point that feeds email and then push, and warning that paid remarketing is over-funded relative to channels a brand already owns.

Gallagher writes as a brand owner inside a supplement business, drawing examples from his own P&L and site behaviour; his channel mix is email plus Meta. Torski writes as a messaging practitioner, with client examples across gaming, fintech, media and delivery, and focuses on push and multi-channel journeys for apps rather than email-led e-commerce.

Read original · Original 2

That practical, build-it-and-describe-it approach matches how Torski explains segmentation, trigger categories and web-to-push handoffs: both describe working setups and the decisions inside them.

Biller focuses largely on how AI tooling is wired into agency operations and client delivery, with email platforms as the working environment.

• Torski focuses on the retention problem itself: who is at risk, which channel can still reach them, and whether the message caused the result.
• Mobile push is central to their work.

Read original · Original 2

Their posts show Claude-built CRM and project systems, a client onboarding flow with custom forms, documents and emails, automatic task creation, a daily AI review of leads across CRM, projects, WhatsApp and email that decides the next action, and client-facing tools such as a Klaviyo deliverability dashboard.

• For a series on running retention operations, the operational detail is the draw: how flows get built, how leads get worked and what a new-client onboarding actually contains.
• A second post shows the pre-launch state — welcome flow, browse and cart abandonment, campaigns built and waiting on stock — useful for a campaign about the work between signing and going live.

The tradeoff is distance from the buyer's core question.

• Biller's material is ecommerce- and email-first; it does not cover mobile app retention, push notifications, in-app behaviour or pre-deletion signals.
• A brief on managing user inactivity and pre-deletion signals would need another contributor for the app-side framework, leaving Biller to speak to operations rather than retention signals.

Examples: Source 1 · Source 2

One post argues that a "we miss you" email after churn is wasted spend, then walks through predictive decay signals, an engagement cliff defined as the number of inactive days that precede cancellation, and a value-first intervention triggered at half that threshold; the line "don't save the subscription; save the habit" is the kind of framing a webinar audience remembers.

• The other post is a worked attribution example: a $14M outdoor apparel brand where Klaviyo claimed $1.8M in email revenue against $1.1M in Shopify email-originated orders, resolved by pulling orders, ad spend, email events and session data into a warehouse and deduplicating — email landed near $803K and paid channels showed a 4.2x blended return.
• That pairs directly with a session on proving what messaging actually caused.

The tradeoff is context: Yuwen's examples are ecommerce and DTC, not mobile apps, and the measurement angle is attribution and warehouse deduplication rather than holdout test design for push campaigns. Their posts also close with a contact link for a consultancy, so a competing platform should expect a service-led rather than platform-neutral framing.

Examples: Source 1 · Source 2

They run a supplement brand and write as an owner: email is not a standalone revenue channel but an efficiency tool that fights the natural decay of Meta performance as audiences get colder; the popup runs on the home page only, deliberately without an extra discount, to avoid the bounce-rate hit that pushes CPMs up; pre-purchase flows follow the winning ad angles; and three campaigns a week go to different segments for new customers, second orders and subscriptions. For an audience that buys media, that operator view offers a different and useful angle.

The tradeoff is narrowness and platform distance.

• The material covers one brand's email, popup and Meta decisions, not mobile apps, push notifications or lifecycle signals.
• A series on user inactivity and pre-deletion signals would use Gallagher as the commercial-efficiency counterpoint, not the retention subject; ask another contributor to provide the app-side retention framework.

Examples: Source 1 · Source 2