#030 | 16 June 2026

Main Story

Your CEP doesn't render your app

CleverTap, MoEngage, WebEngage. Three platforms that every serious mobile growth team has either integrated or evaluated. All three are genuinely strong at what they were built for: ingesting user events, building behavioral segments, firing the right message at the right moment across the right channel.

None of them were built to render UI. That happened later, as an extension of push notification delivery. The reasoning made sense at the time. The architecture left a structural gap that product and growth teams are still paying for today.

Here is the core problem. When a CEP adds in-app messaging, it is adding a rendering capability on top of a system whose entire underlying investment is in data pipelines. The in-app template editor is a feature on a customer data platform, built by a team whose primary job is segmentation, not screen performance. That origin determines every constraint that follows.

A tool built for orchestration that tries to own rendering ends up doing neither well.

The Template Tax

When your growth team builds in-app campaigns inside a CEP's native editor, they're working inside that platform's rendering constraints. Call it the template tax: the accumulated cost of everything you cannot do because the platform decides what can and cannot be rendered.

It shows up four ways. Design constraints that force third-party overlays visually disconnected from your app's actual design system. Creative iteration that stalls because every visual change requires going through an editor bounded by what it was built to support. No access to custom interaction patterns like scratch cards, streak trackers, or gamified reward reveals. And campaign logic that lives outside your codebase, which means A/B tests on visual treatments depend on the platform's tooling, rollbacks depend on the platform's uptime, and any visual change involves someone else's deployment pipeline.

The Performance Gap

The performance numbers are specific. MoEngage waits until images fully download before showing the in-app message. That's rational behavior for a system built around push delivery. It's wrong behavior for an in-app experience that should feel native to the app.

WebEngage has been criticized by teams for trigger delays of 5 to 10 seconds. A 5-second lag between a user action and an in-app response is not a nudge. The user has already moved on. Purpose-built in-app rendering systems fire in under 100ms. That gap isn't about engineering quality, it's about what a platform was designed to prioritize.

The Lock-In That Doesn't Get Discussed

Audience logic can be exported when you switch CEPs. Journey flows can be recreated. The actual rendered experience, the visual templates, the interaction patterns, the conditional display logic, does not export. It lives in a proprietary editor and gets rebuilt from scratch.

The better architecture decouples this. Segmentation and journey logic stay in the CEP. The rendering layer runs separately. When you eventually switch CEPs, you reconfigure the trigger integration. You don't rebuild the experience layer.

What Separation Actually Looks Like

The CEP continues doing what it does well: data ingestion, segmentation, journey orchestration. When the CEP's logic determines that a user qualifies for an in-app experience, it fires a trigger. A dedicated rendering SDK receives that trigger, renders natively from pre-built components already in memory, and returns control to the app in under 100ms.

Airbnb runs 500 or more concurrent experiments using this architecture. Lyft reduced experiment delivery from two weeks to two days. The speed gain is not engineering cleverness, it's what happens when you stop asking one system to own two fundamentally different problems.

The growth team still configures everything from a single dashboard. The difference is that the dashboard talks to a rendering system built specifically for in-app experiences, not adapted from a push notification engine.

The bottom line: your CEP is valuable because of its data infrastructure. Don't ask it to also be a UI rendering engine. The teams running the fastest experimentation cycles are the ones who separated those two responsibilities, and built the rendering layer on something purpose-built for it.

What’s new in Digia?

Free Tool: Generate publish-ready release notes in under 60 seconds

Writing release notes is one of those tasks that takes 45 minutes and produces a document that reads like it was written in 10. Version number buried mid-paragraph, changes listed in the wrong order, no consistent format across releases, and then someone asks for the same thing in Markdown, HTML, and Slack-friendly plain text.

The new Digia Release Notes Generator handles the structure so you don't have to. Enter a version number, a release date, and your changes across six categories: New Features, Improvements, Bug Fixes, Breaking Changes, Deprecations, and Security Updates. The live preview updates as you type. When you're done, export to whichever format your workflow needs.

v2.4.0 · June 15, 2026
New Features → 2 items
Bug Fixes → 3 items
Breaking Changes → 1 item
Contributors → Platform Team, Developer Experience Team
Export → Downloaded as structured .md in one click

Four export formats: Markdown for GitHub, GitLab, and Notion. HTML for documentation sites and email. Plain Text for Slack and Jira. Keep a Changelog format for open-source CHANGELOG.md files. The preview switches between all four without rewriting anything.

Socials

News

Anthropic puts its most powerful model in public hands.

On June 9, Anthropic launched Claude Fable 5, the first publicly available version of its Mythos model. The company says Fable 5 leads on software engineering, knowledge work, and vision, but ships with hard safety limits: in high-risk areas like cybersecurity, biology, chemistry, and distillation, the model blocks responses and falls back to Claude Opus 4.8. Anthropic also released an updated Mythos model on the same day, called Claude Mythos 5. It runs on the same underlying model as Fable 5 but with those safeguards lifted in select areas, and remains available only to a small group of vetted partners.

Anthropic described Fable 5's capabilities as exceeding those of any model it has previously made generally available, and stressed that the public release is possible only because of the classifiers built to block misuse in high-risk domains. The company ran an external bug bounty with over 1,000 hours of testing and worked with red-teaming organizations before going public, with no universal jailbreaks found. Fable 5 is available through the Claude API and usage-based Enterprise plans. Until June 22, access is included at no extra cost across Pro, Max, Team, and seat-based Enterprise plans. After that, usage credits will be required, with Anthropic saying it plans to fold the model back into standard subscription plans once capacity allows.

Your features are only valuable if users adopt them.

AI makes it easy to build new features. But building isn’t the bottleneck anymore - discovery and adoption are. If users don’t encounter a feature in the right context, at the right moment, it simply doesn’t get used.

The result? Missed engagement and wasted revenue opportunities.

Digia solves the distribution problem.

Ship in-app experiences directly on top of your existing data stack - without waiting for an app release cycle or forcing updates.

It works seamlessly with CleverTap, MoEngage, WebEngage, and other CEP tools.

No code changes.
No release cycle.
No Play Store or App Store update.

Your feature or nudge goes live instantly and your data stays where it belongs.

Teams at BBlunt, Dezerv, and Omli use Digia daily to ship experiments and full features without pushing app updates.

Try Digia for free → Digia Studio

Keep Reading