In 2025, most lifecycle marketing programs are laser-focused on what to say and when to say it. But there’s a key lever that often gets overlooked:

How often should we message a user?

Back in 2017, Pinterest launched one of the most advanced notification systems of its time — built not to maximize short-term CTRs, but to protect long-term user engagement. And the results still deserve attention today:

- ✂️ Notification volume dropped by up to 24%
- 📈 Email click-through rates jumped by as much as 31%
- 🔁 Daily active users (DAU) increased among previously under-engaged users
- 📣 Among users who received both email and push, push CTR improved by 21% with less volume overall

The takeaway: Less can be more, if you send the right messages, to the right users, at the right frequency.

Using AI to Solve the Notifications Volume Problem

Pinterest’s challenge was clear: how do you send billions of notifications across email and push — without hurting retention?

Their answer wasn’t more targeting. It was rethinking notification volume entirely. Instead of manually-tuning frequency rules or relying on CTR to guide volume, they built a machine learning-based messaging system with a bold foundation:

- 📊 A weekly notification “budget” per user, dynamically allocated based on long-term engagement potential
- 🤖 Machine learning models to predict both the upside (return visits) and downside (unsubscribes, churn) of sending messages
- 🔄 Decoupling volume control from message ranking, so new campaigns wouldn’t skew results or introduce unintentional volume spikes
- 📉 Recognition of diminishing returns—sending more doesn't always mean getting more

Instead of rewarding users who always clicked with more notifications, Pinterest prioritized incremental utility: who benefits most from each additional message?

That insight mirrors how today’s best CRM programs approach personalization — not just identifying who's active, but who can be influenced by another notification.

Why This Still Matters in 2025

At Grohaus, we work with lifecycle marketing teams who are forward-thinking about automation, experimentation, and data-driven marketing. But even in mature LCM programs, we regularly see:

- 🔁 Overlapping campaigns that unintentionally hit users multiple times a day
- 📉 Spike-and-dip performance from teams A/B testing message content without volume controls
- 🚫 Avoidable notifications opt-outs, or app uninstalls due to messaging fatigue

This is exactly why Pinterest’s system still matters. It reminds us that volume is its own variable — one that deserves just as much strategic planning as content or channel.

3 Takeaways for Lifecycle Marketing Teams Today

You don’t need Pinterest’s engineering budget to apply these ideas. Here’s how LCM teams can adapt this thinking:

For examplary purposes, we'll use D2C eComm brands — especially those rich in SKUs and content — for the following tactical advice.

1. Start with a Weekly Contact Strategy

What to do:
Build a weekly per-user contact cap that spans all channels (email, SMS, push). Many D2C brands manage messaging at the campaign level, but your customer experiences your brand holistically. Start there.

Experiment to try:
A/B test two global frequency strategies for users who are currently receiving 5+ messages/week:
- Control Group: No change (keep sending based on existing campaign logic)
- Test Group: Cap total sends to 3/week, prioritizing highest value revenue-driving campaigns (e.g., abandoned cart, product back-in-stock, etc.)

Track differences in:
- Unsubscribe rate
- Purchase frequency post 60 days
- Click rate

Pro tip:
Use campaign priority logic, or journey suppressions to enforce caps. The best cross-channel marketing platforms have easy ways to manage frequncy at the user level; Braze’s Frequency Capping Rules, or Customer.io’s Message Limits are great examples.

2. Optimize for Value based on Engagement Level

‍What to do:
Pinterest found the biggest engagement gains among users who were lightly engaged — not totally inactive, just not fully activated. These are the users who need just a good reason to come back. You might want to experiment with a similar cohort in your business.

Experiment to try:
Build a custom segment of “marginal” customers:
- Visited site 2–4x in past 30 days
- Clicked 1–2 marketing emails
- No purchase in last 60 days

Then run an engagement experiment:
- Variant A: Typical promotional emails (e.g., “20% Off This Weekend”)
- Variant B: Content-rich campaign (e.g., “How to Style Your Summer Fit in 3Minutes” or “4 Products You Didn’t Know You Needed for Your Home Gym”)

Track:
- Click-through rate
- Purchase conversion rate
- Site visit frequency post 7 days

Pro tip:
These users might not need another discount promotion — they might need inspiration instead. You can create entertaining content to spark curiosity, and then let AI predict who would react better to your new content based on lookalike models. We explored how AI decisioning is available for customers of Braze and Iterable in a separate article.

3. Separate Content Testing from Send Logic

What to do:
When you test subject lines or email layouts, make sure you're not also testing send logic (like frequency or timing). Volume changes can inflate performance and mislead results.

Experiment to try:
You’re testing a new product recommendations email template for a month. Test it head-to-head against your control creative, without changing anything about timing or frequency:
- Same send times
- Same send volume
- Same audience criteria

Control: Current email template
Variant: New modular layout with dynamic product recommendations

Track:
- Click-to-open rate (CTO)
- Revenue per recipient
- Product views per user

Pro tip:
Avoid “multi-variable” tests, especially if you don't have a complex ab testing infrastructure and a large user base. If CTR jumps for the new template design, make sure it’s because your content worked — not because you sent it more often, or to more people.

Final Thoughts: The Message and the Medium Matters

Pinterest’s system wasn’t just about sending fewer messages. It was about sending valuable messages, with a focus on long-term outcomes. Their use of machine learning models and per-user budgeting wasn’t just cutting-edge—it was a practical approach to a complex problem.

And while their tech has evolved since 2017, the principle holds: Companies should respect users' time and attention as much as they respect their metrics.

Want help building a smart strategy that optimizes volume and boosts engagement — without burning your list? Let’s talk.

‍

‍

‍

‍