May 31, 2024 · 26m · a16z
7 Ways to Boost Retention (Both Pre- and Post-AI)
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of The a16z Podcast, consumer and growth expert Bryan Kim presents seven actionable, low-cost retention frameworks for AI-native startups. By adapting proven behavioral mechanics from traditional consumer software to modern AI workflows, founders can significantly boost Day-30 retention rates and ensure long-term product survival.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. How this is scored →
The host as informed peer 3.0 Guest teaching 4.1 Guest disagreement 0.0 The host pushing back 0.3
speaking balance: gold is the host, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 17:15 Pet peeve on superficial streak gamification
The guest forcefully rejects conventional product views that reduce streaks to cosmetic fire emojis, insisting that true retention requires high-effort user actions.
Hardest push from the host ▶ 14:58 Host questioning custom vs default notificationsThe host challenges the guest's presentation by asking whether apps should let users granularly customize notifications rather than relying solely on team-designed defaults.
Biggest teaching moment ▶ 16:04 Guest explaining AI companion-initiated notificationsThe guest educates the host on Character AI's companion notification mechanics, explaining how an LLM bot initiating contact creates a unique emotional nudge for users.
The host holds their own ▶ 19:27 Host linking streaks to user identityThe host demonstrates product insight by framing decade-long Duolingo streaks as identity statements rather than simple mechanics.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The host as informed peer | Guest teaching | Guest disagreement | The host pushing back | Why |
|---|---|---|---|---|---|---|
| Overview: Seven Retention Mechanisms for AI Companies | 2 | 5 | 0 | 0 | The host introduces the topic and asks the guest to explain speed to core product value. The guest provides detailed examples from Google and Perplexity to illustrate model speed impact on D-30 retention. | |
| Method 2: Feature-Gated Onboarding | 3 | 5 | 0 | 0 | The guest breaks down feature-gated onboarding using Lapse and Viggle.ai as primary case studies. The host chimes in at the end to reframe the takeaway as creating product value worth friction. | |
| Method 3: Designing Reciprocity | 3 | 4 | 0 | 0 | The guest outlines give-to-get reciprocity mechanisms using BeReal's consumer dynamic. The host contributes by noting how AI tools alter creator consumption patterns. | |
| Method 4: Building Smart Notifications | 5 | 5 | 0 | 2 | The host actively probes the guest on whether apps should allow user-customized notification settings versus company-driven defaults. The guest explains Character AI's companion notification mechanics. | |
| Method 5: Keeping Streaks Alive | 4 | 5 | 0 | 0 | The guest critiques shallow streak mechanics, contrasting cheap likes with effortful Snapchat interactions. The host builds on the idea by highlighting how multi-year streaks become tied to user identity. | |
| Method 6: Give Them Summaries and Recaps | 3 | 4 | 0 | 0 | The guest explains how LLM summarization powers customized recaps like Oops Finance and Spotify Wrapped. The host expands on the concept by proposing financial percentile rankings for consumer AI tools. | |
| Method 7: Status for Power Users | 4 | 5 | 0 | 0 | The guest cites Eugene Wei's Status as a Service and Civit AI to demonstrate power user status mechanics. The host summarizes the segment by observing that all seven retention tactics are zero-cost product decisions. | |
| Key Takeaways: Cost-Effective Growth Strategies | 0 | 0 | 0 | 0 | Brief wrap-up and standard fund disclosure read by the podcast host. |
Statements from this episode (13)
Targeted Growth Mechanics Can Boost Day-30 Retention by 10-15%
“My estimate is that on average, if you employ some of these combination of these methodologies, you're able to move up your, what I would start with D-thirty. I think you can move that up 10 to 15%. Whoa. Which is, which goes from okay on my list of like 20% D…”
Neglecting User Retention Leads to a 'Burned Forest' Acquisition Crisis
“Retention is the, you know, bread and butter of key consumer products, because if you don't get users coming back and again and again, ultimately, you have what I call this, like, burned forest, Problem where you went and cut all the trees off, and there's no …”
Consumer Products Rarely Recover From Low Day-Zero Retention
“If the new user comes on on day zero and didn't find a product value that they're looking for, They're not gonna stay, and you just, like, lost that user, and it's very, very unlikely that you recover from a low D zero, D one retention. It's very difficult to …”
AI App Day-90 Retention Increased Consistently With GPT Model Upgrades
“There's an example of an AI-native companionship product where they actually did this test where they used GPT-one, GPT-two, GPT-three, and they actually tracked the cohort of user retention using each of the models. And it's very clear as the model quality im…”
Perplexity Integrates State-of-the-Art AI Models on the Day of Launch
“Whenever the state-of-the-art model comes out, he incorporates that into Perplexity on the day of, if not earlier.”
Feature-Gated Onboarding Increases Long-Term Retention Among Activated Users
“And I think what we are seeing is that if you actually gate the usage of the product, behind something that users need to do to unlock. I think that actually ends up leading to higher retention. You're exchanging lower conversion to actual product initiation, …”
Lapse Gated Product Usage Behind Mandatory Invitations to Five Friends
“And what they actually gated the product usage behind is not only a pretty lengthy onboarding process of how the product works, but also you can't use it unless you invited five new friends.”
Viggle AI Restricts Creation Behind a Three-Step Discord Onboarding Process
“Now what Viggle does is it's only available in Discord to create. There are three steps you need to go through to be able to create a Viggle. That includes reading the rules. That also involves actually going through the mechanics of how to build the right seq…”
Users Rarely Re-Enable App Notifications After Disabling Them
“I know this empirically, that once you turn a notification off, the chance of something turning that back on is extremely low.”
Effective Product Streaks Require Effortful Participation, Not Superficial Gamification
“One pet peeve I have on streaks though, is that people think of it as, oh, gamification. I understand streaks. Let's, let's put a fire emoji and a number next to it and it'll be great. It's a, the well-done streaks are a little different. They're very, very ef…”
Snapchat Streaks Require Three Consecutive Days of Video or Image Exchanges
“What that meant on Snapchat is I sent you a video, not a text, not a chat, a video, a snap, or image, image or video, and you sent me back something in a 24 hour period as well. And streak doesn't appear until three days of doing that.”
Bestowing Status on Power Users Drives Product Retention
“And this is interesting, and this is helpful to build retention, because by giving people and bestowing some kind of status, if you will, for power users, the more likely you're able to incent them to continue that status.”
Civitai Relies on Status and Leaderboards for Creator Retention
“One of the companies that sort of focuses on fine-tuned image models and acting as a hub for that, there's a company called Civit AI. It's important for them to have a large, you know, density of fine-tuned image models. And so how do you actually, you know, e…”