Raman Malik

The through-line

Raman Malik's consistent argument is that growth is not a marketing department with a bigger budget but an independent function that bridges product and marketing, with the user funnel itself as the product being obsessed over. "I see growth as an independent function that bridges product and marketing" [1], he says, and the two halves share a single goal: "Drive that user funnel from acquisition down to monetization, but your tools are just different" [1]. Growth product means engineering, design and data science iterating on acquisition, activation, retention and monetisation; growth marketing chases the same funnel through channels, life cycle communications, community and campaigns [1]. The insistence is that neither can operate alone: "I can't work on an onboarding experience without a product team iterating on every flow and without the marketing team thinking about copy, thinking about email drip campaigns. Everything comes in as one cohesive unit to bring this funnel to life" [1]. His scope at Perplexity covers exactly that spread of growth marketing, onboarding, activation, retention and monetisation [6][8][9][10].

The second recurring conviction is that compounding small gains and occasional high-risk bets are both mandatory, and that the discipline lies in knowing which mode you are in. "Everyone in growth is looking for alpha" [1], he says, and the search for alpha is competitive to the point of being binary: "There's no middle ground if you're not first. It literally just means your only other option is to just be a lot better than everyone else" [1].

On when a company earns a growth team

He resists the standard answer. The easy thing to say is post product market fit, "but like what is that even mean?" [1]. His practical trigger is early retention evidence: "as soon as you see that you have early signs in retention, like we're seeing like you know, if you see 30% retention in month three or four, you have the kindling in your fire where it is now worth it to go put gasoline on it" [1]. He is explicit that the investment need not be a fully staffed, multi-million-dollar organisation from day one. "It doesn't have to be we're going to go turn on marketing channels and spend a bunch of money. It could just be, okay, can we get that 30% retention to go to 40%?" [1].

What the first weeks of a growth role actually look like is unglamorous instrumentation work. At Perplexity, a very flat organisation, nothing was built out, though there were engineers ready to jump onto onboarding or retention [1]. The job began with mapping: "your first few weeks stepping into growth is you're just turning over stones. You're trying to build the map of the world" [1]. Asking what is happening in acquisition, engagement and monetisation produces no answers, only more questions, so "your job becomes let me build the right infrastructure so we can actually measure and understand the world", meaning engineering logging, attribution and first-party cookies [1].

On micro-optimisations versus big swings

Malik defends incremental work harder than most growth leaders do: "I think micro optimizations are seriously underrated" [1]. His justification is arithmetic. Build the growth model in a spreadsheet with new users, retained users, resurrections and churned users, and a ten percent improvement in retention or new user activation propagates through the whole system: "you see the entire water level of your active user base just increase. Like that small micro optimization just raised the entire water level of the company" [1].

He hedges that immediately, because the returns fade. "At some point, I'm not going to get to be able to squeeze out an incremental 10 15% without a lot of wasted experimentation time" [1]. Hence a cadence rule: "once a quarter, we have to be taking at least a couple massive swings, right? A big new feature, a big marketing campaign. It has to be high risk, it has to be high reward. I have to be able to stand in front of the company and say I'm going to do something and be willing to fail there" [1]. He sets the bar for those bets at roughly a 25 percent hit rate, with maybe one banger campaign or feature a year that genuinely drives growth or opens a new audience [1]. Viral sharing mechanics of the Spotify Wrapped variety are the archetype he points to for that kind of swing [1].

On the good, the bad and the ugly of A/B tests

Coming from a data science background, he splits testing into three categories. The good is a well-scoped test that "teaches you what you should work on next. It adjusts your direction. It adjusts your cadence" [1], and that catches metrics you are accidentally dragging down and clarifies what is genuinely incremental. The bad is a poorly scoped test with a low minimum detectable effect: "You're going to run it and you're not going to get anything out of it. You just wasted a week of your time" [1]. The ugly is political: "I'm running an AB test so that I can put something in my performance review about a number I moved", a pathology he sees increase as companies get bigger [1].

Testing does not replace judgement in his view; it sharpens it. "you should go into an AB test with a pretty strong confidence of what's going to happen. And this is where like intuition comes in" [1]. Without a prediction of which metric moves and which one might sag, "then you're just flying blind" [1]. The hardest cases are mixed results. His worked example: gating sign-in after a user's fifth query juices activation while some users hit the gate and bail, hurting query volume. Two metrics then compete, and the question becomes whether retained activated users repay the initial query loss over 30 or 60 days. Waiting that long is not an option, so "You kind of have to make a decision right there or figure out how to optimize and like protect that downside" [1].

On milestone metrics and metrics that get gamed

He calls the early behavioural predictors of retention milestone metrics, derived by looking back from users who retain at 30 and 60 days into their first sessions to find what they did differently [1]. For Perplexity, the marker is early query depth: "If I can get a user drawn Perplexity specifically to three queries in that first session, well, now I know I'm really onto something" [1], because that is enough time in product for the user to understand the value and come back. He concedes the imperfection openly, since it is not deep retention [1].

On the objection that any growth metric can be gamed, his answer is monitoring plus breadth. Check whether the metric still correlates with downstream retention once you have finished tinkering with it, because if the correlation breaks, "I'm not being honest with myself around if we're actually helping retention here" [1]. And refuse to stare at a single number: alongside query count he watches time in first session, including whether it hits a ten-minute mark, on the logic that follow-up questions mean a user is "now deep in the Perplexity experience learning and exploring", which is good for the product [1].

On acquisition, word of mouth and partnerships

The current environment flatters AI products: "AI companies have a ton of curiosity traffic right now", and anything with a magical demo will pick up traffic and quick usage [1]. Perplexity's own mix is roughly 80 percent organic word of mouth two years in, which he treats as the best channel a company can have, making the central question one of nurturing rather than replacing it [1]. Awareness still takes repetition: "You need someone to hear about your product three to seven times before they're going to give it an honest try" [1].

That repetition argument is why he rates distribution partnerships so highly. "We can go out and scream from the mountaintops about Perplexity, but it is a lot better when someone else is telling you to use it or it is being bundled into an existing product of yours. That's going to drive a lot more trial" [1]. On the economics of giving a free year of Pro to LinkedIn, Xfinity and Lenny's podcast audiences, he is unbothered: it is acquisition cost, usage is variable, "If you're not using pro, we're not losing money on you. But if you are using it, well, now we're capturing you as a user" [1], and the game becomes conversion afterwards. Asked what he still wants, he frames the answer as audiences rather than logos, naming students as the segment he most wants partnerships to reach at scale [1].

His clearest acquisition mistake was influencer marketing. He came in eager to test creators as a channel and "totally underestimated the amount of effort it is to work individually with creators on specific content and making sure where get the right messaging", compounded by the fact that Perplexity is hard to explain, since "me saying AI search engine means nothing" [1]. He treats such errors as a cost of speed: "If we're not making mistakes, we're not moving fast enough" [1].

On the founder-to-operator transition, and MBAs

Having spent three years grinding on a startup that eventually shut down, he is unsentimental about the emotional adjustment of joining someone else's company: "you are now well aware of how freaking hard startups are. Your ego is bruised, right? You just got your ass kicked" [1]. The harder structural change is narrowing: "founders enjoy the variety if I got to go work on 100 things and I get to steer the ship, but now you're hyper-focused on, okay, I am here to drive growth. I am de-scoping myself just to go do an excellent job at this" [1]. He describes the demoralisation of the founder years plainly, month after month and pivot after pivot with nothing moving [1], and chose Perplexity after conversations with Aravind, Dennis, Johnny and Dmitri because it was a company moving crazy fast and having fun [1].

On business school, he pushes back on tech's default disdain. He used the programme as a funded sandbox, never took an interview, tinkered full-time on ideas, and left a year early once investor conversations made it clear he would learn more going full steam ahead [1]. His recommendation is conditional: MBAs make sense for people breaking into a specific industry such as private equity, product management at Amazon or consulting, where there is a preset path and a community, making it "a very like expensive bridge to breaking into one of those fields" that is nonetheless valuable for many [1].

On presentations and goal-setting

Before Perplexity he founded and led Rhetoric, an app for getting feedback on presentations, positioned as a Grammarly for presentations, and he has discussed how it works, the growing importance of asynchronous presentation-giving, its main use cases and closing its seed round [2]. His own advice on presenting centres on two common mistakes: not soliciting enough feedback, and ignoring the importance of tailoring the sequencing of a presentation's topics to each audience group [2]. He has also described taking pride in defining goals carefully while staying open to revising them, in service of learning and iterating quickly [2], a posture that matches the experimentation cadence he later applies to growth.

Takeaways

• Growth is a bridge function, not a channel team: growth product (engineering, design, data science) and growth marketing (channels, life cycle, community, campaigns) chase the same acquisition-to-monetisation funnel with different tools, and onboarding work needs both at once [1].
• Build the growth function once retention shows kindling, around 30 percent retention in month three or four, and start by improving that number rather than switching on paid channels [1].
• Expect the first weeks in a growth role to be instrumentation: logging, attribution and first-party cookies, because the initial questions produce a hundred more questions instead of answers [1].
• Micro-optimisations raise the entire water level of the active user base through the growth model, but they hit diminishing returns, so commit to a couple of high-risk big swings per quarter with a roughly 25 percent success rate [1].
• Go into every A/B test with a prediction; without intuition about which metric moves and which sags, "you're just flying blind", and mixed-result tests demand an immediate call rather than a 60-day wait [1].
• Define milestone metrics by looking backwards from retained users; at Perplexity, three queries in a first session signals real intent, and guard against gaming by monitoring whether the correlation with retention holds [1].
• Distribution beats self-promotion: partnerships that bundle or endorse the product drive far more trial than shouting from the mountaintops, and a free Pro period is simply acquisition cost since unused Pro costs nothing [1].
• Two of the most common presentation mistakes are failing to solicit enough feedback and failing to tailor the sequencing of topics to each audience [2].