Product Playbooks
Decisions that shape the product itself — what to build next, when to say no, and how founders used real feedback to steer the roadmap without losing focus.
277 tactics · page 5 of 10
“Optimization comes much after your product market fit is there… we had product market fit in a big way on desktop but it took some years to kind of get that same stickiness on mobile where we saw okay our mobile apps our iPad apps are now like at a point where we can now start doing optimization so I think now the pendulum is turning.”
Optimization Only Starts After Product-Market Fit — Not Before
Microsoft had MMP tools early but couldn't get value from them because the mobile apps hadn't found product-market fit yet. Ramit's lesson: don't pay for growth infrastructure before the product is sticky. Once mobile PMF clicked, the same tools became high-ROI. Sequencing matters — optimize only what's already working.
“Our VP is very very strict about this like estimated impact any kind of vendor we have to bring in or anything that we have to do any kind of optimization he's like okay give me the estimated impact… the biggest ideas are the simplest ones… sometimes the simplest things like performance reliability file open speed size of the app those things matter a lot.”
Prioritize by Estimated Impact — Microsoft's VP Rejects Ideas Without a Number
With hundreds of millions of MAU, the Microsoft team rejects any idea that can't show estimated user impact. Every backlog item needs a number before it earns priority. Ramit's counterintuitive finding: the highest-impact ideas at scale are often the unglamorous ones — performance, reliability, transaction completion — not new features.
“I think there's a lot of hubris right like if I don't do anything with AI my investor will not fund me my career will go nowhere so I think that's a very scared approach the problem map and solution have to align.”
AI Must Solve a Real Job — Not Just Exist in the App Because Investors Expect It
Ramit saw many apps integrating GPT APIs without a meaningful use case — the same pattern as early databases bolted onto apps for novelty. His filter: does this AI capability (summarization, RAG, outline-to-doc) map to a real job users need done? AI that doesn't solve a specific pain point is a gimmick that erodes trust and dilutes focus.
“People really didn't want to engage not in the context of the class they weren't looking to take the class and then go into a community and talk about it… we're going to try again with a more contextual approach.”
"If You Build It They Will Come" Community Failed — Contextual Engagement Is the Answer
Zumba built discussion threads around classes — users could chat about their health journey or the workout they just took. Almost nobody participated. The insight: digital community needs to be embedded in the moment of experience, not offered as a separate destination after the fact. Their next attempt targets contextual interaction during the class, not a standalone forum.
“as a product guy i'm really convinced that our usage is really deep like we're starting from a different lego brick like okay you don't edit mask or square pixels you edit like objects so i mean any app that kind of want to copy that has to stop doing what it does today”
Object-Oriented Editing Is the Defensible ML Moat — Competitors Must Rebuild From Scratch
PhotoRoom's core product paradigm — editing semantic objects (cat, clothing, product) rather than pixels or masks — required competitors to completely rearchitect their apps to replicate. This isn't a feature but a foundational approach enabled by on-device ML. The depth of the paradigm shift is what makes the moat: copying a feature is easy; copying a new software metaphor is not.
“in december 2019 we dropped the video just for animation and then we dropped kind of the casual use case to focus on the pro and if you reach local maxima from one vertical like product market fit then you're investing so much on the deck it gets better than the all the local maximas are adjacent like you can reach them after”
Drop Every Use Case That Isn't Your Best One — YC's Local Maxima Principle
PhotoRoom cut video features and casual use cases in late 2019 to go all-in on pro/e-commerce photo editing. The YC framing: once you've reached local PMF in one vertical, adjacent verticals become accessible — but you can't get there without first fully saturating the first peak. Founders who try to serve every use case simultaneously reach no peak at all.
“if i take one example of a mid-size app like like you say we did build this in-house took us 18 months to get to first active actionable results i'd rather see in 2d with one eye than being completely blind like it's a no-brainer”
Building a Media Mix Model Took 18 Months — Still Better Than Being Completely Blind
Thomas Petit's team spent 18 months building an in-house media mix model before getting the first actionable results. The lesson is not that it is too hard — it is that the results are worth it even if imperfect. The new reality is blurrier than pre-ATT deterministic attribution, but blurry signal beats no signal. His heuristic: triangulate with both an in-house MMM and a paid incrementality provider to increase confidence when models converge.
“That was one of the biggest single-step functions in our revenue growth in our history was the day we launched subscriptions — it was also our first unlimited all-you-can-eat offering.”
Launching Subscriptions Was Burner's Biggest Single Revenue Step-Function
Burner started with a credit system but pivoted to subscriptions when the capability launched, pairing it with MMS support — a feature no competitor (including Google Voice) had yet. The combination of a new business model and a genuine product differentiator created the biggest single revenue jump in the company's history.
“I wouldn't want to own ten $1 million apps. I just don't think there's that much leverage in trying to scale things that don't connect to one another — whereas having multiple multi-million dollar apps that are complementary to one another it's much easier for us to add 5 million in revenue to our app.”
App Farms Don't Work — Complementary Multi-Product Does
Greg's experience with building and acquiring additional apps showed that standalone apps in new categories are nearly impossible to grow without an existing funnel. The leverage comes from cross-selling to existing high-intent users — not from operating an unrelated portfolio of small apps.
“The two areas — the biggest mistakes: under-trying to understand the size of the market you're addressing… and not understanding what the ongoing costs of an app are going to be… if I get any amount of volume I'm suddenly spending thousands of dollars a month supporting this app.”
Apps Fail for Two Reasons: Market Too Small or Ongoing Costs Not Modeled
David's post-mortems on 50 failed apps consistently trace back to two errors: an audience too niche to sustain a business, and backend costs that scale faster than revenue. Subscription infrastructure didn't exist for most of his weather app's life — so what should have been a SaaS was inadvertently running as a charity.
“It's also possible to just make cool things and have them have just enough of a business in them that it makes a good living for you, but you don't need all of that infrastructure. If you take the approach of simplicity and straightforwardness and craftsmanship early you can shift and pivot and change as you go.”
Start With Craft and Simplicity — Layer Metrics On Later
David offers an alternative to the data-first subscription playbook: begin with genuine craft, build something people enjoy using, and add monetisation when the app has proven its value. An over-engineered, metrics-driven approach from day one can eliminate opportunities that only emerge through iteration and user delight.
“We would actually ask that question as a part of a survey every week while we were growing RoboKiller. Companies fail because they don't find product market fit but they also fail to grow because they don't understand their product market fit.”
Run the PMF Survey Weekly — Product Market Fit Is a Dial, Not a Gate
At RoboKiller, Ethan Garr ran the PMF survey every single week and cross-referenced results against every product change — not to find PMF once, but to continuously deepen understanding of it. Most founders treat PMF as a binary gate they pass through. The more useful mental model: a dial you keep turning, with each product change moving it up or down for different user segments.
“Just being deliberate and saying every single day have at least one customer conversation is super powerful to stay plugged in and not just think of people as a bunch of numbers.”
Talk to a Customer Every Single Day — Data Without Context Runs Blind Experiments
Sean Ellis discovered this discipline at Logged Me In: a VC who kept asking 'when did you last talk to a customer?' forced him into daily conversations. The payoff wasn't just softer insights — he ran measurably better experiments because he had so much more context. The survey identifies which customers to call; the call gives the qualitative texture the survey can't capture.
“If the majority of the people you get in the door never actually experience the product then they contribute nothing to that north star metric. Activation — how do I actually get them to experience that value the first time — is a big part of that engine.”
Map Your Value Delivery Engine Before Optimizing Any Single Funnel Step
Sean Ellis's 'value delivery engine' diagram maps acquisition, activation, engagement, referral, and revenue as interlocking loops — each one able to move the north star metric. The exercise reveals where the real leverage is: most companies invest in acquisition while leaving a broken activation step that means most new users never reach the aha moment and therefore never count toward real value.
“The commercial strategy that you have — what people buy, when they buy it, and which price they buy it — is going to have such an enormous impact on your renewals and extensions.”
Churn is not just a product problem — the plan, price, and timing of the sale drives renewals
Product teams own churn in most companies, but the actual driver is often the sales funnel above them. Which plan a user picked, whether they came in on a discount, and whether they found the app through paid or organic all predict renewal rate better than in-app engagement metrics alone. Revenue strategy and product need to be looking at the same cohort slice.
“A lot of product teams are going to be like wow our product is amazing, we have increased our renewals by whatever 20% — but if they were to slice it: what is the cohort that bought natively in-app versus the cohort that bought through web, maybe the renewal rates are flat.”
Web checkout's higher renewal rates can be a mix-shift illusion, not a product win
Web-purchased subscriptions renew at significantly higher rates than in-app purchases in many apps — not because of product improvements, but because web attracts a different, higher-intent user. When aggregate renewal metrics improve after enabling web checkout, product teams need to segment by acquisition channel before claiming credit for a retention win.
“From your subscription business you can get money from people that never had a subscription and start new, people that upgraded, people that renew, or people that used to have a subscription, churned, and then came back. Having the history of how these four buckets evolve can tell you a lot.”
Map revenue in four buckets — new, upgrade, renewal, winback — to spot anomalies instantly
Rather than watching a single revenue line, decompose it into four cohort flows: new subscribers, upgrades, renewals, and reactivations. When a single bucket moves, it narrows the cause immediately — a spike in new subscriber revenue points to a pricing or acquisition change, not a product problem. This framework makes historical diagnosis fast and makes gaps in your experiment log visible.
“there might be something that's like medium to high opportunity but insanely high complexity and it's going to take months to build right and so at that point is that worth it or is the high like the medium to high opportunity with like a very low lift worth it right like that second option is going to be where you're going to want to start”
Size each experiment opportunity against engineering complexity before committing to the roadmap
Before finalizing the experiment roadmap, Musetti cross-references expected impact with engineering effort. A medium-impact, low-lift experiment will almost always beat a high-impact, high-complexity one on ROI. Starting with quick wins builds a culture of evidence-driven iteration without burning the team on bets that take months to evaluate.
“i've gone through dozens and dozens of apps screenshotted every single screen in their app put it all in figma and then use that as a starting point for the brainstorm to like get everyone's creative juices flowing”
Screenshot every competitor app screen into Figma to fuel experiment brainstorms
Before ideating experiments, Musetti builds a comprehensive visual library of competitor onboarding and paywall flows in Figma. This gives the whole brainstorm group a shared frame of reference and surfaces approaches no one on the team would have invented independently. Competitive immersion before ideation consistently raises the ceiling on the experiment list.
“If you're an indie developer and you're launching a brandless app last month it's very likely that app to web is a full zero — you can't do it all because you've got a small team, it's less likely to work because you don't have this recognition, and the analysis is actually quite tricky.”
App-to-web works for brands; for indie apps it is almost always a zero-sum detour
Thomas Petit draws a sharp line on the post-Epic-lawsuit app-to-web opportunity: established brands with existing tooling can capture real gains. For a small or newly launched app without brand recognition, the added complexity of different flows, pricing strategies, refund mechanics, and attribution math almost always outweighs any fee savings. His advice: wait until you have both the brand and the resources.
“They can come in and replicate the tool and have a feed but nah — we already have hundreds of thousands of passionate creators who have been riding with us, and you can't replicate that.”
A Facebook clone is a badge of honour — your community moat can't be replicated
When Facebook launched a near-identical rap recording and sharing feature (same colour scheme, same flame-emoji like button animation), Seth Miller laughed. The tool was commoditised almost instantly. What Facebook could not copy in an afternoon was years of community trust, shared culture, and the fact that engaged creators' output and social graph lived inside Rapchat. Community is the durable moat that survives product cloning.
“When you're trying to get to a billion users you're going to be more like 'let me really nudge you to share' — you focus on the top of funnel. Subscriptions focus you more on the bottom of funnel and that was a really big unlock.”
Subscriptions reorient product thinking from top-of-funnel to bottom-of-funnel
Growing toward a MAU-maximising model pushed Rapchat to optimise sharing mechanics and viral coefficients above all else. Once subscriptions were introduced, the team's attention naturally shifted to activation, retention, and the quality of the creator experience — the features that actually drive willingness to pay. Seth Miller found this reorientation clarifying: with a revenue north star, every product decision had a clearer test.
“You take a step back and you say: when my customers come to me, what is the ongoing problem they're trying to solve or the ongoing goal they're trying to achieve? And then you design the features and benefits to support them forever on their journey.”
Build a "forever promise" around the ongoing goal your subscriber is always trying to solve
Every successful subscription is anchored to an ongoing human problem, not a one-time transaction. Netflix's forever promise is 'the biggest selection of professionally created video content in the most efficient way.' VSCO's is 'make your best photos better.' Defining your forever promise first forces you to cut features that serve the product's elegance rather than the customer's ongoing need.
“Essentially what we ended up doing was triaging sort of a different funnel metric each quarter right so one quarter is like we gotta tackle bounce rate all right now we gotta tackle sign up rate now we've gotta tackle pro conversion rate now we gotta tackle retention.”
One Funnel Metric Per Quarter — Serial Focus Turned A Crappy App Into A Profitable One
From 2015 to 2017, AllTrails worked through bounce rate, sign-up rate, pro conversion, and retention — one per quarter — on minimal capital. By end of 2017 they hit profitability. This serial focus is the anti-pattern to roadmaps that spread effort across everything simultaneously. Constraint created clarity: what is the single weakest metric, and can it be meaningfully moved in 90 days?
“Nearly every subscription app i know has 90 plus of their free trials happen within the first 24 hours of install... the fact that all of the trials nearly all the trials happen within the first 24 hours make it relatively easier for subscription apps to have the signal be captured by skadnetwork.”
90% Of Subscription Trials Happen In 24 Hours — SKAdNetwork Captures Nearly All Your Signal
SKAdNetwork uses time-based timers: if no new event arrives within 24 hours of install, it sends back whatever conversion value it has. For subscription apps where 90%+ of trials occur within the first day, this means the algorithm captures the critical conversion signal almost completely. Compare this to gaming or e-commerce where meaningful purchases might happen on day 3, 7, or 30 — those signals are lost under ATT. The 24h trial window is a hidden structural advantage.
“We are just very frugal. I'm not gonna pay a margin of ads to an agency. We have a stable of really strong editors who do all the editing. I shoot with a teammate for two hours on a Sunday — that's the video for the week.”
In-House Creative as Competitive Moat — Frugality Beats Agency Margins
Speechify's four company values are product quality, speed, leading with love, and frugality — and frugality explains why they never outsource creative. Agencies add a margin layer without adding brand understanding. Keeping creative in-house means the people making ads are the same people who understand the product, the users, and the mission. This is the same reason they never outsourced engineering.
“I'm a big fan of imitate, iterate, and innovate. I would rather imitate to start because innovation is hard — you don't know how it's going to perform when it goes live. But imitation you can set benchmarks. We rolled out bedtime Bible stories and sleep Psalms — content people were already finding traction with in the secular space.”
Imitate → Iterate → Innovate — Set Benchmarks Before You Differentiate
Ryan Beck deliberately copied the secular subscription playbook (Calm's sleep content, Headspace's meditation style) and applied it to faith, setting clear benchmarks from the start. This gave Pray.com a proven content structure to iterate on rather than inventing from scratch. The innovation came later — once the baselines were established and they understood what worked for their specific audience, they built things that had no secular analog.
“We redid our whole product platform to allow for testing of onboarding flows, paywalls — having that from the server side so that we didn't have to deploy new apps every time we wanted to test something. We were able to run rapid iterations on our tests and that's really what helped us figure out what could monetize.”
Server-Side A/B Testing Infrastructure — Avoid Redeploying Apps for Every Experiment
Pray.com rebuilt their entire product infrastructure specifically to support rapid server-side experimentation — allowing onboarding flows and paywalls to be changed without a new App Store submission. This architectural decision compressed the feedback loop from weeks to days. For subscription apps where onboarding and paywall conversion are the primary revenue levers, the ability to test without a release cycle is a force multiplier for growth velocity.
“One of the key things I learned is: don't have different road maps. Make sure that your product road map is taking into account the whole organization. When performance marketing is closely related to the technical side of the business, it's all under one roof — and that created a lot of empathy where I was able to have a different perspective than most engineers.”
One Shared Roadmap Across Product and Marketing — No Engineering/Growth Divide
Ryan Beck ran both tech and performance marketing at Pray.com, which gave him unusual empathy for the marketers' needs and let him prioritize marketing infrastructure requests alongside product development on one shared roadmap. The result: no dysfunction between teams, faster execution on growth experiments, and a culture where every engineer thinks of themselves as a growth engineer. He cites Blinkist's CTO-led performance setup as the benchmark that validated this approach.
“We use data science to actually determine who is most likely to subscribe or take an action within our product — it takes a lot of the data signals based on what you're doing in the product, what we know about you, and merges those things together to give us a signal as to whether or not you are likely or unlikely to subscribe.”
AI Propensity Models Predict Who Will Subscribe Before You Ask Them
The Weather Company uses ML propensity models trained on in-product behavior and contextual signals (including weather conditions) to score each user's likelihood to subscribe. High-propensity users get more prominent premium prompts; low-propensity users are left to enjoy the free experience. The key caveat: models must be continuously tuned as economics and competition change — a set-and-forget propensity model drifts into misleading territory.