David Lieb

1. Key Themes

Aggregate Metrics Are Actively Misleading for Early-Stage Products

David Lieb argues that the standard founder dashboard — DAUs, MAUs, and similar aggregate charts — flattens all user behavior into a single line that obscures almost everything actionable. The same graph can represent a healthy power-user core or a product in freefall, and you cannot tell the difference.

"One of the biggest mistakes I see founders make is relying on aggregate user metrics instead of understanding how any individual users use their product." 00:00:08

"If you were just looking at DAUs, this is the graph you would see. And it really doesn't tell you all that much. It basically tells you, yeah, we're not growing. We have some users." 00:08:14

The Dot Plot: A Specific, Named Tool Every Founder Should Build

The core tactical recommendation is to build a two-dimensional grid — rows for individual users, columns for days — and place a dot wherever a user performs the key value action. This is a concrete, buildable artifact, not a philosophy.

"You want to pick an event that your user does in the process of using your product that you think represents value in the product. Maybe it's sharing a photo if you're building a photo app or listening to a song if you're building a music app or processing an invoice if you're building a B2B invoice processing product." 00:02:13

"I would go so far as to say until you have hundreds of users, the dot plot could be your only dashboard." 00:12:29

Behavioral Pattern Recognition Outperforms Analytical Reasoning Alone

A key insight is that the human brain is exceptionally good at detecting visual anomalies in dense grids — better than it is at reasoning from summary statistics. The dot plot is deliberately designed to exploit this.

"What's really cool about this is it lets you figure out patterns that you probably would not have seen with your human brain, just looking at aggregate charts or looking at individual user logs." 00:03:34

"What you find when you look at this in aggregate, you can then kind of zoom out and see an entire page of these, is your brain will start to notice these patterns in a way that you would never have figured out on your own a priori." 00:06:22

The PayPal Fraud Detection Precedent: Visual Anomaly Detection at Scale

The technique traces back to a specific, verifiable origin. Max Levchin and PayPal used human-supervised graph visualization to detect fraud patterns no algorithm had yet been written to find — demonstrating the generality of the method beyond product analytics.

"This is actually an idea that I remember hearing about 10 years ago from Max Levchin, one of the founders of PayPal. They had a big fraud problem at PayPal when they first launched, but they didn't know the patterns to look for. So what they did instead is build a visualization, a graph of all the transactions that were happening on PayPal. And they just had humans sit and stare at screens of these drawings and graphs." 00:06:22

B2B SaaS Churn Is Predictable and Preventable With Dot Plots

Lieb makes the non-obvious case that dot plots are as valuable — or more so — for B2B products as for consumer apps. Enterprise churn is often caused by a champion leaving, and seat activation plus usage density would surface that risk weeks or months before renewal.

"The company bought 10 seats, but only three seats ever activated. Only three of those people ever tried the product. And if you look at their usage, they weren't getting a lot of value from it. Nobody used it more than two days per week." 00:11:05

"The company could have known that this contract was in jeopardy by looking at the dot plot." 00:11:35

Choosing the Wrong Event to Track Destroys the Tool's Value

The most common failure mode is populating the dot plot with vanity events (logins, app opens) that generate dense, feel-good charts but measure nothing about value delivery.

"A lot of founders might want to populate their dot plot with the easiest way to populate it so it feels good and you see a lot of dots. Maybe you'll pick like opened the app or signed into the product. Those are pretty bad events to choose because they don't really measure whether the user is getting real value." 00:12:00

Dot Plots and Cohort Retention Curves Are Complementary, Not Substitutes

The two tools answer different questions. Cohort retention curves tell you whether groups stick; dot plots tell you how individuals actually behave. Neither alone is sufficient.

"Cohort retention curves teach you in aggregate whether groups of users that you acquire stick with you over time. That's very important. You should definitely be measuring that. But the dot plot shows you how those users are actually using your product." 00:12:59

2. Contrarian Perspectives

Your Dashboard Is Probably Your Biggest Blind Spot

Most founders treat a rising DAU line as validation. Lieb argues the opposite: a stable DAU line can completely hide a collapsing retention picture, weekend-only usage, single-session churn, and feature dead-ends simultaneously. The conventional dashboard is not neutral — it is actively deceptive.

"With aggregate data, the graphs that we're all used to talking about, things like DAUs or MAUs, these lump all of your users together. And you can't really get a sense of what any individual user is doing." 00:00:38

One Piece of Paper Per Team Member Is a Valid Analytics Infrastructure

Against the assumption that analytics requires sophisticated tooling, Lieb describes printing paper dot plots and handing them to team members as a primary analytical workflow — not a stopgap.

"We would print out dozens of these pieces of paper with dot plots on them for different samples of our user base. I would print out a piece of paper and hand one of our team members like, here's the iOS users in France. I want you to understand what they're doing." 00:10:11

A Paid B2B Contract Is Not Signal of Product-Market Fit

A name-brand $80,000 annual contract that churns is worse than no contract — it creates false confidence. The real signal is seat activation and usage density, which Lieb's example shows can be near-zero even for a paying customer.

"I worked with a company in the most recent YC batch that had a very name brand customer that signed up and bought their product. I think it was like an $80,000 a year contract... The company bought 10 seats, but only three seats ever activated." 00:10:36

Picking a Wider Time Granularity Makes Your Product Look Better and Tells You Less

Founders instinctively choose weekly or monthly views to smooth noise, but this is precisely backwards — daily or sub-daily granularity is where the truth lives.

"The other mistake you can make is picking a time period that's too wide. Sometimes founders want to make it look better and they pick weeks like week one, week two, week three. It's way harder to figure out what's actually going on unless you look at it at the day or maybe even like sub day granularity." 00:12:00

3. Companies Identified

Bump

A mobile app (founded by David Lieb) that let users share contact information and photos by physically bumping phones. Mentioned as the live environment where dot plot methodology was developed and stress-tested in practice.

"At Bump, we had different symbols that we would put into these cells. So we knew whether you shared your contact information using Bump or if you shared a photo." 00:05:00

PayPal

Global payments platform. Mentioned because Max Levchin used human-supervised visual graph scanning to detect fraud patterns before algorithmic solutions existed — the intellectual origin of the dot plot concept.

"They had a big fraud problem at PayPal when they first launched, but they didn't know the patterns to look for. So what they did instead is build a visualization, a graph of all the transactions that were happening on PayPal." 00:06:47

Spotify

Music streaming platform. Used throughout as the illustrative example company for constructing dot plots — listening to a song as the canonical value-representing event.

"Let's say we're Spotify and we're building a music streaming app and we want to see how our users are using it. Let's pick the event that we're going to chart here being listen to a song." 00:02:38

GitHub

Developer platform. Mentioned because its contribution heatmap on profile pages is a publicly visible, widely recognized implementation of the dot plot concept.

"This idea of dot plots might be familiar to some of you. You've probably seen it at the top of GitHub pages. This is basically what a GitHub graph looks like. They've just wrapped the days around per week." 00:05:00

4. People Identified

Max Levchin

Co-founder of PayPal, later founder of Affirm. Identified as the originator of the visual anomaly detection approach that underpins dot plot methodology, applied first to fraud detection at PayPal.

"This is actually an idea that I remember hearing about 10 years ago from Max Levchin, one of the founders of PayPal. They had a big fraud problem at PayPal when they first launched, but they didn't know the patterns to look for." 00:06:22

Founder of Bump (acquired by Google); now a partner at Y Combinator. Identified as the practitioner who adapted visual user tracking into a repeatable founder tool and who coaches YC companies on this methodology.

"At Bump, we had different symbols that we would put into these cells. So we knew whether you shared your contact information using Bump or if you shared a photo. And it gives you a lot more granularity." 00:05:00

5. Operating Insights

Segment Dot Plots by Attribute and Assign Ownership to Individuals

Rather than one team reviewing one unified view, Lieb's practice at Google Photos and Bump was to slice the dot plot by user attribute (geography, platform, income bracket) and give each slice to a specific team member. This creates accountability for understanding a specific user cohort rather than diffuse ownership of a single dashboard.

"I would print out a piece of paper and hand one of our team members like, here's the iOS users in France. I want you to understand what they're doing. And I would hand another piece of paper to somebody else and say, these are the users on web in the United States who make more than $80,000 a year. Let's see what they're up to." 00:10:11

Use the Onboarding Marker to Isolate First-Cohort Behavior Immediately

Adding a distinct symbol for a user's first day of product use — Lieb suggests a ring around the dot — allows you to sort and filter to see only new users, revealing onboarding drop-off patterns that are invisible in aggregate new-user metrics.

"Another thing you can do to make a record of the first day that a user used the product, the day that they onboarded, you can put another symbol. Like let's say on a user's first day, we'll just draw a little ring around the dot like that just to give us a little bit more signal." 00:03:06

Build the Tool in Ten Minutes With AI Coding Assistants

The dot plot requires no analytics vendor, no data warehouse, and no BI tool. It is a log parser that outputs a 2D grid — a task modern AI coding tools complete in minutes, removing any excuse for not having it running.

"This is a thing that modern AI coding tools can whip up in like 10 minutes. These are best used in conjunction with cohort retention curves." 00:12:59

6. Overlooked Insights

Feature-Level Dot Plots Can Establish Causal Relationships Between Specific Features and Retention

Lieb briefly mentions encoding different symbols for different features — not just different users — which would allow a founder to visually correlate feature usage with subsequent retention. This is a lightweight, zero-cost alternative to formal causal inference or A/B testing for early-stage products, and the implication is significant: you can identify your retention-driving feature before you have enough data to run statistically valid experiments.

"If you change the dots to be different symbols, for example, in our Spotify example, we could choose to represent different features of the product... we could then infer like, oh, maybe the playlist feature is really causal to having people be really active in our product." 00:08:44

Champion-Departure Risk in B2B Is Quantifiable in Real Time — But Almost Nobody Measures It

The YC company example reveals something structurally important: the single biggest cause of unexpected B2B churn (the internal champion leaving) produces a detectable signal in seat activation and usage density weeks before the renewal decision. No CRM, no QBR cadence, and no customer success call would have surfaced this as quickly as a dot plot would have. The overwhelming majority of B2B SaaS companies do not track per-seat activation at this granularity, meaning they are systematically blind to their most common churn vector.

"The champion had gotten excited about this product and bought it. And then the champion left the company. And as soon as the champion left, a new person came in and they said, why are we using this software? We're going to churn. And so they opted out of a renewal clause at the last moment. The company could have known that this contract was in jeopardy by looking at the dot plot." 00:11:35