微笑着给买家制造痛感：Julia Schottenstein 谈 dbt 如何胜出

Schottenstein describes an unusual path from venture capital into product. At NEA she invested in early-stage startups building for technical audiences, and in 2019 she discovered dbt, an open-source data transformation framework. Users described it as an identity rather than a tool, and with cloud data warehouses exploding, she thought that if dbt worked it could work in an extraordinary way.

In 2020 founder Tristan called to say he was raising money, but she lost the deal to Sequoia. She was so convinced that she asked to put nearly 20% of her liquid net worth in personally, but the board vetoed it. A few months later she asked Tristan whether she could join the company that built the product she believed in.

Schottenstein uses the same four criteria for joining an early company as for investing in one. On people, she asks whether you trust the founder to lead, and she was impressed that Tristan could paint a compelling future for the industry while being detailed about day-to-day analytics engineering. On market, she looks for growth and room for a new entrant, noting that the chaos of cloud data warehouses gave dbt the chance to bring structure.

For product, she talks to users and listens for a spark of enthusiasm, such as people who cannot stop talking about it and want to share it with teammates. Distribution, she argues, may matter even more than a good product. dbt had an ecosystem advantage and low-friction open source, while an enterprise-sales company should have a team that knows how to land complex deals.

No company scores 10 out of 10 on all four, she says. When you join, you contribute your time, so ask where the company is weakest and what you bring that could de-risk it.

Schottenstein says the time to think about M&A is when you do not need it. Founders start companies to build enduring independent businesses, and a viable path to stay independent gives you the upper hand in every conversation. Most companies lack that path, so M&A is about creating plan Bs.

For any company, she says, only two or three buyers will find it extremely strategic. Identify where you hold a competitive advantage and inflict pain on those buyers, so that they cannot help noticing you. Crucially, do it in a friendly, open way, because many founders shut down conversations with incumbents by taking too competitive a stance and lose their optionality.

Her example is Transform, which dbt Labs acquired and announced in February. Transform was a pure-play semantic layer with strong technology and Airbnb-Minerva-style roots, and it pressured dbt Labs by being loud about solving hard problems, while dbt had distribution but lagged on product. Because Transform positioned itself as a friendly partner and had already done integration work, the acquisition was easy for dbt to embrace.

dbt Labs recently codified three pillars for competition: hold true to the vision and ignore noise, grow the pie with ecosystem partners, and lean into strengths. The company wants to be a platform, holds its ground on the transformation standard and the semantic standard, and leaves space for partners elsewhere. Schottenstein credits Nick Handel, Transform's founder, with leading that exercise.

She names power with simplicity and a commitment to being open as dbt's main strengths. Critics said they had a SQL templating tool in-house, but founders Tristan, Drew and Connor believed that people who know SQL and work with business stakeholders should be able to create clean production data. Open source reduces friction, produces diverse use cases and network effects, and lets partners build around the roughly 20,000 companies using dbt each week.

When Lenny asks about timing, Schottenstein says dbt arrived as cloud data warehouses boomed. It began as Fishtown Analytics, a consulting firm, and the consulting part of the business lasted almost two years. Whenever the founders met friction while serving clients, they built the fix into dbt, which matured the product.

Schottenstein tells how she, inspired by an ant-farm chapter in Gödel, Escher, Bach, ran a team offsite using rope and sticky notes. Each engineer represented a node in a data transformation graph and the rope its edges, and the group walked slowly through a new algorithm. She wanted every engineer to own the change and anticipate edge cases, rather than letting a few people run ahead.

The project changed how dbt Cloud runs a customer's DAG from an imperative approach to a declarative one, working backward from what must happen for data SLAs to be met in time. She says she keeps looking for memorable moments centered on the mission to give the team ownership.

On open source, dbt open source is the core of data transformation, where business logic is written, and dbt Cloud adds proprietary software for the development lifecycle and productionizing at scale. What goes in the cloud offering are stateful interactions and cross-team or structural collaboration. She stresses deciding which standard truly matters and keeping that open.

Schottenstein says many startups postpone pricing, which she blames in part on the zero-interest-rate era when investors funded GitHub stars and usage. She cites the idea of Madhavan, author of Pricing Innovation, from Pricing Innovation that you do not get to decide whether to have a willingness-to-pay conversation, only when. It is better before you build the product than when sales struggles to sell it.

dbt Labs has a core value of being more concerned with value creation than value capture. Customers say dbt is worth 20 to 35% of what they spend on their cloud data warehouse, and dbt charges a small fraction of that by design. Last year it made its first-ever pricing change, which let it test price elasticity, and she notes that dbt Cloud most often loses deals to dbt open source, which the team accepts.

The pricing work was cross-functional, involving finance modeling, customer conversations, product and communication, with close tracking of conversion and churn. The team spoke with dozens of customers, using product and product marketing, and since people rarely share what they would pay, they used relative value against the warehouse and asked what price feels cheap, fair or too expensive.

Lenny cites Hunter Walk's suggestion that founders be public about selling. Schottenstein agrees: in a Hail Mary situation, transparency and a wider net are better, and with so many companies in the same spot there is no shame in a simple note saying an exit is being explored, the product is interesting and the team wants to stay together. Buyers in these cases are mostly acquiring teams, so a clear data room matters.

Lenny shares how a chance meeting with Airbnb's head of product at a party a year earlier led to the acquisition of his startup. Schottenstein urges founders to make sure buyers know them well before an acquisition, and to use corp dev teams, whose job is to meet everyone, as a route to a product or GM sponsor. For founders in dire straits she advises using investors' networks, adding that investors know many investments return nothing.

On phrasing, she says not to be too clever: everyone knows that evaluating strategic alternatives means selling. If you have runway, avoid saying M&A and talk about partnerships and collaboration. She expects the market to pick up as valuations from the November 2021 peak reset and buyers come off the bench, and she suggests building a list of about a dozen buyers and focusing on two or three.

Schottenstein lists dbt Labs' values: value creation over value capture, transparency (including sharing board decks and having hard conversations in the open), humility, and treating work done well as its own end. Lenny links this to customer-first values at Figma and Snowflake and notes that what separates companies is practicing them. She adds that many employees came from the community and feel ownership because the product changed their work.

On working with opinionated communities, she admits it gets harder as the community grows, but over 30% of dbt Labs' headcount has contributed to its data transformation workflow, across product, data, marketing and engineering. Employees are also in the 50,000-member dbt Slack community and feel it directly when an experience falls short.

She is not a framework person but repeats two sayings: worse is better and tech debt is a champagne problem. The first dbt Cloud scheduler was a big loop over a jobs table that embarrassed the team but worked. It has since been rebuilt several times and now serves 8,000 companies and 10 million runs a month, which a distributed scheduler at launch would not have justified.

Asked what PMs should lean into, Schottenstein draws on her investing years. She still invests in her network, building a group of operators at slightly more advanced companies and asking how they handled open source, pricing and acquisitions. She also calls herself a T-shaped generalist whose breadth across finance, business and product helps her get things done.

The third habit is thinking in power laws and risk. In investing most bets return nothing and a few rare winners pay for the losses, and in product she tries to keep making bets with uncapped upside, even though she says it would be a problem if half of her work went to zero. In the lightning round she recommends Range, Quiet and biographies, and says her team's small process change was doing fewer things and single-threading around one mission.