Julia Schottenstein on M&A Strategy, dbt Labs, and Why Worse Is Better — Notes

Source-grounded literature notes. Citations point to logical sections of the conversation, e.g. [§ Inflict pain]. Own words unless quoted.

Four questions [Adler frame]

Q1 — What is it about? Startup M&A, from someone who has sat on every side of the table — a former NEA investor turned dbt Labs product leader who led dbt’s acquisition of the semantic-layer startup Transform. The M&A material is the spine; wrapped around it are the dbt origin story, an open-core business model, a first-ever pricing change, and a handful of product-operating maxims.

Q2 — How is it argued? Prescriptively, but grounded in concrete cases. Each framework is stated then illustrated: the inflict-pain strategy through the Transform acquisition she ran; pricing through dbt’s first price change; team ownership through a rope-and-sticky-notes exercise; ‘worse is better’ through dbt’s embarrassingly naive first scheduler. The register is the practitioner’s, not the theorist’s.

Q3 — Is it true? The advice is internally coherent and unusually concrete, but the evidence is mostly n=1 and perspective-bound. The inflict-pain claim rests on one well-told case in which she was the buyer, so it describes what impressed an acquirer, not a measured seller win-rate. The numbers — ‘2 to 3 strategic buyers’, a buyer set of ‘maybe a dozen’ — are heuristics, not findings. The dbt-won analysis is retrospective and house-favourable. Read as experienced-practitioner heuristics with good face validity, not validated strategy.

Q4 — What of it? Several transferable moves. Build a company strong enough that M&A is optional — the only real negotiating leverage is the credible ability to walk [§ M&A as Plan Bs]. Identify your handful of genuinely strategic buyers early and inflict friendly pain in the area where you out-compete them [§ Inflict pain]. Have the willingness-to-pay conversation before you build, not when sales is stuck [§ Pricing]. Ship ‘worse is better’ and treat tech debt as evidence of usage [§ Worse is better]. In distress, be transparent and use the code words (‘strategic alternatives’) everyone already understands [§ Selling in a downturn].

Glossary

M&A as Plan B — Schottenstein’s framing: acquisition is fundamentally about manufacturing optionality. The strongest M&A position is not needing to sell, because ‘do nothing, stay the course’ is then a credible alternative. See Inflict Pain.
Inflict pain — the get-noticed tactic: compete hard in the area where you hold a competitive advantage so a potential acquirer cannot ignore you — but do it ‘with a smile’, positioning as a friendly partner to preserve the relationship and the option to be bought.
Strategic alternatives / strategic partnership — the code words. ‘Exploring strategic alternatives’ means you are selling; ‘strategic partnership’ is the soft opener when you still have runway and do not want to say the word M&A.
Buyer set — the short list (a dozen at most, often 2–3) of companies for whom your product is genuinely strategic. M&A outreach starts by building and then qualifying this list, not by mass mailing.
Corp dev — a buyer’s corporate-development team, whose job is to meet every potentially interesting company. Schottenstein’s tactic: take the meeting, decline to sell yet, but push them to introduce a deal sponsor in product or a GM.
Semantic layer / metrics layer — a system where a company defines its business metrics once, so every query returns consistent numbers. Transform was a pure-play in this; Airbnb’s internal version was called Minerva.
Open core — the business model: the standard (dbt’s transformation logic, where you write business logic in SQL) stays open source; the proprietary cloud product handles state, cross-team collaboration, and productionisation at scale.
dbt / analytics engineering — dbt (data build tool) lets anyone who knows SQL build clean, production data assets, work historically reserved for data engineers. dbt Labs began as the consultancy Fishtown Analytics.
Worse is better — shipping good-enough beats waiting for perfect, because real learning only starts once users touch the product. Borrowed from Richard Gabriel’s software-design essay and used as a PM maxim.
Tech debt is a champagne problem — debt accrued from shipping is a sign people are using the product; a luxury, not a failure.
T-shaped generalist — broad knowledge across business, finance, product, and markets, with depth in a few areas; the shape Schottenstein credits for her effectiveness moving from VC into product.
Value creation over value capture — dbt’s core value: charge a small fraction of the value delivered (customers value dbt at 20–35% of their cloud-warehouse spend; dbt charges far less by design).

Claims by section

§ From VC to product

Schottenstein ran the unusual path from professional investor (NEA, early-stage dev-tools/infra/data) into product. She discovered dbt in 2019 and was struck that users described it as an identity, not a tool. Reading the market — Snowflake tripling from $4bn to $12bn that year, cloud warehouses exploding — she concluded that if dbt worked it would work extraordinarily, and tried for a year to win the investment. She lost the 2020 round to Sequoia, asked to put in ~20% of her liquid net worth personally (the board vetoed it), then asked the founder, Tristan Handy, for a job instead. Her gloss: ‘if I was able to invest I wouldn’t be here.’

§ Four dimensions

Her framework for evaluating — or joining — an early-stage company: people, market, product, distribution.

People — chiefly the CEO: do you trust them to lead? Handy could paint a compelling industry future and go deep on the day-to-day analytics-engineering work; that range is rare.
Market — is it growing, with room for a new entrant? The cloud-warehouse chaos was the opportunity, because dbt brought order to it.
Product — can you hear the spark when you talk to users?
Distribution — often more important than product: do they have an unfair advantage in reaching customers? You want a company that is clearly strong at either bottoms-up/PLG or top-down enterprise.

Nobody scores 10/10 on all four. When joining, your edge is that you can dedicate time — so find the weak dimension and ask what you bring to de-risk it.

§ The spark

The product-market-fit tell is not just ‘holy shit, I want this’ but chatter: users who cannot stop talking about it and share it with teammates and other companies. That top-of-mind love does much of the go-to-market work, because evangelists are users who love what you have built.

§ M&A as Plan Bs

The time to start an M&A strategy is when you do not need one. The best strategy is a strong offence — building a genuinely independent company — because that gives you the upper hand in every M&A conversation: your alternative is to do nothing and stay the course. Most companies, though, lack a credible path to standing alone forever, so they must think about M&A — and M&A is always about creating Plan Bs and preserving optionality.

§ Inflict pain

For any one company there are only ever two or three buyers who find it extremely strategic. To get noticed: find the area where you hold a competitive advantage and inflict pain on that buyer — make it impossible for them not to notice you, so their ears perk up. The error she sees founders make is prematurely shutting the door — refusing to talk to an incumbent, taking too competitive a stance — which kills the optionality that is the whole point. So inflict the pain, but stay friendly and open; do it with a smile.

§ The Transform acquisition

dbt’s core is transformations; it was moving into the semantic layer (define business metrics once, serve consistent numbers everywhere). Transform was a pure-play semantic-layer company, technically strong, founded by ex-Airbnb people (Airbnb’s internal metrics layer, Minerva, is famous in the data world). dbt had distribution and ecosystem but was behind on shipping the product; Transform had the product but no distribution. Transform inflicted pain by being loud and technically credible on the hard problems — while positioning as a friendly partner courting dbt’s community. When acquisition time came, the friendly posture paid: dbt knew the product was good, and Transform had already done the integration work. dbt announced the deal in February [2023]. Transform’s founder Nick Handel later led dbt’s competition-philosophy exercise.

§ Competition philosophy

Three pillars, codified at dbt:

Hold true to the vision — competitors throwing shade is mostly noise; run your own race.
Grow the pie — work with the ecosystem to enlarge the opportunity (BI/reporting, then operationalised data, now clean transform assets feeding ML training) rather than slicing it thinly.
Lean into strengths — ambition to be a platform, but leave space for the ecosystem. The ground they will hold is the transformation standard and the semantic standard (better served together); everything else is partnership territory. The posture is long-term.

§ Why dbt won

Two things stand out to her: power-through-simplicity and a commitment to openness. Sceptics asked what was special about a ‘SQL templating tool’ — but the simplicity was the power: founders Tristan, Drew and Connor believed the analysts closest to the business should also build clean production data assets, and dbt’s framework made that hard to mess up while easy to start. Openness drives a flywheel and network effects: low-friction entry → users share it → more companies adopt → a diverse use-case base → a horizontal product → reinvestment → faster flywheel; 20,000 companies use dbt weekly, which pulls in partners who build on the standard. Two further enablers: timing (the cloud-warehouse explosion) and origin — dbt Labs began as the consultancy Fishtown Analytics for ~two years, building dbt to solve clients’ problems firsthand and folding every paper-cut back into the product.

§ Open-core boundary

What stays open source: the guts of transformation, where you describe business logic. What the cloud product monetises: stateful interactions, and any cross-team or structural collaboration, plus productionising dbt at scale. The discipline is deciding clearly what the open standard is (and protecting it) versus what proprietary software should supercharge.

§ Pricing

Most startups have the willingness-to-pay conversation too late — a side effect of zero-interest-rate funding of GitHub stars over revenue. She invokes Madhavan Ramanujam’s Monetizing Innovation (she calls it ‘Pricing Innovation’) [?source]: ‘you don’t get to decide if you’ll have a pricing conversation — only what’, so have it before you build. dbt’s value ‘value creation over value capture’ means charging a small fraction of delivered value (customers value dbt at 20–35% of their warehouse spend; dbt charges far less). Their first-ever price change taught them price elasticity while stakes were low; pricing is cross-cutting (finance modelling and customer conversations and product), and they were happy to keep losing deals to dbt open source. To find willingness-to-pay they talked to dozens, using relative-value framing and the cheap / fair / too-expensive price-point probe. See Monetizing Innovation.

§ The rope exercise

Inspired by a scene in Gödel, Escher, Bach where an ant colony acts as a computer flipping bits, Schottenstein ran a team offsite with a spool of rope and sticky notes: each engineer a node, the rope the edges, working through a new graph algorithm step by step so the whole team — not just the few racing ahead — internalised and owned it. The algorithm itself: shifting how customers’ DAGs run from imperative to declarative (reason backwards from the data SLAs that must be materialised). Her broader point: deliberately manufacture memorable, mission-centred moments so the team owns the work.

§ Worse is better

Two sayings she repeats: ‘worse is better’ and ‘tech debt is a champagne problem’ — both antidotes to perfectionism. dbt’s first cloud scheduler was naive — ‘a big old for loop over a big old jobs table’ — and they were a little embarrassed by it, but it shipped and got the job done. They have since rebuilt it several times because they have real scale (8,000 companies, 10 million runs/month) — but a distributed scheduler with RabbitMQ at launch would have been building for users they did not yet have. ‘We would be so lucky to have tech debt, because that means people are using the product.’

§ Selling in a downturn

She endorses Hunter Walk’s advice that, in a Hail Mary, you should be public about selling — too many companies are in that spot for coyness to work, and transparency casts a wider net (a plain ‘we’re running a process, are you interested?’ note). What buyers acquire is usually the team, so the data room’s most important content is who comes along. Use your investors’ networks; founders over-worry about returning money, but investors understand most early-stage bets return nothing (≈50%) and would rather you land somewhere great. On phrasing: don’t be too clever — everyone knows ‘evaluating strategic alternatives’ means selling. If you still have runway, don’t mention M&A at all; talk partnership and collaboration, because M&A is a dirty word while you can still walk. If you’re out of time, you’re out of time. On the buyer set: build a list of ~a dozen, qualify out the ones who are sidelined (just did a deal, headcount-frozen, spooked by uncertainty), and for the live two or three, run the inflict-pain playbook. She expects the M&A market to thaw as we get far enough past the November-2021 peak that valuations reset and better assets enter.

§ VC skills into product

What transferred from venture: investing in her network (she keeps a network of operators slightly ahead of dbt and mines them on open source, pricing, acquisitions, then ports the best ideas back); being a T-shaped generalist (broad credibility makes her more effective getting things done internally); and power-law thinking (most bets return zero, the wins pay for everything, so keep making bets that could bend the company’s trajectory). Her lightning-round operating change: do fewer things and single-thread the team on one mission.