dbt Labs AI Customer Research: How the Analytics-Engineering Pioneer Builds With Its Community

TL;DR#

dbt Labs, the creator of dbt (data build tool) and the company that coined "analytics engineering," builds its product by listening to one of the largest open-source communities in data — more than 100,000 data professionals, with over 30,000 in its Community Slack. That community is also a commercial engine: by one analysis, it has generated roughly 80% of dbt Labs' inbound leads. dbt Labs formalizes this listening through its annual State of Analytics Engineering report, which surveyed 459 practitioners for its 2025 edition and found that 45% named AI tooling as their top investment priority and 56% still cite poor data quality as their biggest challenge. The lesson for any software company: community-driven discovery scales only when you can capture the "why" behind feedback, not just count GitHub issues. This is exactly where AI customer interviews change the economics — letting product teams run hundreds of conversational, follow-up-driven discovery sessions instead of relying on surveys, Slack threads, and a handful of advisory calls. For analytics-engineering and horizontal SaaS teams, the dbt model shows what community-led product development looks like, and conversational AI research shows how to do it at scale.

How dbt Labs Builds With Its Community#

dbt Labs builds with its community by treating open-source contribution, public feedback, and a recurring research report as the primary inputs to its roadmap. The company started as Fishtown Analytics, rebranded to dbt Labs in 2021 alongside a $150 million raise, and grew dbt from a small open-source transformation tool into the de facto standard for analytics engineering. dbt Core remains open source under the Apache 2.0 license, and according to dbt Labs' own community figures the project has drawn hundreds of code contributors in a given year across tens of thousands of companies.

The mechanics matter. Community members open GitHub issues, debate design in the dbt Community Slack, and present real-world patterns at Coalesce, dbt Labs' annual conference — which includes dedicated feedback sessions that help shape features. The 2025 Coalesce demo of the dbt Fusion engine, with native SQL comprehension, is a recent example of a major product direction surfaced and validated in front of the people who use the tool daily.

This is the open-source flywheel that horizontal SaaS founders study. But it has a structural limit: the loudest voices in a Slack channel or a GitHub thread are not a representative sample of your users. Counting issues tells you what bothers the contributors who file issues. It does not tell you what the silent 95% of practitioners actually need. That gap is the central problem of community-driven discovery, and it is why scaling qualitative signal — through approaches like customer research at scale — has become a defining challenge for data-tooling companies.

The State of Analytics Engineering Report as Structured Listening#

The State of Analytics Engineering report is dbt Labs' attempt to turn ad-hoc community noise into representative, citable data. The 2024 edition drew responses from 456 data practitioners and leaders; the 2025 edition surveyed 459 practitioners and leaders between October 8 and December 27, 2024, with 70% individual contributors and 30% managers. Analytics engineers made up 48% of individual-contributor respondents.

The findings double as a roadmap signal. In the 2025 report, 45% of respondents named AI tooling as their top area of investment for the year ahead, while 38% pointed to data quality and observability — and 56% still cited poor data quality as their most frequent challenge, according to dbt Labs' summary. On budgets, 30% of respondents reported data-team budget increases versus just 9% the prior year, and 40% reported headcount growth versus 14%, as covered by BigDATAwire. The 2024 report had already flagged a softer organizational problem: only 14% of data professionals strongly agreed that their organization sets clear goals for their team.

A once-a-year survey, however, is a snapshot, not a conversation. It tells you that 56% of practitioners struggle with data quality. It does not tell you what "data quality" means to a fintech analytics engineer versus a healthcare data team, or what they tried before giving up, or what would make them switch tools. Surveys flatten rich context into checkboxes — the same limitation that makes AI versus surveys a live debate for every research team. The "why now" lives in the follow-up question a static form can never ask.

Why Community Signal Alone Is Not Enough#

Community signal alone is not enough because it is biased toward the engaged minority and stripped of the reasoning behind each request. A GitHub issue captures a problem statement; it rarely captures the workflow, the workaround, the budget pressure, or the competing tool the user is quietly evaluating. Three structural gaps recur for community-driven companies like dbt Labs:

1. Representation bias. The practitioners who file issues, post in Slack, and submit talks skew senior, opinionated, and highly engaged. The much larger population of quiet daily users — the ones whose churn or expansion actually moves revenue — is underrepresented.
2. Missing the "why." A feature request is a proposed solution, not the underlying job. Without probing, teams build the literal ask and miss the deeper need. This is the core of modern conversational data collection — capturing intent, not just stated preferences.
3. No continuous cadence. An annual report and a perpetually scrolling Slack are two extremes: too slow and too noisy. There is no structured, repeatable way to ask the right 200 users a focused question this week and synthesize it by Friday.

Other data and SaaS companies hit the same wall. The way Databricks approaches AI customer research across its lakehouse base, how Amplitude turns behavioral data into customer voice, and how Datadog runs customer research at observability scale all wrestle with translating a huge, technical user base into roadmap clarity. Community is a moat, but raw community signal is not a research method.

How AI Customer Interviews Apply to Community Discovery#

AI customer interviews apply to community discovery by replacing the survey-and-Slack model with conversations that scale — hundreds of simultaneous, follow-up-driven discovery sessions that capture the reasoning a checkbox never could. Instead of asking 459 practitioners to rank investment priorities once a year, a community-driven company can run continuous AI customer interviews with thousands of users, probing each answer in real time.

For a company structured like dbt Labs, the application is concrete:

• Segment the silent majority, not just the vocal contributors. Invite a representative cross-section of practitioners — not only Slack regulars — and let an AI interviewer agent ask the same opening question, then branch based on each person's answer.
• Probe feature requests for the underlying job. When a user asks for a specific capability, the AI follows up: what are you trying to accomplish, what do you do today, what breaks? That is the difference between building the ask and building the need.
• Replace the intake form on research and beta sign-ups. A conversational intake flow or concierge agent gathers context in the user's own words instead of forcing them through dropdowns — directly addressing why static intake forms hurt conversion.
• Run the State of Analytics Engineering report as an always-on study. Rather than a single annual survey, a continuous study turns the report into a living instrument, with continuous discovery feeding the roadmap quarter by quarter.

This is the model Perspective AI is built for: AI-moderated interviews that follow up, probe, and capture the "why," then synthesize hundreds of transcripts into themes automatically. It is what makes AI qualitative research the default rather than the luxury, and it is increasingly part of the customer research stack modern product teams actually use.

The Analytics-Engineering Context: Why This Matters in 2026#

In analytics engineering specifically, the discovery problem is sharpened by how technical and opinionated the audience is. Data practitioners distrust vague research; they want precise questions and they notice when a survey misunderstands their workflow. dbt Labs earned its standing partly by speaking their language — the term "analytics engineering" itself was a community-validated framing, not a marketing invention.

The market is also consolidating fast. The 2025 report's signal that 45% of teams are prioritizing AI tooling, paired with industry moves like the Fivetran–dbt merger discussed across the data press, means product expectations are shifting under everyone's feet. Companies that can hear those shifts early — and distinguish a loud Slack thread from a real, representative trend — will win. That is why product roadmap validation has moved from a quarterly ritual to a continuous capability, and why teams are rethinking the entire survey layer of customer research.

For the product teams running this, the playbook is straightforward: keep the open-source flywheel, keep the annual report, but add a conversational research layer underneath both so the "why" is never lost. You can see how other technical-audience companies approach it in the way Cursor gathers conversational developer feedback and how Anthropic researches enterprise AI buyers at scale.

Frequently Asked Questions#

How does dbt Labs use its community to build its product?#

dbt Labs uses its community as a primary roadmap input through open-source contribution, public feedback, and an annual research report. Community members open GitHub issues on dbt Core, debate design in the 30,000-plus-member dbt Community Slack, and present patterns at the Coalesce conference, which includes dedicated feature-feedback sessions. The company then formalizes this signal in its State of Analytics Engineering report.

What is the State of Analytics Engineering report?#

The State of Analytics Engineering report is dbt Labs' annual survey of data practitioners and leaders about their pains, priorities, and investments. The 2025 edition surveyed 459 practitioners between October and December 2024 and found that 45% ranked AI tooling as their top investment priority while 56% still cited poor data quality as their most frequent challenge. It functions as structured listening that complements raw community feedback.

Why are AI customer interviews better than surveys for community discovery?#

AI customer interviews are better than surveys for community discovery because they capture the reasoning behind feedback, not just ranked checkboxes. A survey tells you 56% of users struggle with data quality; a conversational AI interview asks what that means to each user, what they tried, and what would make them switch. AI interviewers can run hundreds of these sessions simultaneously and synthesize the transcripts automatically.

Can conversational AI research scale to a large technical user base?#

Yes, conversational AI research scales to large technical user bases by running many simultaneous AI-moderated interviews instead of one-to-one researcher calls. For a community like dbt's 100,000-plus data professionals, an AI interviewer can invite a representative cross-section, ask consistent opening questions, branch on each answer, and probe feature requests for the underlying job — capturing depth at a scale manual research cannot reach.

How can a community-driven SaaS company avoid bias from vocal users?#

A community-driven SaaS company avoids vocal-user bias by deliberately sampling beyond the engaged minority who file issues and post in Slack. Combining open-source signal with structured, representative research — such as continuous AI interviews sent to a cross-section of users rather than only active contributors — surfaces the needs of the silent majority whose retention and expansion actually drive revenue.

Conclusion#

dbt Labs is the clearest case study in community-driven product development: an open-source flywheel, a 100,000-plus-member community, and an annual State of Analytics Engineering report that turns practitioner sentiment into citable data. But the dbt model also exposes the limits of community signal — representation bias, missing context, and no continuous cadence — that every horizontal SaaS company eventually hits. AI customer interviews close that gap by capturing the "why" behind feedback at the scale a community demands, replacing the survey-and-Slack model with conversations that follow up and probe. If you build with a community, keep the flywheel and the report, but add a conversational research layer underneath. Start a study in minutes, explore Perspective AI's AI interviewer, or see how teams are using it to make community-driven discovery representative, continuous, and deep.