Oct 10, 2022 · 46m · mad
Separating Data Hype From Substance | Fireside Chat with Mode Co-Founder Benn Stancil
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of The MAD Podcast, FirstMark Partner Matt Turck interviews Benn Stancil, Co-Founder and Chief Analytics Officer of Mode, to critically evaluate current data industry trends and separates hype from reality. Benn offers candid perspectives on vendor vulnerabilities, architectural bottlenecks in dbt and DAGs, the evolution of BI, and long-term consolidation in the data ecosystem.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Matt holds 14.4% of the talking time here. How this is scored →
Matt as informed peer 4.6 Guest teaching 3.9 Guest disagreement 2.2 Matt pushing back 1.3
speaking balance: gold is Matt, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 22:26 Benn declares dbt's direction fundamentally wrong
Benn rejects the core premise of dbt's product expansion into semantic modeling, forcefully stating that building semantic layers on top of SQL queries is fundamentally wrong.
Hardest push from Matt ▶ 20:19 Matt presses on whether direct SaaS syncs exist in realityMatt directly challenges Benn's theoretical argument against Fivetran by asking if he has actually seen SaaS vendors providing direct database syncs in the real world.
Biggest teaching moment ▶ 26:00 Airport flight analogy for declarative orchestrationBenn re-educates the audience on data orchestration paradigms, using an airport scheduling analogy to show why procedural DAG chains are inferior to declarative end-state targets.
Matt holds his own ▶ 15:38 Matt details Snowflake's specific competitive threatsMatt showcases sharp technical knowledge of the ecosystem by raising specific competitive threats to Snowflake, including Firebolt for interactive queries and Databricks.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Matt as informed peer | Guest teaching | Guest disagreement | Matt pushing back | Why |
|---|---|---|---|---|---|---|
| Benn Stancil's Background, Role at Mode, and Substack Writing | 3 | 2 | 1 | 1 | Matt warmly introduces Benn and compliments his Substack before asking about his blogging history and whether Mode attaches lead generation metrics to it. Benn admits candidly that despite his marketing team's desires, there are no lead tracking metrics attached to his weekly posts. | |
| Defining the Modern Data Stack and the Ponzi Scheme Thesis | 5 | 3 | 3 | 2 | Matt prompts Benn on his provocative claim that the modern data stack is a Ponzi scheme and adds context regarding VC funding and data practitioners starting single-wedge companies. Benn explains how venture pressure forces narrow data tools into unrealistic hundred-million-dollar expectations. | |
| Analyzing Snowflake's Market Position and Potential Vulnerabilities | 6 | 4 | 2 | 2 | Matt framing a thought experiment on how market favorite Snowflake might fail, citing specific competitors like Firebolt and Databricks. Benn uses a smart TV versus simple TV analogy to highlight the risks of platform bloat versus low-cost bare-metal alternatives. | |
| The Future of ETL and Fivetran's Middleman Architecture | 6 | 4 | 2 | 3 | Matt questions whether SaaS companies bypass Fivetran by writing directly to warehouses in practice. Benn admits that while direct SaaS ingestion is cleaner theoretically, Fivetran thrives because building API integrations is unglamorous, difficult work. | |
| Previewing the Critique of dbt and the Limitations of SQL Semantics | 5 | 5 | 4 | 1 | Matt asks for a preview of Benn's upcoming post on dbt. Benn strongly asserts that dbt's expansion into semantic layers via SQL is fundamentally wrong, comparing independent SQL transformations to disconnected animation frames. | |
| Flaws in DAG-Based Orchestration and the Airport Analogy | 5 | 5 | 3 | 1 | Matt asks Benn to revisit his critique of DAG-based orchestration. Benn uses an airport flight scheduling analogy to illustrate why declarative, target-time scheduling makes more sense than procedural domino-pushing. | |
| Deconstructing Data Products and the Data Mesh Framework | 5 | 5 | 4 | 1 | Matt asks about Data Mesh and Data Products, invoking creator Zhamak Dehghani. Benn humorously claims nobody understands either term, calls Yelp the quintessential data product, and describes DataMesh as administrative overhead born from enterprise bloat. | |
| Long-Term Outlook, Market Expansion, and the Reinvention Cycle | 5 | 4 | 2 | 2 | Matt questions whether the five-year cycle of reinventing data infrastructure will ever end. Benn explains how the industry operates like a perpetual motion machine, swinging back and forth between bundling and unbundling tools. | |
| Mode's Product Roadmap and Modern BI Architecture | 4 | 3 | 0 | 0 | Matt transitions to discussing Mode's business and product vision. Benn outlines how Mode is adapting to modern decoupled architectures by consuming external semantic layers like dbt rather than building proprietary governance layers. | |
| Audience Q&A: Industry Fragmentation, Consolidation, and Specialized Databases | 2 | 4 | 1 | 0 | Matt moderates audience questions covering market fragmentation, vertical consolidation, and domain-specific proprietary databases. Benn predicts that standards will inevitably coalesce around central platforms like Snowflake and dbt. |
Statements from this episode (12)
Stancil: Mode is a BI tool built for people who dislike traditional BI
“It is a BI tool basically, but a BI tool built for people who don't like BI.”
Stancil: Modern data stack startups face an inevitable market reckoning
“Pretty quickly you're gonna realize they're all stepping on each other's toes, and that, that fallout has to go somewhere. Like, not all of these companies can be the next Figma that they all now say that they are, and so it's what happens then, and I think it…”
Turck: Fivetran generates over $100M in annual revenue
“And is actually as far as I know, one of those, like, companies are over a hundred million in, in, in revenue.”
Stancil: Native SaaS warehouse connections could undercut Fivetran's core revenue
“You know, it doesn't undercut everything for Fivetran, but it potentially undercuts the big sources, which I imagine are the things that are, like, the real drivers of revenue for them.”
Stancil: dbt's SQL-based semantic layer is not a real semantic model
“They're moving in, like, the semantic layer tool, so basically they're saying, give us raw data, and we will tell you, like, apply semantics to it. The way that they do that now is through SQL. So, like, semantics are kind of, like, air quote semantics. It's b…”
Stancil: dbt's initial ease of use may become its downfall
“Dbt is, I think, really popular because it's so easy to get up and running, but it may also eventually be, like, if it had an undoing, to me that would be the undoing, is the thing that was really easy to get up and running doesn't actually solve the real prob…”
Benn Stancil: Data orchestration should be deadline-driven rather than task-triggered
“The way we've built DAGs are kind of like, when do I start the, like, when do I kick over the first one, and then we line it up such that we hope the thing arrives at the end. And the way it would make more sense to me is you just tell the thing, I need this t…”
Benn Stancil: Analytics tools will eventually embed directly inside workflow software
“You could potentially incorporate data into that, such that the data sort of surfaced to people in the moment they need it in the product you're trying to use to solve the problem, instead of going to a dashboard and, like, clicking on some stuff. So I think t…”
Stancil: Data mesh is an uninteresting side effect of failed centralization
“I guess, ah, seems hard to manage and practice. The way I've seen people describe it is basically it's the thing that you naturally create when you're a very big organization and you can't have a centralized data team that can possibly centralize everything, w…”
Stancil: Enterprise data market won't support 1,000 unicorns
“The median outcome is the data philosophies of Facebook and LinkedIn and all these kind of early tech companies gets adopted by the enterprise, and so all these sort of modern data tools that tech companies buy today go off and get sold to Coca-Cola and Caterp…”
Stancil: Semantic modeling layers should exist outside of BI tools
“Rather than that living in your BI tool, we think that should live in, in a more generalized layer, and then we just consume from it.”
Stancil: Integration standards will be set by Snowflake, not open standards
“I don't think the way we get there is sort of like through open standards and stuff like that. I don't think anybody will actually adhere to that. I think most likely what happens is Snowflake basically says, hey, if you do things in this particular way, we ca…”