Conversation
Highlights

How Qualytics Built Enterprise Data Quality Infrastructure by Validating First, Building Second

Gorkem Sevinc spent 18 months building a product nobody wanted.

The radiology second opinion platform seemed defensible. He had co-founded it with Johns Hopkins radiologists, invested nights and weekends into development, and brought genuine domain expertise. There was just one problem: patients didn’t even know radiologists were doctors, let alone understand why they’d need a second opinion on their reports.

“Turns out nobody was interested in second opinions in radiology,” Gorkem recalls. “Most patients didn’t even know who a radiologist is, that it’s an actual doctor that is reading your images and generating a report that goes to your clinician.”

The company eventually pivoted to litigation cases and achieved a modest exit, but the lesson crystallized. In a recent episode of BUILDERS, Gorkem, now Founder and CEO of Qualytics, shared how that failure shaped an entirely different approach to building his current company—a data quality platform experiencing explosive inbound demand as enterprises realize their data foundations can’t support AI democratization.

The Whack-a-Mole Problem

Before founding Qualytics, Gorkem operated as a CTO and Chief Data Officer at a financial planning and wealth management company. The recurring problem was maddeningly specific: his team managed 100+ data sources feeding into Snowflake, supporting analytics and ML models across the organization. Everything looked architecturally sound—until dashboards broke and KPIs reported incorrect numbers.

“I’m putting all these controls around my data. I have my Snowflake, I have 100 different systems feeding data into this centralized data repository. I have all sorts of crazy analytics, ML models, AI models built on my data foundation. That’s all great except the data is wrong and KPIs are off,” Gorkem explains.

The reactive pattern became numbingly predictable: executives escalate, data team writes a custom rule to catch that specific issue going forward, everyone moves on until the next break. “We end up taking a reactive approach and we’re writing a piece of code for that scenario. Kind of like software testing, but this data testing, data quality, right? We write a rule to catch a specific issue going forward.”

His team was managing hundreds of these custom rules, yet they were still getting yelled at. “My team was writing and managing a lot of these rules and we’re still being yelled at because we’re being taking a highly reactive approach. It’s expensive. Whack a mole.”

After evaluating incumbents and other startups focused on data observability—the “did I get the right amount of data at the right time” verbiage—Gorkem concluded they were solving the wrong problem. “What I need is my business people to say this data is fit for purpose. It is actually able to be used.”

The 100-Conversation Rule

As the pandemic began in 2020, Gorkem and his co-founder started building Qualytics with a hard constraint: validate with 100 prospects before writing production code.

His network of CTOs and CDOs, plus PE portfolio connections from his previous company, provided access to potential customers who could stress-test his hypothesis. The goal wasn’t confirming the problem existed—he’d lived it. The goal was understanding whether prospects recognized they had the problem, how they currently addressed it, and what they’d actually pay to solve it.

“I wouldn’t start an idea and go after it if I hadn’t talked to 100 people first,” Gorkem states. This wasn’t informational interviews about pain points—it was NSF I-Corps-style validation: show rough sketches, understand what happened when they hit the pain point, probe how it hurt them financially or operationally.

The first six months focused on competitive research and customer conversations while his co-founder built the initial prototype. Only after validation did they line up beta customers and raise their pre-seed round.

Repeatability Lives in Weekly POC Patterns

Early customers came from Gorkem’s network, but the inflection point arrived in late 2022 when Qualytics closed their first paying customer with no prior relationship—a large enterprise that required navigating procurement without SOC 2 certification or multi-year track record.

Gorkem’s approach: convert the free POC into a paid one. Not for revenue, but for commitment. “I convinced one to make it into a paid proof of concept… there is shared risk. Now I’m not taking all the risk for you do a free POC and then you walk away.”

Getting the MSA signed upfront forced both parties through legal and security review before technical validation started. That customer 10x’d their initial contract and remains with Qualytics today.

But one successful customer doesn’t prove repeatability. For Gorkem, product-market fit crystallized when POC behavior became predictable by week. “When I started seeing the trends between proof of concepts that we’re running… when I see everybody getting replicability of what I would expect within that poc, I can say, okay, week two, I’m expecting this. Week three, I’m expecting this.”

That pattern recognition—emerging in late 2023 and early 2024—enabled them to build ROI calculators and internal champion enablement materials. “Now I can start thinking, let me put an ROI calculator together for you. Let me actually help you sell this internally.”

The patterns were consistent enough to systematize and hand off to professional sellers. Qualytics raised their Series A.

Simultaneous Sales and Marketing Leadership Hiring

Gorkem made a structural decision that runs counter to typical founder instinct: he hired sales and marketing leadership at the same time, not sequentially.

“I felt very strongly that sales and marketing leadership need to start at the same time and that I am keeping them accountable to go to market motions and go to market metrics, not just sales metrics and marketing metrics because the two have to be working in lockstep.”

Before Series A, the company operated with essentially no marketing budget. “We had spent no money on marketing and we had an absolute shit website.” The company unveiled their rebrand two days before this conversation—a corporate identity finally matching product maturity.

For sales hiring, Gorkem leveraged his and co-founder Eric’s practitioner credibility to hire for complementary skills rather than domain expertise. “I can teach you data quality, I can teach you data operations. You don’t have to be an expert in data quality to be able to do the sales motions. I can teach you that. I cannot teach you how to grind, how to have that grit within you.”

The division of labor: technical practitioners (Gorkem, Eric, customer success, sales engineering) maintain empathy and credibility with buyers who’ve lived similar problems. Enterprise AEs and BDRs bring the sales motion discipline.

In-Product Engagement as Leading Churn Indicator

Beyond net revenue retention, Gorkem instruments granular product usage: frequency of data quality operations, number of anomalies discovered, how actively customers edit rules. These are leading indicators—engagement drops precede churn by months.

“I pay a lot of attention to within a poc, within our customers… how often are they doing DQ operations, how many anomalies are they finding, how many rules are they editing, what’s the kind of amount of activity that I have within my product.”

When engagement signals decline, he doesn’t schedule a QBR for next quarter. “I’m not going to do a QBR with you. I’m going to pop into a weekly meeting that you have with my team and as the CEO, I’m going to have a little conversation with you what’s working, what’s not working and adjust accordingly.”

The intervention latency matters. “If I only look at NPS once a year, I have already failed.”

Vertical-Specific Business Outcome Translation

Qualytics’ most important positioning decision was refusing to sell “data quality” as a horizontal data engineering tool.

“Data quality is not a data problem. It’s a business problem. It has to be a collaboration between business users and data users. I’m not selling to a data engineer. I’m selling to an organization with the goal that they’re going to engage business users.”

In practice, this means translating technical capabilities into outcomes each vertical cares about. Insurance companies hear about premium leakage. Banks hear about OCC and SEC data controls and regulatory compliance requirements. Supply chain companies hear about their specific operational problems.

“Let’s say that you’re an insurance company and you have premium leakage problems or you are in supply chain and you have different problems. Those may not be called data quality, but they’re data quality behind the scenes. So when I understand that vertical and I can talk to a bank about their data controls and their regulatory requirements from the OCC and SEC, they take us a lot more seriously than here’s another sales guy pitching me on a data quality product.”

This requires research into each vertical’s regulatory environment, operational pain points, and executive-level business metrics. When you walk in speaking their language about P&L impact, you’re not another vendor—you’re someone who understands their world.

The AI Forcing Function

When Qualytics launched, some enterprises classified data quality as “nice to have.” AI adoption eliminated that optionality overnight.

“Thanks to all the AI innovations that are happening. Every single enterprise has AI initiatives and AI budgets,” Gorkem explains. The implications are stark: “Imagine this. I have, through AI enablement, every single employee in my organization, in my 20,000 person organization can go ask questions, interrogate data and come up with insights that are previously unknown. Somebody didn’t build that dashboard. You are actually interrogating it and coming up with it. Great. Do you trust that data?”

For most enterprises planning to democratize data access through AI interfaces, the answer is no—and that realization is driving demand. “In order to trust that data, you have to have a robust data governance program. You have to have a data quality product like quality. You have to have your data catalog, et cetera. You have to have your data ecosystem in order.”

The result: “The amount of sales activity, the amount of inbounds that we’re getting is exploding now. And I think that’s going to be 10x in just 2 years very easily because of the amount of AI adoption that is happening.”

Gorkem’s timing wasn’t accidental. He built during the “nice-to-have” phase so the product would be ready when external forces—AI adoption, in this case—made data quality mandatory infrastructure.

His ultimate goal: “I want my company to be the name that is synonymous with data quality.”

For a founder who once spent 18 months building something nobody wanted, that ambition rests on a foundation of ruthless validation, repeatable POC patterns, and the understanding that the best enterprise products solve business problems, not technical ones.