The Data Gap Is Dead

Attention used to be the whole game. For a decade, we heard “brands are buying your attention,” “everyone is tracking your attention,” “optimize for attention.” I think we’re past that. We more or less know where attention lives, and the volume of behavioral information is monumental. The next evolution isn’t just better ad targeting or your favorite creator doing a sponsored post. It’s twofold: first, shaping how AI/LLMs answer you, because those outputs are becoming the new top-of-funnel discovery. Second, generating unique data and proprietary insight so compelling that a bigger company acquires you simply to be half a step ahead.

Big companies still move at lightspeed when they want to; budget, reach, and distribution are huge advantages. But the old data moat (infra, headcount, tooling) isn’t what it used to be. Off-the-shelf AI and commodity data infrastructure mean small teams can collect more, synthesize faster, and produce stronger, clearer signals with a fraction of the cost. This doesn’t magically erase execution risk, but it does level the playing field on capability. Anyone can catch up on the mechanics now. The difference is who operationalizes it and who doesn’t.

Don’t “Do Data Later”

The data firehose keeps getting bigger: the world is on track for ~181 zettabytes of data in 2025, up from ~149 ZB in 2024. That’s why small teams can learn faster than ever. And AI isn’t just noise: mainstream estimates put gen-AI’s annual economic upside in the trillions (low-to-mid single digits). Translation: the capability gap is now execution, not access. Meanwhile, modern data infra is radically accessible (usage-priced warehousing, per-second compute), another reason the “we’ll do data later” story doesn’t hold.

When I’m looking at a startup, I care about the usual suspects, traction, retention, payback, gross margin, because they show whether the engine actually runs. But personally, the metric at the forefront of my mind is decision velocity. AI makes it possible to ingest messy inputs (product events, sales notes, support threads, market chatter), squeeze out patterns, and turn those patterns into concrete bets faster than a human-only pipeline ever could. When the path is unclear, and it’s always unclear, those machine-accelerated insights are the difference between wandering and compounding.

The Moat, The Upsell, and The Necessity.

I’m seeing three storylines over and over:

• Data as a moat. This is more or less always someone’s go-to answer about how they are thinking about data, and yes, ideally, data is a moat, but we are seeing data moats fall apart quickly with this new speed. What’s even stronger is innovative proprietary collection and structure; your vantage point lets you answer questions competitors can’t. Not just “we have data,” but we have the right data with the right join keys and freshness to change outcomes. Go find a way to get customer insights through every YouTube and TikTok comment on your competitor’s new feature. That defensibility shows up in win rates, renewal odds, and pricing power.
• Data as an upsell. Insight layers are sneaky-good revenue. The moment you can benchmark a customer, forecast their next best action, or remove effort from their workflow, you create attach points that become really sticky. And, I should add, it’s not just core customers that should come in contact with these upsells. Eventually, data allows you to open up new revenue streams or even business models that attack new customer demographics: Sell the decision to your core user, and sell the aggregate to an adjacent buyer. Brands pay for profit guidance while agencies/3PLs license anonymized category benchmarks; factories buy ‘next-failure’ alerts while OEMs and insurers purchase aggregated failure telemetry—same data, second revenue line.
• Data as table stakes. At Late Seed to Series A, a real “data play” isn’t a nice-to-have; I think it’s expected. If a team at that stage can’t show how data shapes product, sales motion, or customer economics, I’m concerned. If it’s already a portfolio company and there’s no data spine, I’m worried.

VCs need a data spine, too

Founders aren’t the only ones who should be running on insight loops; VCs should, full stop. At our firm, we’re actively organizing deal, company, and market data to push past intuition and pattern-matching skills. The goal isn’t a magic 8-ball; it’s sharper priors and faster updates. In the very near term, I want to answer questions at “operator resolution,” not cocktail-party resolution: At what revenue, market, round size, valuation, and team size does a legal-tech process-serving marketplace cross into profitability, and what’s the modeled probability of a 10× on a $500k initial check? That level of detail is now possible. The limiter isn’t AI; it’s whether we collect cleanly, think rigorously, and connect dots others don’t.

This is where it gets interesting for portfolio support, too. If you know that companies with X retention and Y sales cycle in Z sub-sector benefited from specific pivots, mentors, and pricing moves, you can route that playbook to the right founder at the right time. Insight becomes an operating advantage the moment it’s actionable and transmitted.

Does this “kill VC”? In one sense, yes, it kills the past framework, but the better frame is that VC gets reborn: more backbone to intuition, more evidence, less mystique, and a potential benefit to the VCs with more deal count. That’s good for founders, and it’s how VCs will earn a seat at the table five years from now.

AI isn’t the product; it’s the amplifier.

The exciting part is how ruthlessly AI compresses the loop. Collection scales (you can instrument more, listen more, enrich more). Synthesis improves (you can classify, summarize, and prioritize signals without an army). And, crucially, the output can be compiled into an opinionated next step: who to call, what to build, where to raise prices, and which segment to kill. That alone is the roadmap. Take this lightly, but forget what you think v2 and v3 should be; listen to the data. It won’t write your strategy, but it will surface the highest-leverage questions faster than you could on your own.

There are plenty of fintechs that “do cap tables” or sell valuation tools. Carta stands out because of the insight surface area it sees and synthesizes. We’re talking 50,000+ startups, 2.5M+ security holders, and trillions in equity managed, enough longitudinal signal to publish private-market trends, hiring and comp analyses, ownership dynamics, and round mechanics with real statistical weight. That dataset, not just the workflow, lets them ship benchmarks operators actually trust. In other words, elite because of data gravity, not just product.

Don’t Rely on “Product Roadmap”

I’ll ask “Show me your product roadmap,” and “Walk me through your customer roadmap.” Here’s my take on that, especially at the pre-seed/seed stage: if you can 1) pivot faster because you have stronger, fresher insight or 2) make more money because you own proprietary insight, you actively mitigate the risk we’re all underwriting, unforeseen change. A shift in customer mix, a new competitor with big spend, a tool that obsoletes half your feature set, or an unexpected revenue seam opening up…that’s the job. Founders find it and execute. Not the VC. When I meet a Late Seed or Series A company and there’s no credible data spine, no way to detect, decide, and act at speed, I’ve already learned enough about the founder. It’s harsh, but it’s true.

On buying vs. owning data.

I see both ends. Some pre-seed teams are impressively prepared, with a collection built into the business model from day one. Others outsource the early steps, buying lists, renting insights, and stitching together fragmented third-party data to get by. I get it. Do what you have to do to get motion. But let’s be honest: that’s a short-term bridge, not a great long-term play. If your edge depends on a vendor, every competitor can also pay; you don’t have an edge, you have an expense line. Over time, you need to tip toward ownership: a vantage point others can’t easily replicate.

AI’s pace is closer to a platform shift than a feature wave. If you are not rapidly internalizing what’s possible, you’ll get lapped. New tools will replace old workflows; new distribution channels (including AI agents themselves) will reroute demand; new pricing curves will punish slow movers. The teams that survive will be the ones already wired to detect and adapt because their data loops actually run. Today.

Where I Stand

Data isn’t just “important.” It’s the operating system for how modern startups learn, decide, and make money. At the validated stages, I would expect to see it clearly. Show how you listen to the market, how you translate that into product and revenue moves, and how AI helps you do it faster than humanly possible. Show investors where your insights are truly proprietary. Show me that your advantage compounds the longer you operate. If you can do that, I wouldn’t really need a 10-page roadmap. We can already see the company you’re building. If you can’t, the rest of the story doesn’t matter, because a new tool or a new competitor will write it for you.

- Rron Thaci