Sell the Data, Bundle the Software Free

01 /The Concept

Bystreet's tagline is "Close more deals in hospitality." The startup has focused on a single market to start: hotels across the Caribbean and Latin America, where it has mapped more than 8,000 properties.

At the core of the platform is that property database.

Users can filter hotels across a wide range of criteria: branded versus independent, owner-managed versus managed by an operator, open, pre-opening, under construction, or in planning – among many others.

Beyond property information, the database includes contact records for the people responsible for different operational functions at each hotel: IT procurement, food and beverage purchasing, guest entertainment, and other departments where vendors might have something to sell.

But knowing that a hotel exists, and that a specific person oversees the relevant function, is only the starting point. Walking in cold and making a sale is hard. First, you need to understand how the property actually operates – well enough to make an offer that's meaningfully specific to its situation.

For that, Bystreet's AI generates a detailed research brief on each hotel or hotel group: its history, strategy, target guest profile, positioning, and other details that give sales teams the context to build a sharp, relevant pitch rather than a generic one.

The product still looks like an early-stage build – which is entirely reasonable, since the platform launched only yesterday, announced by Bystreet on the Y Combinator website as part of its current accelerator cohort.

02 /Why It Matters

Two months ago Bystreet entered YC with a completely different product: Gowatch, an AI platform that helped companies monitor what was happening with their customers and competitors by pulling signals from across the web.

On its face, Gowatch had a dramatically larger addressable market than a hospitality database covering the Caribbean and Latin America. So why the sharp pivot to a far narrower niche?

The sales case for broad products is weak. If your product is supposed to be useful to everyone, you end up marketing to no one in particular. The result is low conversion rates despite high spend on wide-reach advertising.

A niche product like a Caribbean hotel database changes the math entirely. The target audience narrows to companies already selling into that region – or looking to enter it. You can reach them directly, with a very specific offer. The cost to acquire each customer can be orders of magnitude lower than spray-and-pray advertising.

Narrow focus also gives you the time and resources to make the product genuinely good.

Building "an AI that scrapes the web for company intelligence" has no defensible edge anymore. The technology is commoditized – a target customer could largely do it themselves with off-the-shelf tools.

What Bystreet's AI does instead is dig much deeper. Rather than surface-level web scraping, it pulls and cross-verifies data from what the startup describes as "thousands" of public, government, industry-specific, and even offline sources – building a structural knowledge graph of who owns what, who operates what, and who makes which decisions across the Caribbean and Latin American hospitality industry. That makes the database not only more valuable but substantially harder to replicate.

Building these kinds of "deep" databases with rich, domain-specific knowledge and contacts that are genuinely hard to find elsewhere is an emerging trend gaining real traction – with investors and buyers moving away from sprawling generic databases toward narrower, more authoritative ones.

DataLane ([related review](/review/nikakoj-chatgpt-ne-pomozhet-tebe)) raised $22.5M late last year for a platform that helps B2B sellers reach small, offline local businesses – food service, beauty, and healthcare. Its tagline: "Find the buyers who spend their time in their establishments, not on LinkedIn" LinkedIn captures at most 10–50% of small offline businesses in the US; DataLane reaches the rest. Like Bystreet, DataLane builds industry graphs – mapping ownership structures, management chains, and decision-making authority.

Throxy ([related review](/review/pravilno-vybrannaja-model-raboty-startapa-srazu-otsekaet-kuchu-konkurentov)) takes a similar approach to data aggregation – pulling from specialized industry and government databases rather than generic sources – and has pushed it one step further: instead of selling database access, it sells booked meetings. Its AI finds qualified prospects, conducts outreach, qualifies interest, and only then schedules a meeting with the human sales rep. Starting from zero, Throxy reached $1.2M in expected annual revenue with a team of three.

Leadbay ([related review](/review/ishhi-pokupatelej-kotoryh-trudno-najti)), a Y Combinator graduate from late last year, promises to surface SMB prospects that "no one else can find" – building its database from specialized sources that generic tools miss. Its claim: 3x more small US and French businesses than LinkedIn, with contact quality so high that 99% of qualified leads it surfaces can't be found even through popular prospecting tools like Clay.

03 /Opportunities

Writing software with AI coding tools is now accessible to almost anyone. The practical consequence: selling software that processes well-known data with well-known algorithms is becoming increasingly difficult.

But the quality of any software's output is fundamentally determined by the quality of the data it processes. Which means a new business model is emerging: sell the data, and give the software away as a bonus.

The catch is that data is collected by software. So if the data you plan to sell can be gathered by a script that essentially scrapes LinkedIn, its value will be low.

The real play in data businesses is combining multiple specialized sources with proprietary matching algorithms. The goal is to create something that isn't impossible to replicate – just painful and expensive enough that buying is cheaper and faster than building.

Alternatively: sell the outcome, not the data. Throxy proved the model, going from zero to $1.2M annual run rate with three people – by selling booked meetings rather than database subscriptions.

The broader shift: the primary value of a startup may be moving from the software it builds to the data that software is trained on and processes. Which raises a pointed question for founders – are you planning to sell your customers data? If so, how hard is that data to collect, cross-reference, and keep current? And how much does the difficulty of replication protect the business?

The direction worth pursuing: software platforms whose core defensible asset is hard-to-aggregate data that delivers genuine value to customers. The entry point is identifying where that data is genuinely scarce in your domain – and whether you have the capability (or can build it) to collect, enrich, and package it for resale.