ScoreData Corporation

A phone rings. Somewhere, a model decides who should answer it.

It is the most ordinary moment in business: a customer calls. A claim, a complaint, a question about a balance. For most companies, what happens next is luck - whichever agent is free picks up. ScoreData Corporation thinks luck is a terrible business strategy. In the half-second before the line connects, its software has already read the caller's history, scored the likely outcome, and routed the call to the agent most likely to resolve it. No human pressed a button. A model did.

ScoreData is an eleven-person company in Palo Alto, California, which is to say it is small enough to fit in a conference room and ambitious enough to take on Databricks in a footnote. It builds an AI/ML platform called ScoreFast, and its pitch is refreshingly unglamorous: take the messy, legacy, half-forgotten data a business already owns, and turn it into decisions that happen in real time. Fraud or not fraud. Stay or churn. Buy or pass. Score it, and move on.

Pictured: not a robot, not a rocket - a routing decision. The least photogenic, most valuable thing ScoreData makes.

Everyone has the data. Almost no one can use it in time.

Here is the uncomfortable truth that ScoreData was built around. By the mid-2010s, every bank, insurer and call center was sitting on mountains of data. They also had data scientists, dashboards, and slide decks full of insight. What they did not have was a way to act on any of it before the customer hung up.

The bottleneck was never the math. It was the maintenance. A predictive model is a perishable thing - it goes stale the moment the world it was trained on moves. Keeping models fresh meant armies of specialists retraining, revalidating, redeploying. By the time the new model shipped, the fraud pattern had changed and the customer had already left. Insight arrived, reliably, one quarter late.

The industries feeling this most acutely were the ones where a single decision carries real money and real risk: financial services, insurance, banking, telecommunications, healthcare. A wrongly routed call, an undetected fraudulent transaction, a churning customer no one flagged - each one has a price. ScoreData decided the answer was not more data scientists. It was a model that mostly takes care of itself.

Five founders, one slightly contrarian idea.

ScoreData was founded in 2014 by Vas Bhandarkar, Danny Yang, Mudit Chandra, Prasanta Behera and Pankaj Jha. Bhandarkar, the CEO, is the sort of Silicon Valley operator whose resume reads like a history of AI's commercial false-starts and breakthroughs - he was an early employee at two companies, Remedy Corporation and Selectica, both of which made it all the way to a Nasdaq IPO. He had watched, up close, how hard it is to turn a clever algorithm into a business that pays its bills.

So the founding bet had an unfashionable twist. While much of the industry chased ever-larger models, ScoreData paired machine learning with old-fashioned econometrics - the discipline of measuring how economic variables actually move. The wager: a model that understands the economics of a decision will make better, more profitable ones than a model that only pattern-matches. Less magic, more accountability.

It was, admittedly, the less exciting pitch at the AI party. It is also the one that tends to survive the morning after.

ScoreFast: the model that does its own homework.

The platform is called ScoreFast, and the name is the promise. It ingests external and legacy data - the stuff buried in systems no one wants to touch - and builds predictive models that update continuously, with minimal manual intervention. The point is not that humans never look at it. The point is that the model does not wait for them.

What it produces are run-time scores: numbers a business can act on in the moment a customer is in front of it. The use cases are deliberately practical.

A platform whose biggest selling point is what you don't have to do. Hard to put on a billboard. Easy to put on a P&L.

The short, deliberate history of a long bet.

Money talks. So, increasingly, do the models.

ScoreData has raised roughly $3.9 million across seed and Series A rounds - modest by the standards of an industry where pre-revenue startups raise nine figures on a slide. In June 2020, it closed Series A financing from Impact Venture Capital, a Silicon Valley firm with a taste for early-stage AI and machine learning. Existing investors RecruitGroup and Asha Jadeja Motwani came along, which is usually a quieter signal worth more than the headline.

The reach is the surprising part. For a company you could seat at one long table, ScoreData operates across three countries - the United States, Japan and India - largely through its ties to the Recruit ecosystem in Tokyo. Its competitive set, per the analysts, includes names like Databricks, Altair, Fivetran and Qlik. ScoreData is not trying to be those companies. It is trying to be the one that already had your data working before you finished your coffee.

Eleven people. Three time zones. One platform that, ideally, you forget is running. That's the whole trick.

Make every decision a scored one.

Strip away the platform diagrams and the mission is simple. ScoreData wants the moment of decision - which agent, which offer, fraud or not - to be informed, instant, and quietly automatic. It wants the data a company already owns to earn its keep. And it wants the people who run the business, not just the people who model it, to be able to ask the question and get a usable answer.

There is a healthy skepticism every reader should bring to a sentence like that. AI promises are cheap. What separates ScoreData's version is the unglamorous bit: the insistence that a model is only useful if it stays current and a decision is only valuable if it's on time. That is not a moonshot. It is plumbing. And plumbing, when it works, is the most underrated technology there is.

The phone rings again.

Return to that ordinary moment. A customer calls. For decades, what happened next was a coin flip dressed up as a process. The right agent, the right offer, the caught fraud - all of it depended on someone, somewhere, having done the analysis in time. Usually they hadn't.

ScoreData's wager is that the coin flip is over. In its version of the moment, the model has already read the room. The call routes itself. The fraudulent charge never clears. The customer about to leave gets a reason to stay - before they even know they were leaving. The decision is made in the half-second no human could fill, and it is made the same way every time: scored, current, accountable.

ScoreData is not the loudest company in artificial intelligence. It may end up being one of the more useful. Eleven people in Palo Alto are betting that the future of customer engagement is not a smarter robot but a faster decision - one that's already been made by the time the phone stops ringing. The rest of the industry is welcome to chase the spotlight. ScoreData would rather catch the call.

Where we came in: a ringing phone. Where we leave: the same phone, answered better, by a decision nobody had to wait for.