There is a persistent, often unspoken challenge for economists who want to track the earliest phase of any new industry. Financial technology—what everyone now simply calls fintech—offers a prime example. In 2010, the sector was barely defined, let alone properly categorized. If you were tasked, as I once was, to quantify new fintech startups from that period, you would have almost certainly started with the International Standard Industrial Classification, ISIC. Specifically, code 6201: Computer programming activities.
Now, this is a category so broad that it swallows everything from simple app builders to, say, core banking infrastructure developers. If I’m being honest, the “fintech” label simply wasn’t front-of-mind for many entrepreneurs back then. Most company registrations didn’t mention “finance” at all, and yet, it’s precisely these firms—quietly registered as “computer programming” businesses—that went on to reshape banking as we know it. There’s a small irony in that.
But, say you do want to isolate fintech entrants. The first step, if you follow a methodical approach, is to pull all company registrations under ISIC 6201 for the year 2010. In some countries, that’s straightforward. In others, it involves requests, negotiations, or outright luck. The reality of cross-country company registry data—at least as of the mid-2010s—is frustratingly uneven. If you’re reading this as a policymaker, you might be in a position to improve that, but, for now, let’s take what we can get.
So you have a list. Now what? Most entries will be generic: “ABC Solutions,” “XYZ Software.” Very few announce, “We are a financial technology company.” Even fewer specify the business model in any standardized way. Here’s where early-stage funding data comes in handy. There is usually a lag—six months, a year, even two years—between incorporation and first capital raised, but, if you cross-reference your ISIC 6201 firms against databases like Crunchbase or local venture registries, you can filter out those that attracted seed funding for financial products.
This step, while data-intensive, narrows things down considerably. If a company registered as “software development” receives a $500,000 investment for, say, a peer-to-peer lending platform, there’s little ambiguity about its identity. It is, for all practical purposes, a fintech. Of course, there are exceptions: some fintechs are bootstrapped or funded by parent corporations, and these slip through the cracks. But, for a baseline measure, seed-funding data is one of the better signals available.
Integrating these datasets, however, isn’t clean work. Names change, founders launch multiple ventures, databases don’t always align. In my experience, the matching process is a blend of algorithmic filtering and human review. There’s always a handful of records that just don’t fit neatly—companies that look like fintechs, but aren’t, and vice versa. I wouldn’t trust any dataset from this era that claims perfect accuracy.
There’s also the problem of timing. A company registered in 2010 might not launch its actual product until 2011 or later. Some are shells, held in reserve for future ventures. Others pivot dramatically: what starts as a generic IT services company suddenly becomes a digital payments provider after a change of heart—or strategy—by its founders. Should we count these? It depends on the purpose of your analysis, honestly. For trend mapping, I lean toward inclusion. For impact assessment, perhaps more selectivity is warranted.
Let’s talk guidelines. First, don’t expect ISIC codes to do your job for you. They’re a starting point, not a solution. Second, when integrating early seed-funding data, document your methodology clearly—every assumption, every filter. The next analyst (maybe your future self) will thank you. Third, use keywords with care. “Payments,” “finance,” “lending,” “bank,” “remittance”—these are all useful, but context matters. “Bank” in a company’s name does not always mean financial services; sometimes, it’s a surname or a historical oddity.
I realize this all sounds rather painstaking, and it is, but I’ve yet to find a shortcut. The closer you look, the more you realize that classification, at least in the early days of an industry, is as much art as science. If you have doubts, you’re probably approaching it correctly. Overconfidence is the enemy of good data.
On a policy level, the lesson here is that registration systems need to evolve with the economy. If regulators want a clear picture of emerging sectors, allow new companies to self-identify with secondary or even tertiary industry tags. Free-text business descriptions are a rich source for future analysis, especially as natural language processing improves. But even then, don’t expect perfection.
Quantifying 2010 fintech entrants with ISIC 6201 is possible—if imperfect. Start with computer programming activities, layer in early-stage funding, and apply as much human judgment as you can spare. Accept the uncertainty; document it; move forward anyway. I have always found that clarity comes, if at all, only in hindsight.