The ETF market is growing fast. Global assets rose 33% in 2025 and, on PwC's projection, is expected to reach $35tn by 2030, up from roughly $22tn today. Growth on that scale does not make the data describing the market cleaner. It does the opposite, and the infrastructure serving the industry has not scaled to keep up.

You can see the strain in something as basic as the headline count. Ask the major ETF data providers how many funds trade globally and the answers range from about 13,700 to more than 15,800, a spread of over 2,000 products, because each one counts something slightly different, from exchange-traded products that are not funds through to separate share classes. None of the figures is wrong; they simply describe the market through different lenses. What is missing is a single source the whole market reconciles back to, and the gap grows as the market does.

The market grew - the plumbing did not 

The growth itself is well documented. PwC puts global ETF assets at around $19.5tn at the end of 2025, after record net inflows of $2.1tn over the year, and in its survey of 72 industry executives most expect the market to be in the range of $30-35tn by the end of the decade. What those headline numbers hide is how much faster the data describing the market grows than the assets themselves.

A single fund cross-listed on Xetra, the London Stock Exchange and Borsa Italiana produces separate trading records on each venue, in different currencies, under different tickers, settling against different calendars. Multiply that across share classes, the launches and closures that happen every day, and a steady run of corporate actions, and the volume of reference and market data the industry generates climbs well ahead of AUM. There is plenty of data. It is just scattered.

What fragmentation actually looks like 

Fragmentation is not a vague complaint about too much data. It has a specific shape, and anyone who has tried to build a clean view of the European market already knows it.

The data comes from thousands of sources: exchanges, issuers, index providers, regulators, custodians, and a long tail of private and quasi-public feeds. No common standard governs how any of it is published, so formats run from structured files to PDFs to web pages built for human eyes rather than machines. Identifiers do not line up, so the same instrument carries different codes across systems, and mapping them is manual work that breaks the moment something changes. Markets trade across time zones on a 24/5 clock, in several currencies and languages. Frequencies range from sub-second tick data to quarterly holdings disclosures. Threaded through all of it are gaps, outliers, broken records, restated figures, and the churn of corporate actions, each of which has to be caught and checked.

So a large part of the working day, for the people who rely on this data, goes on assembling and cleaning it rather than using it. Reconciliation is constant, it is mostly done by hand, and it runs against a deadline.

Who pays for it 

The same structural problem lands differently depending on where you sit.

For an ETF issuer, fragmented data means competitive intelligence is always slightly stale. By the time flows have been pulled from several sources, cleaned and reconciled, a shift in market share can already be weeks old, and product and distribution calls get made on a lagging picture.

For a capital markets desk, the issue is venue fragmentation. One product trades across Euronext, Xetra, the London Stock Exchange, Cboe Europe, SIX and Borsa Italiana, and a clean cross-venue view of spreads and market-maker behaviour means stitching together exchange downloads and terminal screens. A spread blowout on one venue can sit unnoticed until an investor complains about an execution.

For a fund selector or portfolio manager, the universe runs past 10,000 products. A defensible, repeatable selection process means comparing cost, tracking, liquidity and holdings across that universe, and most of the effort goes on getting the underlying data into a usable state before any analysis starts.

For a fund administrator or custodian servicing these products, the cost is reconciliation: inconsistent identifiers and classifications across sources that have to be aligned before anything reaches a client report. Four different desks, one root cause. The data exists; nobody is working from the same version of it.

What a "single source of truth" actually requires 

"Single source of truth" is easy to claim and harder to build. A few things separate a real one from the phrase.

The first is bottom-up construction. If the data is assembled from the individual share-class level, the totals reconcile upwards: sum the share classes and you get the fund, sum the funds and you get the segment, sum the segments and you get the market, and every level ties out. Top-down estimates do not behave this way, which is part of why providers' headline figures disagree in the first place.

The second is ETF data standardisation. Identifiers and classifications have to be consistent across the whole universe, so that the same instrument is treated as the same instrument wherever it turns up.

The third is automation with validation built in rather than bolted on afterwards. Ingestion, processing, analytics and distribution have to run as one pipeline that catches errors, gaps and corporate actions as they occur, instead of waiting for someone to notice them later.

The fourth is delivery that meets each desk where it works: an API for systematic users, a web application for analysts, and a machine-readable interface for the tools that increasingly sit between people and their data.

None of this removes the need for judgement. The pipeline still needs people who know the market to validate what comes out of it. Automation handles the volume; expertise handles the edge cases.

How ETFBOOK approaches it 

This is the problem ETFBOOK was built around. The platform consolidates public, quasi-public and private sources into one structured system with ready-to-use analytics, serving the full B2B chain from issuers and asset managers through to liquidity providers and service providers.

Coverage today is Europe and the United States: about 3,700 ETF products in Europe and 5,400 in the US, more than 9,000 in total, accounting for roughly 99% of assets under management across the two regions. Asia-Pacific is on the roadmap. Because attributes are collected at share-class level, the data filters down by asset class, region, replication structure, ESG status and trading characteristics, so a user can move from the whole market to a narrow sub-segment without the numbers coming apart. Flows and trading activity sit in the same view. Delivery is T+1, one day between capture and availability, with operations running through the trading week.

The feature list is not really the point. The point is that the work of sourcing, cleaning and reconciling, the work that currently eats into so many people's days, comes off the client's desk.

The gap widens from here 

As the market climbs towards the numbers PwC describes, the distance between how much ETF data exists and how much of it is usable gets wider, not narrower. More products, more venues, more share classes, more of everything that has to be tracked. The question for anyone who depends on this data is changing. It used to be whether the data could be obtained at all. Now it is whether it can be trusted and acted on without a team reconciling it by hand first. 

That is a fair test to put to any data provider, this one included. Can the numbers be reconciled from the bottom up? Are identifiers consistent across the universe? Does validation happen inside the pipeline? Is the data where you actually need it? An industry that cannot yet agree how many ETFs exist will need good answers well before 2030.