What alternative data actually costs

A reader who has priced a market data API arrives at alternative data expecting the same shape: a pricing page, a free tier, a rate limit. There is none of that. This page is about what is actually there instead, what the market spends in aggregate, and what is available to somebody who is not a fund with a data budget.

How it works

Alternative data is any input to financial analysis that is not a price, a filing or a press release. The SEC's examination staff, in a 2022 risk alert, gives the working list: satellite and drone imagery of crop fields and retailers' car parks, analyses of aggregate credit card transactions, social media and search data, mobile geolocation, and email data from consumer apps.

The supply chain has three links and the reader is at the far end of none of them. Somebody collects exhaust data for another purpose — a loyalty programme, an app SDK, a payments processor, a satellite operator. A vendor licenses it, cleans it, maps it to tickers and turns it into a panel. A fund buys the panel under an annual contract.

The sales motion is the product's defining feature. Checked in September 2026, the pricing pages of the best-known vendors resolve to a demo request: Sensor Tower says plainly that it customises pricing plans for each customer, YipitData and Thinknum return a 404 on /pricing altogether, and Similarweb shows the split in one company — its marketing products carry a free trial, while Stock Intelligence, the product sold to funds, carries a talk-to-sales button.

The one vendor that describes its pricing model still gives no number. SafeGraph sets out the axis the whole industry prices on: an annual fee based on a custom mix of rows, columns and usage rights. That is the shape of every contract in this market, and knowing it is more useful than any single figure.

The sector is also consolidating, which matters for anyone evaluating a vendor. Orbital Insight's site now redirects to a different company. Earnest Analytics was absorbed into Consumer Edge in 2025. M Science states on its own About page that it is a subsidiary of Jefferies Financial Group — a bank owns one of the named independents. And App Annie, the subject of the enforcement action below, was renamed data.ai and later acquired by Sensor Tower.

What it costs

Nobody outside the contracts knows, and the published estimates come from firms that sell into this market. That is not a reason to discard them; it is a reason to quote them with the interest attached.

Neudata, a data-scouting firm that sells the report these figures come from, estimated in February 2025 that investment managers spent a minimum of $2.5 billion on alternative datasets in 2024, and that on an adjusted methodology the true figure could have been as high as $12.7 billion. Its 2025 projection ran from $3.3 billion to $15.4 billion.

That spread — five times between floor and ceiling, from one firm, in one sentence — is the most honest thing anyone has published about what this market costs.

At the dataset level the same firm puts the average at about $1.1 million a year, with a minority of datasets earning over $20 million a year from investment clients. Its own commentary adds the caveat that matters: the average is misleading, because revenue is concentrated in a small number of datasets. It also states what the methodology excludes — internal infrastructure, headcount, cloud services. The licence fee is the smaller half of the bill.

The only per-firm figure available is older. A 2022 survey by Lowenstein Sandler, a law firm that advises both buyers and sellers here, reported that nearly three in four users of alternative data spent between $1 million and $5 million a year on it. The firm's later survey publishes adoption rates rather than dollars, so that 2022 range is the most recent number of its kind and should be read as four years old.

A note on the figure you have probably seen. The widely repeated projection of buy-side alternative-data spend that circulated for years traced back to AlternativeData.org, a site historically operated by one of the vendors. That domain no longer resolves. Every article still citing it is citing something a reader cannot open, which is worth remembering whenever a market size appears without a document behind it. No regulator publishes a spend figure at all.

The cost that never appears on a price page

Diligence, and it falls on the buyer.

The SEC's examination staff has found advisers using data from non-traditional sources without written policies designed to address the risk of receiving material non-public information through it — conducting, in the staff's words, ad hoc and inconsistent diligence of providers, and lacking procedures for assessing the terms and legal obligations attached to how the data was collected. The same alert is careful to say that alternative data does not necessarily contain such information.

The other side of that arrangement is that the provider can be charged too. In September 2021 the SEC announced what it described as its first enforcement action charging an alternative-data provider with securities fraud: App Annie and its founder settled charges concerning how the company's app-performance data was generated and what customers were told about it, paying more than $10 million.

For a buyer, the practical reading is that a contract for a dataset carries a compliance obligation with staff time attached, and that obligation is priced into nobody's quote.

What you can do about it

Start by asking whether you need a panel at all. Most questions a private investor brings to alternative data — who is buying, who is selling, what a fund owns — are answered by the public disclosure record, which is free at source and catalogued here under insider, 13F and congressional trade trackers. The deadlines that make it late are set out in when a trade shows up in the public record.

Look for the consumer edge of the same vendor. Several of these companies sell a self-serve product to marketers built on data related to what the funds buy. It is not the investment product, the panel construction is different, and it is purchasable.

When you do ask for a quote, ask on the industry's own axis. Rows, columns, users and usage rights — specifically whether you may redistribute, publish or show derived figures to clients. That last term moves the price more than volume does, and it is the one most often discovered after signature.

Budget for the half of the cost that is not the licence. Storage, pipeline work and the compliance procedure the SEC's staff expects to see documented. A dataset that arrives as monthly files is a data-engineering commitment before it is an edge.

Treat every market-size figure as a sales document until you can open its source. Every number on this page is linked to the document it came from, and each of those documents was published by somebody with a position in this market. There is no neutral estimate, and a figure with no retrievable source is not a figure.