Why free market data ends up costing more than paid data

Free market data is not a scam. It is a different product, built to a different standard, for a different purpose — and it works perfectly well until the moment your decisions start depending on it. The cost is real, it is just paid somewhere other than the invoice.

Why free data exists at all

Nobody gives away market data for charity. Free feeds exist because they serve someone's business model: a broker wants you trading on their platform, a website wants your pageviews, an app wants your attention. The data is a customer-acquisition cost, not a product with its own quality budget.

That matters, because quality costs money in specific, unglamorous ways. Recording every tick rather than sampling costs storage. Backfilling a dropped connection costs engineering. Applying corporate actions across intraday history costs a process somebody has to maintain. When data is a marketing expense rather than a product, those line items are exactly what gets trimmed.

So the question is not whether free data is dishonest. It is which of those line items were trimmed, and whether the ones that were matter to what you are doing.

The four places it shows up

1. Sampling — you are seeing a fraction of the trades

Most free feeds do not deliver every trade. They deliver a snapshot every second, or every few seconds, or every time the price changes by some threshold. For watching a price move, that is entirely adequate.

It stops being adequate the moment you use anything volume-based. Volume profile, market profile, VWAP and every order-flow technique ask how much traded at this exact price. Answer that from sampled data and your point of control can land at the wrong level. You are not getting a slightly blurrier picture; you are getting a picture of a different session.

2. Gaps — the data that quietly is not there

Connections drop. That is not a fault; it is the internet. The question is what happens next.

A paid feed generally backfills: on reconnect it sends what you missed, and your local history has no hole. A free feed usually just resumes. The two minutes you were offline are gone permanently from your database.

The cost is deferred and therefore invisible. You do not notice on the day. You notice months later when a backtest produces a result that cannot be reproduced, and you cannot tell whether the strategy changed or the data did.

3. Corporate actions — the cliff that is not a move

When a stock splits 1:5, a ₹2,500 share becomes a ₹500 share overnight. Nothing happened economically. If historical prices are not restated, your chart shows an 80% collapse that never occurred.

Most free feeds do adjust the daily series, because that is the visible one. Far fewer adjust intraday history. So your daily chart looks correct and your five-minute chart of the same stock has a cliff in it — and any indicator reading that intraday series is computing on a fiction.

4. Reprints — history that changes after the fact

The most corrosive one, because it undermines the thing data is for.

If a closed candle can change later, your backtest is not reproducible. Run it in March, get one number. Run the same code on the same period in June, get a different number. Nothing in your strategy changed — the past did.

You cannot debug this. You cannot even detect it unless you happen to have recorded the earlier values. It simply erodes your confidence in every result you produce, without ever announcing itself.

Where the money actually goes

The costs of free data are all deferred, which is exactly why they are underestimated:

• Time. Hours spent investigating a discrepancy that turns out to be the feed, not your code. This is the largest cost for most people and nobody ever counts it.
• Bad decisions. A support level drawn from a wrong point of control is a level you will trade against with real money.
• False confidence. A backtest built on gappy data can look excellent. You size up on the strength of it. That is the expensive one.
• Rework. Discovering the problem after building a system on top means rebuilding, not patching.

Set against that, a data subscription is a small, predictable, visible number. That visibility is precisely why it feels expensive and the alternative does not.

When free data is genuinely the right choice

Not everyone needs to pay, and it would be dishonest to pretend otherwise.

Free is fine if you are learning and want to see what a chart looks like; you trade daily or weekly and hold for months; you are checking prices rather than analysing structure; you are not backtesting anything you will risk money on.

It stops being fine when you trade intraday, so the detail inside the bar is the thing you are trading; you use volume-based analysis, so distribution is your signal rather than decoration; you backtest before committing capital; or you are building something for other people, where a data fault becomes a support ticket and a refund.

The honest version of the argument is not free data is bad. It is that free data is priced for one use and you may be putting it to another — and the mismatch is invisible until it is expensive.

How to check what you actually have

Five questions, and any vendor worth using answers them quickly in writing:

1. Is your tick data recorded, or reconstructed from bars?
2. What happens to my history when my connection drops?
3. Are corporate actions applied to intraday data, or only the daily series?
4. Can a closed candle ever change afterwards?
5. Does open interest arrive with the price, or as a periodic snapshot?

You can test three of those yourself in an afternoon — that is what a trial period is for. Record some bars, come back next week, compare. Pull your network for two minutes during market hours and watch whether the gap fills. Load a stock that split and compare resolutions.

Whatever you conclude, conclude it deliberately. The worst position is not using free data; it is not knowing which kind you are using.