You paid for the signup. The screen they land on is full of data and empty of direction.
Centaur · Research
August 21, 2026 · 10 min read
Trade discovery data is a structured record of the trades that public traders disclose, built so another product can query it. A pipeline reads what traders post on platforms such as X and Telegram, identifies the messages that describe an actual position rather than commentary, and converts each one into a typed event: an open, an increase, a decrease, or a close, carrying the asset, the direction, a price where one is available, a timestamp, and a link back to the original post.
Related events group into positions, and positions are scored against market data over fixed windows so that different traders can be compared on the same basis. What you get is not a sentiment score and not a buy signal. It is an auditable log of who took which position, when they said so, and what happened next, including the positions that went against them.
Start with the number that should be on the wall of every growth team in this industry. 774 million people owned crypto in August 2026, an all-time high.
Over the same stretch, spot volume across the top ten centralised exchanges fell from $4.5 trillion in Q4 2025 to $1.95 trillion in Q2 2026, a decline of more than 55% in two quarters (ownership per Crypto.com Research, August 2026; volume per CoinGecko's Q2 2026 industry report, July 2026).
More holders. Less trading. The industry has become efficient at acquiring people and much worse at activating them.
The spend behind that acquisition is not small and it is not flat. Coinbase reported $1.06 billion of sales and marketing expense in FY2025, up 62% year over year, against revenue growth of 9% (Coinbase FY2025 results, February 2026). Marketing grew roughly seven times faster than the top line.
And the funnel keeps a lot of what it acquires in a holding pattern. eToro, the largest social trading platform in the world, reported 40 million registered users against 4.02 million funded accounts in Q1 2026 (eToro Q1 2026 results).
That is one company's ratio rather than an industry rate, but if your own funded-to-first-trade number looks better, you already know which step is costing you the most.
The usual suspects get the blame. KYC friction, deposit friction, app performance. All real, all measurable, all already owned by somebody on your team.
The step that tends to have no owner is the one immediately after. The account is funded, the user is in, and the product presents a chart, an order book, a search bar and several hundred markets.
The screen is not empty. It is full of data and empty of direction, and none of it answers the question they walked in with: what should I trade?
That moment is more decisive than it looks, because the decision window is short. 66% of UK investors aged 18 to 40 finalise an investment decision within 24 hours, and 14% within 60 minutes (FCA, December 2024). For a large share of users the first session is not the top of a considered funnel. It is the whole funnel.
Now consider what the industry has offered in response, which is more markets. MEXC listed 879 new perpetual contracts in sixteen months, more than any competitor, and still fell from second place to seventh with volume more than halving (CoinGecko, State of Crypto Perpetuals 2026, May 2026).
Listing velocity has stopped working as a differentiator. To someone who does not know what to trade, a longer list is not a richer product. It is a harder question.
Off your platform, and then sometimes back. The reliance concentrates exactly where you least want it. 26% of US investors have acted on a social media personality's recommendation, but among investors with less than two years of experience it is 57%, against 10% for those with ten years or more (FINRA Investor Education Foundation, December 2025).
The cohort leaning hardest on outside voices is the cohort you just paid to acquire. They are not doing it out of loyalty to those sources either: only around 13% to 15% of Gen Z name social media or influencers as their most trusted source, despite roughly half using them to learn about investing (FINRA Foundation and CFA Institute, Gen Z and Investing, 2023).
Which brings us to the finding that reframes the problem. Among US retail investors who bought meme stocks or viral investments, 78% said they rely on the investments displayed in their mobile trading app, ahead of the 69% who cited social media groups (FINRA Investor Education Foundation, December 2025).
Your app is already the most influential source your users have. The question is not whether to shape what they trade. You are shaping it now, by omission.
Three surfaces, in rough order of integration effort.
Trader positioning on the asset pages you already have. A user lands on an asset and, alongside price and depth, sees how tracked traders are currently positioned in it, long against short, and which of them opened recently. The cheapest of the three, because the page already exists and you are adding a module to it.
A discovery feed of what just happened. A chronological stream of opens, increases, decreases and closes across the tracked universe, each item linking back to its source post. It gives a user with no thesis somewhere to start, and gives your app a reason to be opened on a day the user had no plan.
An auditable view of who is worth watching. Not a follower count and not a screenshot. A ranked view built from timestamped public positions, scored over fixed windows, with a minimum sample threshold and the losing calls left in. This is where most implementations go wrong, and the last section explains why.
Note what none of these are.
None is copy trading, and none executes anything. This is a discovery and context layer sitting in front of your existing execution, and that distinction is not cosmetic. It is the difference between a feature you can ship and a feature that changes your regulatory posture.
The ongoing cost is the problem, not the initial build.
| Source capture | X moved to pay-per-use pricing in February 2026, and developers on X's own forum have documented per-post pricing changing mid-integration |
|---|---|
| Extraction and deduplication | Commentary and general market takes must not become trade events, and traders cross-post the same call to several platforms. Without both, one call becomes three signals and your rankings are wrong |
| Position lifecycle and scoring | Adds, trims, exits, inferring state when a trader goes quiet, then scoring over consistent windows with a defensible answer for missing prices |
| Maintenance | Handles, formats, platforms and behaviour all change continuously. This never finishes, which is why the question usually resolves to buy |
There is a second cost that rarely makes the estimate: all of it has to be right before the feature ships at all. A discovery surface built on a noisy extraction layer does not degrade gracefully. It shows wrong positions to people who are making decisions, which is worse than shipping nothing.
If engineering time is already the constraint on your roadmap, and for most exchanges it is, that is the real argument for licensing this layer rather than building it.
The right instinct is decision support, not signals, and the difference is narrower than it sounds. A ranked view built on transparent, published criteria is information. The same view with a "top pick" badge on it, or a button that executes the trade for the user, is something else.
Keep the ranking mechanical, keep the losing calls visible, and skip the copy button, and you stay on the side of that line where most exchanges want to be. The full regulatory detail is a piece on its own; for now, that is the operating rule.
The three surfaces above are not hypothetical. They are what Centaur's API and MCP server expose today, over REST or directly to an AI agent.
The feed is trade events as they are extracted from public posts on X and Telegram, in order, each one linking back to the original message. Asset positioning is a live read of who is long and who is short on a given market right now, across every tracked trader, refreshed as new events land.
And rankings are built on win rate, median return and actual performance, not on raw accumulated returns or follower count, with a minimum number of tracked positions required before a trader appears at all, adjustable if you want to set the bar higher. A trader with a handful of lucky calls does not outrank one with a long, proven track record.
Every one of those is a single documented call. None of them executes anything, moves funds, or requires a change to how your product already handles trading. It is a layer that sits in front of what you have, and it is the layer between having several hundred markets and giving your users a reason to pick one.
Your acquisition funnel is working. The step after it is not. A funded user staring at several hundred markets has one question, your product does not answer it, and the places they go instead are ones they do not particularly trust. That is a discovery gap, not a UX problem, and it is the cheapest large improvement available to most exchanges right now.
It closes with three surfaces, and they do not require you to rebuild anything: positioning on the asset pages you already have, a feed of what just happened, and a ranked view built on verified performance instead of a follower count. Centaur already runs all three.
Your users are already following these people. Give them a defensible basis for it.
__________________________________________________________________________________
Talk to us about an integration.
Nothing in this article is investment advice. Centaur measures the trades a trader disclosed publicly, which is not the same as their complete trading record, and it does not verify broker or on-chain activity.
