AltIndex

Not investment advice. This is a review of a software product, not a recommendation to buy, sell or hold any security. Nothing here is financial advice, and past performance figures quoted by any vendor are not a reliable indicator of future results.

Signals, Features And Scale

What Alternative Data Is, And Why Anyone Bothers

The premise is a decade old on the institutional side and only recently affordable to individuals, which is the actual news here.

Company financials arrive quarterly, audited and late. By the time a filing tells you revenue rose, the rise happened months ago and the price moved on the announcement. Everyone gets that information simultaneously, which is precisely why it is hard to profit from.

Alternative data is everything else — the observable traces a business leaves while operating. If a retailer's web traffic climbs for eight weeks, something is happening before any filing says so. If a software company posts forty new engineering roles, that is a statement about its plans. If employee ratings fall sharply, that often precedes visible trouble.

Hedge funds have bought this for years: satellite imagery of car parks, credit card panels, app telemetry. It is expensive, and the expense was the moat.

AltIndex's proposition is packaging a retail-accessible version of that idea, in the way Danelfin packages a different feature set. The signals are shallower than an institutional dataset, but they are real signals rather than opinion, and that is a meaningful difference from the average stock-tip service.

Financial statements tell you what already happened, on everyone's screen at once. Alternative data is an attempt to watch the business rather than the report.

Reading The 61% Win Rate Properly

This figure is the centre of the marketing and it deserves neither dismissal nor acceptance. It deserves unpacking.

First, win rate is not return. A strategy can win 61% of the time and lose money, if the 39% of losses are larger than the wins. Win rate without average win and average loss is an incomplete description of any strategy — and the missing half is the half that determines whether you made money.

Second, compared with what? In a rising market a randomly chosen basket of large-cap stocks wins more often than it loses. The number that matters is performance against a relevant benchmark over the same period, not against zero.

Third, over what period, and was it live? A result derived by testing rules against history is a very different claim from one recorded in advance and audited afterwards. Backtests are famously flattering because the rules were chosen knowing the outcome.

Reading a company's actual filings — with a long-context model like Claude, say — remains the unglamorous alternative. Fourth, who is measuring? This is the vendor's own figure about its own product, unaudited by anyone else.

None of that means the number is dishonest. It means the number, as published, cannot tell you what you would need to know, and any investment tool advertising a headline performance statistic should be read this way as a matter of routine.

Where These Signals Genuinely Work — And Where They Mislead

The honest position is that alternative data is strong for a particular kind of company and close to useless for others.

It works best on consumer-facing businesses with a digital surface. App downloads genuinely track adoption for a mobile-first company. Web traffic genuinely tracks interest for an e-commerce retailer. Job postings genuinely signal expansion. For these, the data is close to the actual business.

It works poorly on B2B, industrial and regulated businesses. A company selling turbines to utilities has no app, and its website traffic tells you nothing about a sales cycle measured in years. The signal is not weak here, it is absent — and a scoring system that produces a number anyway is producing noise.

Reddit sentiment deserves particular scepticism. It measures attention, not quality, and the 2021 meme-stock episode demonstrated exactly what following retail social sentiment can do in both directions. High Reddit volume tells you a stock is being discussed — which is genuinely predictive of volatility, and not at all predictive of direction.

Insider and congressional trade tracking is the most defensible signal in the set, because it is disclosed regulatory data about what people with unusual visibility actually did with their own money. It is also public, lagged by filing deadlines, and widely watched.

The pattern to internalise: these signals are best as a prompt to investigate, and worst as a reason to trade.

• Strongest on consumer apps, e-commerce and digital-first companies
• Weakest on B2B, industrial, and long-sales-cycle businesses
• Reddit sentiment predicts attention and volatility, not direction
• Hiring data is a decent expansion signal and a slow one
• Insider and congressional filings are the most solid input, and also the most public

Autopilot And The MCP Server Are Different Risks

Two features sit apart from the research tooling and each deserves its own note.

Autopilot applies rules to copy a portfolio automatically. Automation in investing is not neutral — it removes the pause between a signal and a decision, and that pause is where most people avoid their worst trades. If you use anything like this, position sizing and a hard limit on what any single rule can do matter far more than the rule itself.

The MCP Server is a more interesting inclusion and a genuinely modern one. It lets an AI assistant query AltIndex's data directly, so you can ask a general model about your portfolio and have it answer from this dataset.

That is useful and it introduces a specific failure mode worth naming: a language model given data will produce a confident narrative from it, including from signals that are noise. A cited research tool such as Perplexity at least shows you its sources. The model does not know that web traffic is meaningless for an industrial supplier. It will explain the number anyway, fluently.

Treat generated commentary as a summary of the data, never as analysis of the business.

1. Start on the free tier and score a portfolio you already hold and understand.
2. Check whether the scores agree with what you independently know about those companies.
3. Look specifically at how it scores a B2B or industrial holding, where the signals should be weakest.
4. Track the picks on paper for a full quarter before any money follows them.
5. Compare that paper result against a plain index over exactly the same window.

What A Subscription Actually Buys

This site does not publish plan prices. The structure is what stays true, and it is conventional here.

AltIndex is freemium with subscription tiers. There is a free level — including a screenshot-based portfolio scoring path that lets you see a score without connecting anything — and paid monthly or annual levels above it. Annual is offered, and no card is required to start.

What is metered is depth of access rather than usage: how many holdings you can score, which signals and alerts you can see, and whether features like Autopilot and the screener are available. That is the normal shape for research subscriptions and it means the free tier is a genuine preview rather than a trial clock.

There is a separate institutional offering for funds and ETF issuers, which is a different conversation and a different contract.

The one budgeting point specific to investment tools: the subscription is a certain, recurring cost set against uncertain, variable benefit. On a small portfolio an annual research subscription can be a meaningful percentage of returns, and that arithmetic is worth doing before subscribing rather than after.

Vocabulary This Category Assumes

Four terms do most of the work, and two of them are routinely misused in marketing.

Alternative data

Information about a company from outside its financial reporting — web traffic, app downloads, hiring, sentiment, imagery. Used to observe a business between filings.

Win rate

The share of picks that ended profitable. It says nothing about size, so a high win rate is fully compatible with losing money overall.

Backtest

Applying a strategy to historical data. Flattering by construction, because the rules were chosen already knowing what happened.

Insider trading data

Here it means legally disclosed transactions by executives and legislators, filed with regulators — public record, not anything improper.