How we work
A Market Insights page has one job: put the facts about an asset in one place, in a form you can check. Three rules keep it that way.
Numbers come first
Every figure on a page, from the RSI to the score, is calculated from market data before a word is written. CleaRank Financial AI never makes up a number. It explains numbers that already exist.
One method for every asset
Apple and a small-cap miner go through the same steps with the same thresholds. Nothing is tuned by hand to make one asset look better or worse.
Say what it is
A simulation is labelled a simulation. A score describes data. An analyst target is credited to analysts. None of it is a personal recommendation.
Where the data comes from
Prices, financial statements, earnings history, analyst ratings and insider filings come from TwelveData, a licensed market data provider. For cryptocurrencies we also use CoinGecko when a coin's history is not available from TwelveData.
- What a day means. Stocks, ETFs, futures and currency pairs use the daily bar the exchange or provider publishes for that trading day. Crypto trades all the time, so a crypto day runs from 00:00 to 24:00 UTC.
- Before a page updates, the data is checked. If a feed returns something that doesn't make sense, such as a price that jumped for no reason or a field that went missing, the refresh is stopped and the page keeps its last good version.
- Prices on the page move between refreshes. The quote at the top updates while you read. The analysis below it is tied to the price at the time it was written, which the page shows next to the date.
The indicators
We use the textbook definitions, calculated on daily closing prices, so you can reproduce any of them in a spreadsheet or on a charting platform.
| Indicator | What we calculate |
|---|
| RSI (14-day) | Wilder's Relative Strength Index over 14 days. The first average gain and loss are the simple average of the first 14 changes; after that each day is smoothed as (previous average × 13 + today) ÷ 14. RSI = 100 − 100 ÷ (1 + average gain ÷ average loss). |
|---|
| SMA (50 and 200-day) | The plain average of the last 50 or 200 daily closes. Price above both, with the 50-day above the 200-day, is the classic uptrend. |
|---|
| EMA (20-day) | An average that weights recent days more, with a weight of 2 ÷ 21, started from the simple average of the first 20 closes. |
|---|
| MACD (12, 26) | The 12-day EMA minus the 26-day EMA. Above zero means the short-term average is above the longer one. |
|---|
| CCI (14) | The Commodity Channel Index on the typical price, (high + low + close) ÷ 3, with the standard 0.015 constant. |
|---|
| Stochastic %K (14) | Where the latest close sits inside the last 14 days' high to low range, from 0 to 100. |
|---|
How we read RSI. 70 or above is overbought territory and 30 or below is oversold. In between is neutral, and the text may call 60 and above bullish momentum and 40 and below bearish momentum. The headline, the signal tiles and the written analysis all read the same stored value, so they always agree.
The CleaRank score
The 0 to 100 score at the top of each page sums up what the data says today. It comes from the CleaRank Financial AI scoring model, a fixed set of rules we built and test, so the same inputs always produce the same score. The part of the system that writes text never sets or changes it.
How it is built
- 1
Each factor below is turned into a reading from −1 (strongly negative) to +1 (strongly positive).
- 2
The readings are combined using the weights for that type of asset. If a factor has no data, for example a stock with no analyst coverage, it is left out and the other weights are scaled up to fill the gap.
- 3
The score is 50 plus 50 times that weighted average, rounded and kept between 3 and 97.
| Factor | What it looks at | Stocks | ETFs | Crypto, forex, commodities |
|---|
| Trend | Price against its 50 and 200-day averages, and the 50-day against the 200-day. Distances are scaled to the asset's own volatility, so a currency sitting 0.2% under its average isn't treated like a stock 10% under. | 25% | 40% | 50% |
|---|
| Momentum | Which RSI band the asset is in, and whether MACD is above or below zero. | 15% | 25% | 30% |
|---|
| Valuation | P/E and price-to-sales against the sector median. Within 20% of the sector P/E is neutral. Losses count against. | 15% | – | – |
|---|
| Revenue growth | Revenue growth against the sector median. | 10% | – | – |
|---|
| Profitability | Net margin, return on equity against the sector, and debt to equity. | 10% | – | – |
|---|
| Earnings record | How often recent reports beat estimates, and the latest surprise. | 10% | – | – |
|---|
| Analysts | The split of buy, hold and sell ratings, and the median price target against today's price. Used only when at least 3 analysts cover the stock. | 15% | – | – |
|---|
| Performance | The 1-year return, or 6 months when there isn't a year of history. | – | 20% | 20% |
|---|
| Cost | The expense ratio. 0.10% or less scores best; above 1% scores worst. | – | 15% | – |
|---|
A worked example (illustrative numbers)
A stock trading above both moving averages (trend +0.4), with neutral-to-firm momentum (+0.1), a P/E well above its sector (valuation −0.6), growth a little ahead of peers (+0.3), solid margins (+0.4), a good earnings record (+0.5) and mildly positive analysts (+0.2).
Weighted together: 0.25×0.4 + 0.15×0.1 + 0.15×(−0.6) + 0.10×0.3 + 0.10×0.4 + 0.10×0.5 + 0.15×0.2 = 0.175. The score is 50 + 50 × 0.175 = 59, which reads as Hold. The strong business is held back by the expensive valuation.
What the labels mean
80 to 97Strong Buy
60 to 79Buy
41 to 59Hold
21 to 40Sell
3 to 20Strong Sell
A label is shorthand for where today's data sits on the scale. It is not advice to buy or sell, and it knows nothing about your goals, timeline or situation.
Keeping a record
From October 2026 we store every asset's score each day. Once there are enough months of history, we will publish how the scores compared with what prices did next, whatever that shows. A score that has never been tested shouldn't be trusted, including ours.
How each asset class is analysed
The same principles apply everywhere, but a bank stock, a bond ETF, Bitcoin and the euro behave very differently. Each page type has its own checks, its own score weights and its own limits.
Stocks · 580+
Stocks
A stock page combines price behaviour with the business behind it.
- Trend and momentum: price against its 50 and 200-day averages, RSI, MACD and the other indicators above, run as simple pass or fail checks.
- Valuation and quality: P/E, price to sales, margins, return on equity and debt, each compared with the median of the company's sector so a software firm is not judged against a utility.
- Revenue and earnings: recent results, how often the company beat estimates, and an earnings reality check showing how the share price actually moved after past reports.
- Wall Street: the analyst consensus, the spread of 12-month price targets and recent rating changes, credited to the analysts.
- Insider trading: recent buying and selling by directors and officers from public filings.
- The long record: drawdowns and how long recoveries took, growth of $10,000 against the S&P 500, and seasonality when there are three or more years of history.
ETFs · 160+
ETFs
A fund is judged on what it holds, what it costs and how it has performed, not on company fundamentals.
- Fund facts: net assets, expense ratio and yield, with what the fee costs a year on $10,000.
- What's inside: sector weights, so you can see whether a "broad" fund leans heavily on one industry.
- Returns: total returns as the fund family reports them, next to price-only returns, clearly labelled.
- Score: trend and momentum carry most of the weight, with 1-year performance and cost making up the rest.
Crypto · 80+
Cryptocurrencies
Crypto has no earnings or analysts, trades every hour of every day and moves far more than most stocks.
- Market position: market capitalisation, rank and trading volume, with CoinGecko filling gaps in a coin's history.
- Days run on UTC: a crypto day closes at 00:00 UTC, and pages update every day of the week.
- Wider risk levels: trade plan stops start at 7% or more, because a tight stop on a coin that moves 5% a day is noise, not protection.
- Young coins: simulations never run longer than the coin's own history, so a token with one year of data gets no ten-year projection.
Forex, metals and futures · 20+
Currencies, metals and commodities
A currency pair or a metal has no earnings and no built-in tendency to rise over time, so it gets its own treatment.
- Key levels: recent highs, lows and moving averages that traders watch.
- Trend scaled to volatility: a currency 0.2% under its average is a small move, and the score treats it that way.
- No built-in growth: simulations keep the middle path at today's price and show only the range volatility allows.
- Leverage warning: trade plans note that leveraged losses can exceed a deposit and that news can gap a price through a stop.
Comparisons · 140+
Head-to-head comparisons
A comparison page runs both assets through the same method on the same dates: price rebased to a common start, key metrics side by side, momentum and sentiment, risk measures, long-term scenarios and a trade plan for each. The verdict explains which one the data favours on each point rather than declaring a winner on everything, because the better choice depends on what you are looking for.
Trade plan levels
The trade plan turns the score and recent volatility into example levels. They show how far a stop and targets would sit for this asset at today's price, so you can size a position with your own risk. They are not instructions.
- Direction. A Buy or Strong Buy score gives long levels, Sell or Strong Sell gives short levels, and Hold gives a neutral plan.
- Entries. One entry at today's price and a lower-risk entry about 2% better (1% for a neutral plan).
- Stop distance. Starts at 3% for most stocks, 7% for crypto and other assets that move more than 4% a day, and 12% for those moving more than 10%. It is never tighter than 1.5 times the asset's average daily move, is widened slightly when RSI shows a stretched market, and is capped at 30%.
- Targets. 5%, 10% and 15% from the entry for a long or short plan, and 3%, 5% and 8% for a neutral plan.
- Position size. The calculator starts at 1% of your account at risk per trade, which you can change. Trade plans are only shown when there are at least 50 days of price history.
Simulations
Price scenarios (Monte Carlo)
The scenario table and chart show a range of possible outcomes, not a forecast. They come from a standard statistical model of price movement, using the asset's own daily volatility from its history (capped at 50% a year).
For the middle path, the asset's past growth rate is blended with a long-run market average of 8% a year. The more history an asset has, the more weight its own record gets, up to five years. The result is capped at about 12% a year. Currencies and commodities get no growth assumption at all: their middle path stays at today's price and only the spread widens.
We show the 10th percentile (a poor outcome), the median and the 90th percentile (a strong outcome). A projection never runs longer than the price history behind it, and the yearly price table stops at five years.
Growth of $10,000
What $10,000 put into the asset at the start of its available history (up to 10 years) would be worth today, next to the same amount in the S&P 500 through SPY. Where dividend data is available, the figures assume dividends are reinvested and the page says so. An S&P 500 fund is not compared against itself.
Seasonality
Average returns by calendar month. They are only shown when there are at least three years of history, because fewer years produce patterns that are mostly noise.
Analyst forecasts
When Wall Street analysts cover a stock, the page shows their consensus: how many rate it a buy, hold or sell, and the median of their 12-month price targets, with the number of analysts. These come from the data provider and are clearly credited to analysts, not to CleaRank. Where no analysts cover an asset, we say so and don't invent a target.
CleaRank Financial AI and the written analysis
CleaRank Financial AI is the analysis system we built to cover hundreds of assets with the same care every day. It has three parts: the data checks that decide whether a refresh can go ahead, the scoring model that produces the 0 to 100 score, and a language model that turns the results into plain English. Every page is built from a fact sheet our system prepares first, holding every number the page will use: prices, indicators, the score and the reasons behind it, valuations and analyst data.
- 1
The facts are computed first. Every figure is calculated from market data before any writing starts. Each day this means working through millions of price, volume and fundamental data points across the 840+ assets we cover.
- 2
The text is written from those facts. CleaRank Financial AI explains what the numbers mean and how they fit together. It cannot add a number that is not on the fact sheet.
- 3
Every sentence is checked automatically. Numbers, tickers, dates and directions are compared against the fact sheet. Anything that does not match is corrected or removed before the page publishes.
- 4
Our team reviews the analysis every day. Accuracy comes first for us, so people check the daily output as well as the rules behind it. When something is wrong we fix it at the source, so the same mistake cannot appear on other pages.
Updates and dates
Every page is refreshed daily. For stocks, ETFs and futures, new figures arrive after each trading day's close, so the "Last updated" date moves on trading days and stays on Friday's date over a weekend. Crypto updates every day. The date shown at the top of a page, in its structured data and in our sitemap is always the time the figures last changed, not just the time the page was redrawn.
What this analysis can't do
It can't see the news first
The method works on prices and reported figures. News, guidance and events show up only once they reach the data.
It doesn't know you
Nothing here accounts for your goals, timeline, taxes or risk tolerance. Treat it as research, not advice.
The past is only a guide
Indicators, simulations and seasonality all lean on history, and markets change. Wide ranges are there for a reason.
Who reviews the analysis
The method and the daily output are reviewed by the CleaRank team: market strategists, analysts and financial journalists who check the figures, the wording and the rules behind them. Each Market Insights page names the team member responsible for it.
More about the people and the company on our About us page.
Spotted a mistake?
Every Market Insights page has a "Report an error" link that opens an email with the page already filled in. You can also write to info@clearank.com. We check every report, fix the page and its stored data, and look for the same problem on other pages.
See also our editorial code for how we publish news and reviews.
