Methodology
Both of our apps are built on the same conviction: a market view is only useful if you can take it apart. This is how that view is assembled, and where it stops.
The principle
Most tools resolve that complexity by hiding it — a single arrow, a colour, a verdict. We take the opposite position. Competing evidence is organised rather than flattened, so a user can see where factors align, where they conflict, how current the data is, and how much of the picture is actually present before forming their own view.
How a score is built
Each market has a defined set of macro inputs behind it — policy and rates, growth, inflation, positioning, risk appetite, volatility and price structure. Every input is gathered on its own schedule and stamped with the time it was published, not the time it was read.
Raw inputs arrive in incompatible units: percentages, index levels, contract counts, basis points. Each is converted onto a common scale relative to its own history, so a move in one input can be weighed against a move in another without the largest number automatically dominating.
Related inputs are grouped so that a single theme cannot overwhelm the result simply by being measured several ways. Five separate readings of the same interest-rate story stay one rate story, rather than becoming five independent votes.
Families are combined into a single reading on a fixed scale. The result is smoothed deliberately, because a research view that flips on every tick is not a research view — it is noise with a number attached.
The contribution of every factor is retained alongside the final figure. That is what makes a score inspectable after the fact: the number never travels through the app without the evidence that produced it.
Inputs
Scores are only as good as their inputs, so it is worth being specific about what those are and where they come from.
Policy rates, expectations for where they go next, rate differentials and real rates.
The level and shape of government curves, and how they are shifting between maturities.
Headline and core measures, plus the direction of travel rather than the latest print alone.
Survey data, production, labour-market readings and the momentum behind each.
Commitments-of-Traders data showing how concentrated the market already is.
Risk appetite across assets and how much uncertainty is currently being priced.
Trend and impulse, kept deliberately separate from the slower macro families.
The economic calendar with forecasts, previous values and post-release actuals.
Inputs are drawn from official macroeconomic releases, regulatory filings and licensed market-data feeds. We do not originate market data and we do not resell it — it is transformed into research context, and the age of every input travels with the reading it produced.
Confidence
Every score is published with a confidence reading, because a strong signal drawn from thin or contradictory evidence deserves to be treated differently from the same signal drawn from a full, agreeing dataset. Four things determine it.
Low confidence is information, not an error state.
What confidence measures
Do the factors point the same way, or is the score a narrow average of inputs that disagree?
How much of the expected evidence is actually present for this market right now?
How recent are the inputs, and how much has their influence decayed since publication?
Has the reading held together over recent updates, or has it been moving erratically?
Data quality
Inputs age at different rates — a positioning report and a policy expectation do not go stale on the same schedule. Each carries its own decay, so its pull on the score fades as it ages instead of counting fully until it is replaced.
Where a source fails or has not reported, the app reduces the affected reading and surfaces the gap. It does not substitute a neighbouring value or carry the last figure forward indefinitely to keep the display looking complete.
Around significant scheduled releases, readings for the affected markets are held back rather than presented as though the release has already been absorbed. The window is tied to the impact of the event and to the currencies it touches.
Every score shows when it was calculated. If the pipeline stalls, that is visible on screen as an ageing timestamp rather than concealed behind a figure that looks freshly computed.
Regime context
Before any individual market is assessed, both apps establish the broader state markets are trading in — how risk appetite, the dollar, volatility, yields, growth and inflation currently line up. That state is shown first, and individual scores are read against it.
A defensive reading during a broad risk-off phase is ordinary. The same reading while risk appetite is expanding is a conflict worth investigating. Without the regime layer, those two look identical.
Context first, instrument second.
Limits
A methodology page that only lists strengths is marketing. These are the boundaries of the approach, stated plainly, because knowing where a tool stops is part of using it well.
A score describes the balance of evidence as it stands. It does not predict where a market will be tomorrow, and it carries no target, no horizon and no implied probability of a particular outcome.
Nothing in either app tells a user what to buy, sell or hold. There are no signals to follow, no entries, no exits and no position sizing. The apps do not know a user's circumstances and do not attempt to.
This approach describes conditions and context, which move over days and weeks. It is a poor tool for intraday timing, and we would rather say so than imply a precision the method does not have.
Sources revise, arrive late or fail. Where that happens the apps degrade the affected reading and show the gap rather than filling it silently — but no amount of handling makes imperfect data perfect.
Being able to see why a score reads the way it does is valuable, and it is the point of the design. It is not the same thing as the score being right, and we do not present it as such.
The products
