Methodology

Mizaan is algorithmically curated market intelligence. The signal — clustering, ranking, entity extraction and the session recaps — is computed deterministically from data we hold, with no model in the loop. On top of that, each story carries an AI-written summary: original prose synthesised from the story’s facts and its cited sources, clearly labelled and never copied from source articles. Rankings are facts about coverage; summaries are AI interpretation — we keep the two visibly distinct, and the iris accent marks what we computed.

AI summaries can be wrong. They are informational only, not investment advice, and every one links the underlying sources so you can verify.

Where the data comes from

Market prices, company profiles and historical candles come from free-tier market data providers, fetched on a schedule and served from cache — a page never calls a vendor directly. News is aggregated from official RSS and API feeds listed in our source registry, weighted heavily toward regulators, central banks, exchanges and statistics authorities. Every source carries an explicit licence note, and we store headlines, short deks (≤ 40 words), timestamps and links only — never the body of a third-party article.

How clustering works

Every incoming headline is classified against committed dictionaries (tickers, entities, topics) and embedded locally. We then group headlines that describe the same event within a rolling 48-hour window, using semantic similarity plus a boost when stories share the same tickers or entities. One central-bank decision covered by nine outlets becomes one cluster, not nine. A cluster’s title is always the headline of its highest-trust source — we never synthesise a title.

How the importance score is calculated

The 0–100 score is a documented, weighted formula:

• Distinct source count — how many independent outlets cover the story. This is the single strongest signal, and we show it on every cluster.
• Source tier — regulators, central banks and primary sources count for more than secondary outlets.
• Market impact — the size of the move in the instruments the story concerns.
• Primary-source presence — a bonus when a regulator or central bank is among the sources.

The result is multiplied by a recency factor that decays with age but never falls to zero, so heavily covered news doesn’t vanish the moment it ages.

What “computed” means

Our session recaps and the per-instrument “how it moved” blocks are written by a deterministic template engine from our own numbers — the same inputs always produce the same words. The engine chooses its sentence form from the shape of the data (a streak, a reversal, a new high or low), never asserts a cause the data doesn’t support, and says less rather than padding with filler. Because each recap is reproducible from its stored inputs, you can always trace a sentence back to the figures behind it.

Delays and currency

Every price states its delay and source. Global exchanges carry a live layer; GCC exchanges (QSE, Tadawul, DFM, ADX) have no reliable free intraday feed and are shown as end-of-day, clearly labelled as delayed. All values are stored in USD; QAR is a presentation conversion at the official peg of 3.64 QAR = 1 USD (reviewed 2026-01-01). Percentages, ratios and index levels are currency-neutral and are never converted.

Questions about a specific number? Most figures link back to their source or their instrument. Start from the wire or markets.