Methodology

What TickerWhale analyses

Every trading night we rebuild the research on every company we cover, from five kinds of information:

• Markets & options

  Prices, volume, volatility and what the options market is pricing

• Company & earnings

  Filed fundamentals, margins, earnings and guidance

• Wall Street

  Analyst consensus, rating changes and target revisions

• News & developments

  Company news and events, grouped and attributed

• Crowd & social

  Retail sentiment and attention, where available

The five research dimensions

The TickerWhale Score is a single 0–100 reading built from five dimensions, each analysed on its own terms and then weighted into one number. It answers “how strong is the case for this business right now” — it is not a price target and not a recommendation.

Every dimension has to be there. If one genuinely cannot be computed for a company on a given day, we publish no score for it rather than a partial one — and no weight is quietly moved onto the dimensions we do have.

How companies are compared

A margin or a multiple means nothing on its own — it only means something next to something else. So each figure is measured against the company's real peers rather than a fixed benchmark: its industry first, widening to its sector or the broader covered universe when there are too few comparable companies reporting that particular figure.

Most of the dimensions are then ranked across everything we scored that day, so a factor reading of 70 places a company in the top third of the market we cover rather than scoring it in isolation. Extreme outliers are trimmed before any average is taken, so one unusual peer cannot distort a whole group.

Where no real peer comparison is possible, the figure is left out of the score rather than filled in from an assumed value.

The TickerWhale Model Target

A modelled view of what a company is worth, expressed as a range with a central estimate.

• Several independent methods, not one. Cash-flow models, earnings power, peer multiples and forward estimates each produce their own answer.
• The methods fit the business. A bank is not valued like a software company, and a property trust is not valued like either. A fast-growing company that is not yet profitable is valued on what it is actually doing rather than on earnings it does not have.
• The methods cross-check each other. They are blended rather than picked between, and a method that lands far from the rest is pulled back toward the group so it cannot dominate the answer.
• Safeguards against thin or extreme evidence. When the models disagree sharply, or the answer implies an improbably large move, the estimate is reined in — and the stock page tells you a safeguard applied rather than hiding it.
• No number when the evidence is not good enough. If too few methods produce a credible answer, we publish nothing for that company instead of a figure we would not stand behind.

Each estimate shows how many independent methods stood behind it and how wide the range is — more useful than a single confidence word, because they tell you why to trust it or not. The Model Target is a Pro feature; everything else on this page is free.

Where AI is used

AI does real analytical work at TickerWhale. It reads and groups the day's news, interprets what a development means for a company, weighs several pieces of information against each other, writes the daily market read and decides which stories led it, interprets the crowd where we have it, and answers your research questions through Willy.

What AI does not do is set or alter TickerWhale's canonical scores, ratios, rankings or valuation outputs. Those are produced by the model and data pipelines, and the AI is given the finished figures to work from.

That line is enforced rather than requested. Before written commentary reaches the site it is automatically checked back against the stored numbers: a figure that does not match, or a claim about a company's strongest factor that is not true, blocks publication.

Numerical integrity

• Scores are published once per trading night as a single complete run, and never for a session that has not finished.
• Every figure carries the data behind it — a stock page shows the market session and the timestamp of its numbers; ranking pages are stamped with the trading day they came from.
• Live prices are shown as live and never overwrite the official daily record.
• Identity is checked, not assumed: a defunct company's record cannot overwrite the current issuer, and a price and share count from different share classes are refused rather than published.
• A news story is only attached to a company when the article itself provides enough evidence. A data provider's own tag is never sufficient on its own.

Limitations — worth reading

• Scores and value ranges are model estimates built from historical and modelled data. They are not investment advice, and they can be wrong.
• A score reflects the most recent data we hold, so it can lag a breaking event or a newly filed set of accounts.
• Value ranges are sensitive to assumptions. Small changes in growth or discount rates move them, which is why the width of the range and the number of methods behind it matter more than the single central figure.
• Scoring uses the most recent revised filings — current analytics rather than a point-in-time record. Backtests are not available.
• News matching is deliberately conservative: we would rather miss a story than attach one to the wrong company, so a real story can occasionally be muted.
• The crowd reading is available for the companies with enough recent activity to measure. Where it is absent, the sentiment dimension is built from news and Wall Street alone.
• Coverage is limited to the companies we track. The absence of a score is not a negative signal.