Purpose and methodology.
Where this came from
Most of my investing was either passive — index funds and mutual funds — or tips, where someone suggests a stock on the basis of something they have heard. I wanted a framework for investing in mid-cap stocks consistently, so I built one.
Framework overview
Four dimensions, applied in a fixed order, none of them optional. Sector is read first, because the same balance sheet means different things in adjacent industries.
Regulatory tailwinds
- Government policy and scheme analysis
- Budget allocation trends by sector
- Scored sector-level sentiment
- Demand mix: government-backed versus private
Quality score (0–10)
- Profitability and margin direction
- Cash quality: how much reported profit arrives as cash
- Growth durability, not growth headlines
- Balance-sheet resilience and working-capital discipline
- Ranked within sector peers, not against absolute thresholds
Network & ownership
- Institutional cross-holdings graph
- Promoter and insider activity trends
- Board quality and independence
- Key customer and supplier concentration
Entry timing
- Where the stock sits in its own price cycle
- Used for timing the entry, not for trading
- Trend alignment across multiple horizons
- Even good companies carry cycle timing risk
Technical analysis is usually associated with short-term trading. We use it for something narrower: timing. A fundamentally excellent company bought at the peak of its price cycle can still underperform for a year or more. The cycle position tells us where in the price cycle the stock is, not whether the company is good. The quality score handles that. Together they filter for companies that are both worth owning and worth buying at this point in time.
How a company gets in
Exclusion comes before scoring. A company is thrown out for reasons written down before we met it — things like a deteriorating top line, working capital drifting the wrong way, informed money leaving the register, solvency strain, or a price with no defensible value underneath it. Failing any one of them is disqualifying on its own. They do not net off against strengths elsewhere.
Surviving that is not the same as being interesting. A company is admitted to the researched set only if something positive actually fires: growth available cheaply, a sector or regulatory tailwind, or accumulation by investors with a track record. Most companies clear every exclusion and are admitted by nothing. That is the normal outcome, and it is reported as such rather than smoothed into a middling score.
What is in a deep dive
Around thirty sections. The score and every input that produced it; multi-year cash behaviour; the shareholder register diffed quarter by quarter with the price range each move could have happened in; sector-specific operating ratios against a cohort of listed peers with percentiles; adverse media screened against legal names rather than tickers; a behavioural review of our own decision; and a thesis ledger stating in advance what would falsify the position, scored against what has actually happened since.
Every number is computed from a dated snapshot of a primary filing. A language model writes the interpretation and never touches the arithmetic, the score, or the verdict. A model allowed to adjust a number will eventually adjust it toward the story, and there is no way to audit that after the fact.
Sections disappear when their input is missing. An absent number is never rendered as a zero or as a neutral, because a neutral is a claim and absence is not.
Next
Sector-wise specialisation. The same four dimensions, but with the financial layer read through the operating metrics that actually matter in each industry rather than a common set applied everywhere. A leading indicator in EPC is not a leading indicator in speciality chemicals, and the score gets sharper the more that distinction is built in.