Implement a multidisciplinary approach in the building of analysis models to ensure selected securities are attractive based upon a broad array of metrics, and not unduly influenced by any one analysis method. Three unrelated disciplines, asked the same question about the same company. Where they agree is worth something. Where they disagree is worth more.
A single method, applied with enough conviction, will find what it is built to find.
Every analytical method has a shape, and that shape decides what it can see. A valuation screen finds cheap companies, including the ones that are cheap for excellent reasons. A quality model finds excellent businesses, including the ones whose excellence is fully in the price. A macro model finds markets that have run ahead of their economies, and says nothing at all about whether a particular company deserved to.
So every company in the benchmark is put through three unrelated disciplines, and none is allowed to dominate.
The market and sector work, brought down to the individual company. A model of the share price against the same macro factors, estimated over a long window and then expressed as the percentage difference between model price and actual price, adjusted for goodness of fit exactly as the market and sector models are.
Four windows are estimated rather than one, from one year to ten. A company that looks expensive on every horizon is a different case from one that looks expensive on the shortest window alone, and collapsing that into a single number would hide it.
The ten ratios from the previous chapter, used here for what ranking set aside: the relative magnitudes. Read across five categories, they describe a company's financial character rather than filtering it.
This is where the discarded distance comes back, and it is exceedingly useful.
Four models built from the accounts, each answering a question the other two disciplines cannot: is this business predictable, does management earn more than its capital costs, is the price sensible against the growth, and is the competitive position holding.
Set out in full below.
In the days before regulation levelled the playing field, informational asymmetry meant a premium on relationships with corporate officers, and the challenge was obtaining material information before anyone else.
The asymmetry has not gone away. It has inverted. All material information is now disseminated instantly and universally, and the challenge is managing the overload better and faster than everyone else looking at the same screen.
The first two disciplines are both directed at that. They determine which information has relevance and how much; they collect and manipulate it uniformly; and above all they free up time for the heavy accounting forensics, which is the part no model does for you.
Eight years of trend revenue and EBITDA growth, mapped against the actual annual numbers. The smaller the deviation from trend, the more predictable the company.
The window is chosen to capture a full business cycle. Any company posting an operating loss in any year of the window is deemed unpredictable, without argument. Companies are then ranked on predictability, and that ranking is one of the four inputs.
One of the most basic analyses available and among the most useful: how well management uses the capital in the firm. Return on invested capital divided by the weighted average cost of capital.
The question it asks is simply: for every unit of currency financed, how many units are generated? A company with a cost of capital of 4% earning 20% on invested capital produces five units for each one financed, and an efficiency factor of five.
Mercia calculates the cost of capital itself rather than accepting a vendor figure, because a ratio is only as sound as its denominator.
There is no shortage of valuation measures, and several were already used in the screens. What none of them does is make provision for growth.
So the measure here is the price to earnings ratio against the eight year average EBITDA growth rate. A reading of 1 is precisely fair: growth and multiple are the same number. Below 1 is undervalued; 1 to 2 is fair; above 2 is expensive.
Using realised growth rather than forecast growth is the whole point. Forecasts are an opinion about the future. This is an observation about what the business has actually done.
This has its roots in the notion of an economic moat, as propounded by Warren Buffett. If a company can maintain, and preferably expand, its margins over time, it holds a strong position in its market.
It is a simpler measure than Porter's five forces, and no less effective for it. Take the profit margins over time and assess whether they are expanding, consistent, or contracting. A company that cannot hold its margin is telling you about its competitive position more reliably than any strategy document will.
The part that is not a model, and should not pretend to be.
Almost everything above follows given rules. But companies are as unique as people, and some analysis has to be specific to the company in front of you. A filtered, focused list of quality ideas is what creates the time to do it.
As a rough guide, the things worth looking for:
Much of this is intuitive, and genuinely one of those ten thousand hours things: spend long enough looking for something and you develop a sense for when it is there. Every company plays the numbers game to some degree. What matters is telling ordinary quarter-end management from the thing that actually changes the investment case.
Mercia runs a forensic panel across the universe to surface the quantifiable half of that list automatically, and reports it under a deliberately high bar: a flag is raised only where it would change a decision. A company with nothing to answer for gets a short note saying so, rather than a page of padding.
Equal thirds, and the reason it is not 40/30/30.
The three disciplines produce three percentile rankings against the whole universe, and the composite is an equal third of each.
The 40/30/30 rule that opened the first chapter describes where share price movement originates, and it is why the framework builds outward from the market. It is not a recipe for weighting three analytical methods, and using it as one would double count: the market and the sector have already spoken, at full strength, in the allocation matrix. Letting them speak again inside the stock score would say the same thing twice and call it corroboration.
Where a company is too young to carry a ten year econometric history, its econometric leg scores zero rather than being quietly removed. Renormalising over the two remaining legs would promote a company for the absence of evidence, which is the opposite of what the absence means.
MSFT · United States · Information Technology. Three unrelated disciplines, each ranking this company against every other company in the benchmark.
| The four fundamental models | Reading |
|---|---|
| Predictability deviation from eight year trend |
0.05 |
| Management efficiency ROIC 27.1% against WACC 10.6% |
2.56x |
| Valuation against growth P/E 27.3 against 20.1% growth |
1.36 fair |
| Margin trend 8 year slope |
expanding |
The point of three. The three disciplines disagree often, and that is the design. A company can be cheap against the macro and mediocre on the accounting, or excellent on the accounting and priced for it. Reading them separately, and then together, is the whole exercise. This is a fraction of one company's page; the client area carries nine such panels for every name in the benchmark, each with written analysis of what the numbers are actually saying.
Written analysis, as it appears on the page. Read the decade window first: with 120 observations and an R² of 0.94, the long-run macro relationship for Microsoft is unusually well specified, and it places fair value at 440 against 493. That is the number to anchor on.
Why this company: it sits outside the market used in chapter one and outside the sector used in chapter two, so that the three worked examples across this section cannot be assembled into an allocation. The selection rule is fixed and runs with the models each month.
The full company pageThe models are the beginning of the work, not the end of it.
Everything described in these four chapters is computed for every company in the benchmark, every month, and published on a page for each name: the composite and its three legs, the econometric model across four horizons, the ten screens, the four fundamental models, the forensic panel, insider dealing, consensus, and the filings themselves.
Alongside every one of them is written analysis. Not a description of the chart above it, which the reader can already see, but an explanation of the tension in the numbers: where two models disagree, which one the evidence favours, and what would have to change for the conclusion to change. That is the part that takes judgement, and it is the part the models exist to make possible.