Header · Squarespace header, theme White · logo: Evergreen with white fill · nav folders: Services, Insights, About, Clients · header button (native) links to the portfolio review form (https://zfrmz.ca/r3nj67YWFfJs6AhVsK7h) · Clients folder adds Client login (NBIN), Q8 · cart icon removed · mobile: hamburger, folders expand in place, button at the bottom
Book a portfolio review
Article · a blog post in the Articles collection · title, author, date and tags from the post settings · the byline portrait is an image block · URL /articles/july-2026-revisiting-the-first-decision-in-our-research-process · SEO title “Revisiting the First Decision in Our Research Process | Qube Commentary, July 2026” · meta description from the conversions document

Qube Commentary · July 2026

Revisiting the First Decision in Our Research Process

Body · post body in two columns (the editor’s side-by-side columns): text 8 of 12, the issue panel 4 of 12, stacked on mobile · H2 and H3 as published, Roman numerals set as numbers · tables rebuilt as HTML in a Code Block · text as published, light editing only

In brief

Noah Clarke on how Qube rebuilt the universe filter, the first decision in our research process, after testing it on fifteen years of data.

A first glance through a restaurant window. A quick scan of a book’s inside flap. These small filters decide where the evening goes, or what we read for the next two weeks, long before any considered choice is made. Our research process at Qube has a similar decision at its beginning. Before any analyst opens a financial statement, before Karlen and Patrick review any of their subsequent work, before our Portfolio Steering Committee considers a new position, we have to decide which companies are worth looking at in the first place. The list of publicly traded companies that we could entertain exceeds 10,000, but our research team will go through somewhere in the range of 300 valuations during the year (some passed on early, some making it through deeper dives). The gap between those two numbers is filled by a screen we call the universe filter. It is the least visible part of our investment process but can be quite consequential. This article is about a research project completed earlier this year, looking at how that filter has been performing.

The Original Idea

Our universe filter has been part of the research process for many years, with its specific criteria evolving over time as our thinking has evolved. The version we reviewed has been in place since 2017, when we made a substantive change to how we screen the global stock universe. Some readers may recall this commentary’s spring issue from that year, which described the change at the time. The decision was made against the backdrop of a market that was trading at elevated levels relative to historical norms, and our judgement was that a more rigorous quality screen would help us identify durable businesses that might otherwise be overlooked in a broadly expensive environment.

The academic foundation for the change came from a paper published in 2000 by Joseph Piotroski, then an accounting professor at the University of Chicago.1 Piotroski proposed a ninepoint scoring system, built entirely from a company’s published financial statements, that turned out to be useful for separating high-quality businesses from low-quality ones. The score has since become one of the most widely studied quality measures in academic finance, cited in hundreds of subsequent papers and adopted in various forms by practitioners around the world.

Figure to rebuild DIAGRAM: The Piotroski Framework, 9-point quality test: weak (0-3), stable (4-6), strong (7-9). Original PDF, page 11.

The Piotroski framework grouped its nine signals into three categories. The first was profitability: was the company earning a return on its assets, was its cash flow healthy, and were those measures improving year over year. The second was leverage and liquidity: was the company managing its debt load responsibly, and was it avoiding the temptation to issue new shares to fund itself. The third was operating efficiency: were margins improving, and was the company getting more revenue out of its asset base. Each company was awarded a point for passing each individual test, and the resulting score, on a scale of zero to nine, served as a quick proxy for fundamental quality.

Our own implementation in 2017 began as a fairly close adaptation of this framework, with a few additional screens (for instance, a minimum market capitalization) layered on top. Over the years that followed, the filter was refined. The original nine Piotroski signals were consolidated into five factors that suited our screening workflow more cleanly, and a dividend factor was added on the view that a sustainable dividend was itself a signal of corporate health. The five factors in the version we examined this year were these: companies that pay reliable dividends, companies that are reducing their debt, companies whose return on assets is rising, companies whose gross margins are expanding, and companies that are not diluting their existing shareholders by issuing new stock.

Each of these is a real signal of corporate quality, and a portfolio assembled with attention to all five would not be a foolish portfolio. For most of the historical period we examined, the filter performed its job competently. That said, there is always room for improvement.

The Project

The advantage of having fifteen years of data, twenty-four countries of coverage, and roughly twenty-four thousand annual observations of companies in the global stock universe is that you can put a model through a fairly serious examination. For each year from 2010 to the end of 2025, we asked which companies the filter would have selected, using only the financial information that would have been available at the time. We then tracked how those selections performed against the broader market over the following year. We repeated the exercise across the full fifteen-year window.

The pattern that emerged was instructive. In its early years, the filter generated a modest positive edge. In more recent years, that edge has flattened, and in some years inverted. Across the full period, the filter’s selections produced returns that were close to neutral relative to the broader market. To put it more plainly: the filter, judged purely on its own selections, was no longer adding the kind of edge it had once added.

This finding deserves careful framing. It does not mean the filter has caused harm. Our portfolios have outperformed our benchmark over the long run, because the rest of the process (the phase reviews, the analyst work, the Steering Committee oversight) has done a great deal of compensating work. The filter has been one input among many, and the system as a whole has held together. What the project told us was something more specific. The first input in the chain was no longer pulling its weight and was shaping the work of everyone downstream in ways we had not fully appreciated.

The deeper we looked, the clearer the pattern; the filter had developed a pronounced sector tilt over time, passing through companies in Consumer Staples at a much higher rate than companies in Health Care or Technology, structurally, across the fifteen-year window.

Some of this was traceable to the dividend factor. The most successful growth companies of the past two decades have, almost as a rule, declined to pay dividends. Their managers have judged, correctly in most cases, that the cash was better deployed inside the business than returned to shareholders. However, the filter treated the absence of a dividend as a strike against the company, regardless of how thoughtfully that decision had been made. Therefore, a company would need to excel on all other metrics if it didn’t pay a dividend — limiting the number of non-dividend payers making it through our filter.

The rest was traceable to the kinds of businesses that look attractive under a checklist of incremental improvements. A company like Colgate or Procter & Gamble can reliably trim a little debt each year, raise its return on assets by a hair, expand its gross margin by a few basis points, and buy back a slice of its shares. These are the hallmarks of well-run mature businesses, and they show up beautifully on a quality screen, more so than others that might have had higher upside. The same pattern shows up in parts of Utilities and Real Estate, for similar reasons. A filter that subtly favours some sectors and avoids others ends up doing the same for individual companies. There is a sports analogy that applies here, though it has become a bit of a cliché. In the spring of 2000, the New England Patriots used the 199th selection of the NFL draft on a backup quarterback named Tom Brady. By the standards of every scouting model in use at the time, he was the 199th most promising player available. Six Super Bowl rings later, every team in football has rewritten its evaluation models.

The lesson is not that the old models were bad, but that even reasonable models can miss the most important cases and when they do, the right response is to study what they missed and update accordingly.

Two Ideas That Go Together Like PB & J

The new filter combines two ideas, one focused on valuation, the other focused on quality.

The valuation idea is the more important of the two. It asks, of each company in the universe, a question that takes a moment to explain but is intuitive once stated. How cheap is this company relative to its own typical valuation, after adjusting for whether the overall market happens to be cheap or expensive at this moment? The adjustment matters more than it might first appear. If interest rates rise sharply and every stock in the world becomes cheaper, that is not a signal about any individual company, but rather a signal about the broader market environment. The filter should be looking for companies that are cheap for reasons specific to themselves, and not for reasons specific to broader market conditions.

A small concrete example may help. Visa has traded, over the past decade, at an average of roughly fifteen times revenue, occasionally reaching eighteen, occasionally dipping to twelve. If Visa is at fifteen today, it is at its own normal price. If it is at twelve, it is meaningfully cheap by its own standards. If it is at twentytwo, it is meaningfully expensive by its own standards, even though a company at twentytwo times revenue might be a bargain in some other industry. The point is to compare each company to itself, in its own context, rather than to apply a single standard across the entire universe.

The quality idea, given a third of the total weight, retains four of the original five factors: debt reduction, improving return on assets, gross margin expansion, and stable or shrinking share count. The dividend factor came out.

Quality remains a useful check against the kind of company that looks cheap because it deserves to be cheap.

We also continue to exclude most Energy and Basic Materials sector companies from the universe before the ranking step. The reasoning is that, in our view, the valuations of commodity businesses are largely driven by commodity prices rather than by company-specific fundamentals, which makes the cheapness signal in those sectors quite noisy.

Tested against the same fifteen years of data, the new filter produced a positive median annual edge of roughly six and a half percent and was positive in every one of the years in the comparable period. Sectors that the old filter had systematically discounted now appear with much greater frequency. For instance, some companies which seemed to never make it through the filter now passed much more frequently. Microsoft passed in seven of the fifteen years. ASML and Mastercard passed in all fifteen.

What the Filter Does Not Do

There is a temptation, when a model produces good results, to assume it is doing more work than it really is. I want to push gently against that temptation, because the most important lesson from this work is not about the filter at all. It is about where the filter sits in our process.

The universe filter produces a list of roughly three hundred companies, and from that list, the actual research begins. I don’t mean to belabour the point, as I know Karlen’s article touches on what follows, but our universe filter is best explained within the context of our overall research program.

The analyst is assigned a company from the research universe and begins building the initial valuation. They model the business, identify what we call the narrative (the story that explains why the company exists and how it makes money), and surface any early concerns. That first pass is then reviewed by one of our portfolio managers, Patrick Choi or Karlen Daschuk, who provides direction on what comes next. A significant number of companies are kicked out at this point. The screen may have viewed them favourably, but on closer inspection we often find issues that no quantitative measure could have caught.

For companies that pass the first phase, the analyst is then tasked to take a deeper dive. They build the investment case in full, or, just as often, the case against. They review filings, model out a range of scenarios, address every question and point of contention posed by the supervising portfolio manager, and stress-test the assumptions that drove the initial valuation. By the time a company has cleared this second review, it has typically had thirty to fifty hours of analyst attention invested in it.

What follows is the final stage of analyst work: a complete recommendation, supported by a full valuation, a written narrative, and an explicit assessment of risks. This is reviewed once more, usually by Patrick. Further revisions are often requested, and our analysts get very used to seeing red font. If the recommendation survives all of this, it is presented to the Portfolio Steering Committee.

The Committee is made up of the firm’s advising representatives, myself included. We decide together whether the company is suitable for entry into our Kaleo portfolios. Not infrequently, the Committee declines a recommendation that both the analyst and the reviewing portfolio manager supported. Sometimes the position has less upside than current holdings. Sometimes it is too illiquid for the size of our book. Sometimes the timing is wrong. Whatever the reason, the final decision rests with the Committee.

The way decisions move through our process is, in itself, part of the discipline. Our analysts work through a defined process, with portfolio managers reviewing the work at each stage, and the Steering Committee making the final call on every position. No single input, including the universe filter, is asked to carry more weight than it should. The phase reviews are where we expect to catch what a quantitative screen cannot see. The Steering Committee discussions are where we expect to catch what the phase reviews might miss. Each step adds a layer of judgement that the previous step did not provide.

The new filter simply improves the starting point of that chain. It does not change the work that happens after the starting point, nor should it. The screen produces a list of candidates. The candidates are evaluated with models and human judgement. The decisions about portfolios are made by a committee whose responsibility for client outcomes is unambiguous.

A Subtle Improvement

I have tried not to oversell what we did. We replaced one screening tool with an improved version. We tested it carefully against fifteen years of historical data and updated the first step in a research process whose other steps have served us well for many years. The filter that came in is, as best we can tell, better suited to the market we find ourselves in now.

The deeper lesson here is not really about filters. As with the companies we hold in our portfolio, we must return, every so often, to the first decisions we make and ask whether the underlying assumptions still hold.

We now have a new piece of code saved on our server. Every January, I will press a button, and a list of about three hundred companies will be compiled. The analysts will get the list shortly after that, in a spreadsheet, and a new batch of research will begin. Almost nothing about that scene will look any different from how it looked last January. The work that produced the change will be invisible inside it. This is, I think, what most subtle improvements look like.

Notes

1 Piotroski, J. D. (2000). Value Investing: The Use of Historical Financial Statement Information to Separate Winners from Losers. Journal of Accounting Research, Vol. 38, Supplement, pp. 1–41.

Disclaimer. The content of this report is for general information purposes only and not intended to provide specific personalized advice, including, without limitation, investment, financial, accounting or tax advice. Please contact Qube Investment Management Inc. to discuss your particular circumstances. Commissions, management fees and expenses may be associated with investment accounts. Please read the investment management agreement before investing. Many investments are not guaranteed and are not covered by the Canada Deposit Insurance Corporation or by any other government issuer. There can be no assurances that an investment will be able to maintain its net asset value or that the full amount of the investment will be returned to you. Values change frequently and past performance may not be repeated. Qube Investment Management Inc. is a registered portfolio management firm in the Provinces of Alberta, British Columbia, and Ontario and was registered as a portfolio management firm on June 25, 2012. Any return period cited before this date was prior to QIM being registered as a portfolio management firm. Inception was Jan 1, 2011 and all returns are for a modeled portfolio initiated at $500,000. Your actual returns may vary according to your individual portfolio. The modeled returns are calculated inclusive of dividends, adjusted to the Canadian currency, and are determined via the IRR (Internal Rate of Return) method. The gain/loss shown are simple (non-compounded) returns for periods up to one year. If the time since inception date is more than one year, then the return shown is an annualized return. For comparison purposes, the Kaleo model(s) are reported as gross returns before investment management fees. Individual investor level returns will differ as the fees agreed to in your Investment Management Agreement (IMA) are subtracted from the gross return. At any one point in time, the composition of the Kaleo model may change. Currently, the focus for our models (Kaleo A and Full) is to invest in a globally diversified portfolio of liquid stocks with a minimum market capitalization of $1 billion. Our diversification strategy is to have similar industry weightings between our Kaleo models A and Full, which in turn will have similar weightings to the MSCI World Index. Our investment mandate is to not have any one industry sector or sub-group exceed 2.0 times the percentage weighting assigned to that group by the MSCI Index unless the sector or sub-group composes less than 5% of the total index. Please refer to your Investment Policy Statement (IPS) for more details. Index comparisons are based on the total return index defined by either: 70% of the MSCI Index and 30% of the S&P TSX Total Return Index; or 50% of the MSCI Index and 50% of the S&P TSX Total Return Index. All index returns are inclusive of dividends, adjusted to the Canadian currency, and, similar to the modeled portfolio, determined via the IRR method. Please note that, as total return indices are not actual portfolios, these returns do not include the cost of management and/or trading fees. Past performance is not indicative of future results and there is no assurance that our model portfolio will achieve its objectives or avoid significant losses.

More from this issue · Summary block on the Articles collection, filtered by the issue tag, three items · link to the Archive

More from the July 2026 issue

Letter From the Editors: An Examination of Process

By Noah Clarke, MA; Karlen Daschuk, CFA, CFP®; & Michael Baker, MBA, CFA, CFP®

Noah Clarke, Karlen Daschuk and Michael Baker on AI, geopolitics and why process matters most, introducing the July 2026 Qube Commentary.

Read

The Qube Code: Guiding Principles

By Michael Baker, MBA, CFA, CFP®

Michael Baker sets out the Qube Code, the six principles that guide how Qube invests and advises, through the lens of Batman.

Read

Eleven Bowls Poured

By Karlen Daschuk, CFA, CFP®

Karlen Daschuk on what a decade of Qube’s in-house research records reveals about the craft of picking stocks.

Read
Comp All pages