How Harvst approaches research.
Research should make its position clear. This page sets out how Harvst arrives at a view, what it does with the evidence that argues against one, and what happens when a conclusion has to change.
The scenario running through this page is hypothetical. It describes no real company, no real market development and no research Harvst has published, and it contains no figures or dates drawn from actual events. It exists for one purpose: to show how Harvst's approach works from beginning to end, including the part where a stated conclusion turns out to need revising. Every block of the scenario is labelled where it appears.
From question to research.
Good research begins by defining the question clearly. Harvst starts with the development, uncertainty or change that matters, then narrows the research around what needs to be understood.
The situation
An industrial components manufacturer becomes unusually cheap relative to its own history.
Consider a listed manufacturer that supplies parts into several end markets. Its share price has fallen a long way over a year while its reported earnings have held up, and it now trades at a much lower multiple of those earnings than it has for most of the last decade. Commentary offers two ready-made explanations, and they point in opposite directions: the business is a bargain the market has overlooked, or the market has correctly worked out that the earnings are about to fall.
Both explanations are assertions. Neither is research.
Narrowing it
The question is not "is this cheap". It is "are these earnings durable".
A low multiple only matters in relation to what the earnings do next. So the useful question is a narrower one: are the current earnings a reasonable base for the next few years, or are they the peak of something that is about to reverse?
Framed that way, the research has a shape. It has to establish what is actually producing the earnings, whether that source is repeatable, and what would have to be true for it to stop. It also has an answer that can be wrong in a way a reader can check later, which the broader question does not.
How the evidence changes the picture.
Research should not begin with a conclusion and work backwards. Harvst considers the evidence that supports a view, the evidence that challenges it, and the context that can change what the evidence means.
Evidence for
The earnings look better supported than the share price implies.
The company has been winning work in a part of its market that is growing for reasons unrelated to the economic cycle, and that work carries longer contracts than the rest of the business. Management has been reinvesting rather than distributing, which is what a business expecting more of the same tends to do. The balance sheet carries little debt, so a weak year would be uncomfortable rather than dangerous.
Evidence against
A large share of the improvement traces back to one customer.
Reading the disclosure rather than the summary, a substantial part of the recent growth comes from a single large customer that is itself in an expansion phase. Margins have also improved faster than volumes, which usually means pricing rather than efficiency, and pricing is the first thing to go when a competitor decides to buy share.
This does not overturn the case. It does establish that the case depends on something narrower than it first appeared, and it identifies exactly where the research should be watching.
Context
The same evidence means different things in different conditions.
Customer concentration in a growing end market is a strength while that market is expanding and a serious vulnerability when it stops. Improved pricing is durable where a supplier is genuinely hard to replace and temporary where it is not. Neither piece of evidence carries a fixed meaning; each depends on conditions that can change without the company doing anything at all.
This is why the research states what the case rests on. A reader who knows the conclusion but not its dependencies has no way of telling when it has stopped being true.
From analysis to a view.
Research is most useful when the reader knows what it concludes. Harvst brings the analysis to a considered view and states that view plainly, including the uncertainty that matters to it.
The view
The discount is real, and it is not free.
The research concludes that the earnings are more durable than the share price implies, but that the market is not simply wrong: it is pricing a genuine concentration risk that the optimistic reading ignores. The conclusion is that the business is undervalued on the condition that its largest customer relationship holds, and that this condition, rather than the multiple, is the thing worth watching.
The uncertainty that matters
Stating what would make this wrong.
A view is only checkable if the reader knows what would falsify it. Here that is specific: a reduction in the largest customer's orders, a renegotiation of pricing on renewal, or a competitor establishing itself with that customer. Any of the three would remove the support the conclusion rests on, and would do so before it showed up in reported earnings.
Research that ends at "the shares look cheap" cannot be assessed later, because nothing was claimed precisely enough to be checked. That is not a stronger conclusion. It is an unaccountable one.
When the evidence changes.
New evidence can strengthen a view, weaken it or change it entirely. When that happens, Harvst can revise its conclusion without rewriting the one that came before it.
The development
The condition the view rested on is tested directly.
Some months later the largest customer announces that it is bringing part of this work in-house over the next two years. It is a smaller share than the whole relationship, and the transition is slow, but it lands precisely on the dependency the research identified.
The reassessment
What changed, and what did not.
The durability of the earnings outside this relationship is unaffected, and the balance sheet is unchanged. What has changed is the shape of the case: a business whose growth was concentrated in one expanding relationship is now a business managing a known, scheduled reduction in it. The discount that looked like an overreaction is better described as a reasonable price for a slower business.
Because the original research stated the condition explicitly, this is a reassessment rather than a rescue. There is no need to reinterpret what was written; the trigger was named in advance, and it happened.
The business is undervalued, on the condition that its largest customer relationship holds.
Published with the dependency stated, and with the three developments that would invalidate it named.
The condition has failed. The discount is now a reasonable price for a slower business, not an opportunity.
Revised on the announcement, for the reason the original research said would matter. The earlier view remains published, unaltered.
We can change our view. We don't change our history.
Explore Harvst research
Macro & Markets
Research on economic, policy and market developments, and the conditions affecting markets, sectors and companies.
Sectors & Themes
Research on industries and broader developments affecting businesses and markets.
Companies
Research focused on individual businesses and the developments affecting them.
What Harvst is
Harvst is an independent research publication. It is not an investment firm and not a fund. It does not manage money, holds no client funds, and accepts no compensation from any company it covers.
Nothing published here is advice, a recommendation, or a signal to act on.