Open research · Conviction Core
Every name in Conviction Core has a written thesis, published before the trade. This page explains what gets written, when it gets published, and what makes a position get sold.
The ordering
Research is easy to grade generously after the fact. A note written once the price has already moved will always look prescient, because the person writing it already knows the answer.
So Conviction Core inverts it. The reasoning for a name is written and published, timestamped, before any order is proposed — not after the fill, not at the end of the quarter. That ordering is the whole integrity claim: it removes the option to edit the reasoning once the price has spoken.
What follows is the member's decision. An order is proposed to them with the thesis attached; they approve it or they do not. Approved orders are then placed through their own connected broker account. Money is never pooled, and nothing is placed without their approval.
The order never runs ahead of the reasoning. Step 1 always precedes step 4.
The funnel
A book of this kind is defined at least as much by what it refuses as by what it holds. The screen is deliberately unkind, and the common outcome of a deep dive is not a position.
The first read is forensic, not narrative. Cash conversion, receivable and inventory days, related-party flows, auditor changes, promoter pledging, contingent liabilities. A good story attached to bad accounts is still bad accounts, and it stops there.
Small and micro caps are screened on traded volume and impact cost. If a position could not be exited at a sensible price, it is resized or refused. You cannot sell what has no bid.
Names that fail do not quietly disappear. The reason a name was refused is written up and kept in the same internal record as the names that were bought, so when the company comes round again the work starts from what was already concluded rather than from a blank page.
The document
Not a target price and a paragraph of enthusiasm. A thesis is a structured document with the awkward parts made explicit, because those are the parts that decide the outcome.
Debt and its maturity ladder, working-capital intensity, capital employed and what it earns, promoter pledging, off-balance-sheet and contingent items. Read over several years, not one.
Whether reported profit turns into cash, and where it leaks if it does not. Operating cash flow against profit across the cycle, capex against depreciation, and what the growth is actually being funded by.
Related-party transactions, auditor tenure and resignations, board independence in practice, related lending, unexplained subsidiaries, and the gap between what management said last year and what happened.
Written in full, as the strongest version of itself — not a token paragraph. What has to go wrong, how it would first show up in the numbers, and how much of the price already assumes it does not.
The specific, checkable conditions that would mean the thesis is wrong, written down before the position exists — so the exit rule is not invented later, under pressure, by someone holding a loss.
Targets, not promises. The book sits away from them whenever the evidence does not support closing the gap.
The exit rule
A position is held until the reason for owning it stops being true. That moment has a name — a thesis break — and it is defined in writing before the position is opened, not after it starts hurting.
Break The numbers stop supporting it
Cash conversion deteriorates, the return on capital that justified the price erodes, or the growth turns out to be funded by the balance sheet rather than the business.
Break Governance turns
An auditor resignation, a jump in pledging, related-party activity that was not there before, or a disclosure that contradicts what management said. This is treated as terminal, not as a discount.
Break A named trigger fires
One of the invalidation conditions written into the original thesis actually occurs. It was defined in advance precisely so it cannot be argued away in the moment.
Break Liquidity fails
Traded volume falls to a level where the position can no longer be exited sensibly. The risk changes even if the business has not.
Churn is a cost. So is holding something whose reason has quietly expired. Both are decided against the written thesis, not against the last price.
The thesis is re-read against the new facts and the outcome is written down. If it is a break, an exit order is proposed to the member with that reasoning attached — and, exactly as on the way in, it is placed only once the member approves it, through their own broker account.
Where things stand
A record is only worth reading if it includes what is inconvenient. Two things are true about Conviction Core right now, and neither is hidden.
Deliberately under-deployed
The book is a long way from its own cash target, and that is a decision rather than an oversight. Capital is deployed when a name clears the gates and not because a target weight says it should be. Held cash is a position, and it is explained rather than dressed up.
Cash weight and position count as at the latest published book snapshot. Bucket weights for micro, small and mid currently sit below their targets as a direct consequence.
No track record
Conviction Core is young. The performance series behind it covers too short a window to say anything meaningful, so it is not charted, not annualised, and not presented as a track record — because it is not one. When there is a run long enough to be worth judging, it will be shown in full, including the bad parts.
What can be judged today is the process: what a thesis has to contain, when it gets published relative to the order, and the written rules that decide a sale. That is what this page is for.
Access
Conviction Core is invite-only. Entry is a single-use code, valid 30 days, and begins only after separate documentation is signed. Members keep their own broker account, money is never pooled, and every order is approved by the member before it is placed.
Theses are published to members before the order is proposed, and are read inside the portal after signing in. There is no public archive of them on this site.