Monitoring portfolio and watchlist companies against your thesis

Continuous monitoring for investment research is easy to start and hard to keep useful. Point a set of alerts at your portfolio and watchlist, and within weeks you have more headlines than anyone reads, with no clear link between any of them and the reason you hold or watch a company. The fix is to monitor the thesis rather than the company: write down what you believe, decide which observable signals would support or weaken it, grade the sources those signals come from, and keep a person between a detected change and any conclusion. This page is about research process. It is not investment advice, and your compliance team decides which sources and methods are permitted.

Write the thesis as claims you can check

A thesis that lives only in a memo cannot be monitored. Break it into assumptions, and for each assumption write what you would expect to see if it holds and what would suggest it is breaking. The weakening column is the one most often left blank, and it is the one that protects you.

Each row becomes a monitoring rule. A signal that does not map to any row is still worth a glance, but it should not be allowed to quietly reshape the thesis.

Portfolio and watchlist need different monitors

The same company can sit on both lists for different reasons, and the monitor should follow the reason.

• Portfolio companies are about early warning. You care most about weakening signals, governance events (an auditor change, late filings, executive departures) and anything that would trigger a review under your own process. Coverage matters more than precision here: a false alarm costs a few minutes, a missed signal costs far more.
• Watchlist companies are about entry conditions. You are waiting for specific evidence that the thesis has become true enough to act on, so rules are narrower and tied to the conditions you wrote down when you added the company.
• Sector context sits underneath both: regulation, peer events and supply chain news that affect several holdings at once. Monitor it once at the sector level rather than repeating it for every company.

For tracking funding rounds and deals across a sector, see funding and M&A tracking; for finding new names to add to a watchlist, see AI-assisted deal sourcing.

Grade sources before you trust them

A monitoring system treats every source as equal unless you tell it otherwise. Investment conclusions should not. Give each source a tier and store the tier on every signal it produces.

1. Primary and official: regulatory filings, registries, court records, official company announcements and documentation.
2. Original reporting: outlets that report first-hand and name their sources.
3. Secondary: aggregators, rewrites and newsletters that summarize others.
4. Unverified: forums, social posts, anonymous claims and rumour columns.

Rules follow from the tiers. A tier-four signal can prompt a question but cannot change a rating on its own. A conclusion should rest on tier one or two, or say plainly that it does not. Agree in advance with compliance which source types are off limits, including anything that might be material non-public information, and encode that list in the monitor instead of relying on memory. The note on deciding which sources may answer covers this habit in more depth.

Design each monitor from a thesis row

With assumptions and sources defined, a rule becomes concrete enough to review:

company    Example Components plc (portfolio)
assumption pricing power is holding
question   has list pricing, plan structure or discounting changed?
sources    company pricing page, annual report, investor presentations (tier 1)
material   price change, plan added or removed, new discount programme; ignore copy edits
cadence    weekly, Monday 07:00, Europe/London
notify     email the covering analyst; no automatic note to the investment committee

The cadence should match how fast the signal can move and how fast you could act on it. Weekly is plenty for pricing pages; filing deadlines may justify a daily check during reporting season. Running more often does not make a slow signal faster; it only produces more empty occurrences to review.

Keep a person between a signal and a conclusion

The biggest risk in automated monitoring is not a missed signal. It is a detected change that turns into a conclusion without anyone deciding that it should. A summary that says “pricing power weakening” because a single page showed a promotion is worse than no summary at all.

• A signal is not a finding. Record what changed, where and when, with a quotation. Interpretation is a separate step with a named author.
• Look for the counter-signal. Before writing an interpretation, check the weakening column for the same assumption. Confirmation bias is easier to resist when the form asks for it.
• Keep open questions open. If no source answers a question, record it as unanswered rather than filling the gap. The note on missing evidence as an open question explains why the distinction matters.
• Check citations, not just their presence. A link shows where a claim came from, not that it is right or relevant; see citations that hold up.

Keep a thesis log for each holding: an append-only list of date, signal, source tier, interpretation and effect on each assumption. Reviewed quarterly, it shows whether the thesis is being tested or merely decorated. When a signal justifies a full review, the steps in automated due diligence research take over.

Where GoCux fits

GoCux does not read the open web or supply market data. It is a workspace in which this kind of research stays reviewable. Pulse runs a recurring question over sources you select, with a cadence, timezone and definition of material change that are approved together as one exact version; each run is then reviewed change by change, and nothing acts on its own. For deeper questions, a research plan names the question, the ordered steps and a source allowlist, and it has to be approved before it runs; execution cannot reach beyond the approved plan, and an allowlist with no entries lets nothing in. Findings backed by a quotation are kept apart from questions where no source turned up. Filings and memos you import can be searched within a project, returning passages with their citations. The finance solution page shows how these pieces support evidence-led research.

Checklist for thesis-linked monitoring

• Each holding has a thesis broken into assumptions, with supporting and weakening signals.
• Portfolio monitors prioritize early warning; watchlist monitors track written entry conditions.
• Sector context is monitored once, not repeated for every company.
• Every source has a tier, and the tier travels with each signal.
• Off-limits source types are agreed with compliance and encoded in the rules.
• Cadence matches how fast the signal moves and how fast you could act.
• Signals and interpretations are separate records with named authors.
• Unanswered questions stay visible, and the thesis log is reviewed on a schedule.