Actionable insights — Full Investment Case: KKR

1. Score every candidate on the same fifteen metrics, and publish the worst one

1. Fix the metric list in advance — business model, management, moat, industry, risks, balance sheet, capital intensity, capital allocation, profitability, dilution, historical growth, outlook, valuation, owner's earnings, value creation.
2. Score each out of ten with a one-line justification, and take the simple average. No weighting, so a single bad score cannot be diluted away by adding good ones.
3. Read the lowest score first — it is the shape of the risk you are underwriting.
4. State whether you are accepting that weakness, and why. If the reason is "everyone does it," say so explicitly so it can be challenged.

• An unweighted average hiding a veto-level flaw: 4/10 on dilution costs only 0.4 points of the total.
• "Industry practice" as a defence. It explains why the number is high; it does not mean the shareholder is not paying it.

2. Value a multi-part financial by its parts, not by a group multiple

1. Split the business into segments that earn differently — here fee income, insurance, and balance-sheet holdings.
2. Value each on the measure appropriate to it (a multiple of fee-related earnings, a book-value or spread-based measure for insurance, a mark for the holdings).
3. Sum them, subtract corporate debt, and express the result as a per-share fair value.
4. Cross-check with the group's own multiple history — agreement between two unrelated methods is the point.
5. Say which segment the discount is concentrated in, because that is what has to re-rate.

• Sum-of-the-parts as a permanent explanation for a permanent discount. A conglomerate discount that never closes is a valuation, not an error.
• Which segment's multiple is doing the lifting; insurance and fee streams should not be capitalised at the same rate.

3. Ask where the capital comes from and whether it can be withdrawn

1. For any manager or financial, separate the capital that has a maturity from the capital that does not.
2. Permanent or long-dated capital changes behaviour: it removes the obligation to invest and exit on a calendar.
3. Quantify it, as a share of total assets, and track whether it is growing.
4. Check the cost: insurance float is only free if the underwriting roughly breaks even — the spread earned versus paid is the number to find.

• Permanent capital that carries a guaranteed crediting rate — annuity liabilities are cheap funding only while the spread holds.
• Regulatory capital requirements at the insurer limiting how much of that money can actually be moved into illiquid assets.

4. Read consolidated leverage down to who can be sued

1. Take the headline debt figure and split it into recourse (the parent must repay) and non-recourse (only the fund or asset backs it).
2. Non-recourse debt inside consolidated funds is not the manager's obligation; exclude it from the parent's leverage.
3. Then look at direct corporate debt against cash and fee-related earnings — that is the real balance sheet.
4. Note what remains at risk anyway: reputation, seed capital, and the fee stream if the funds fail.

• Reputational recourse: a manager that lets a fund fail may face no legal claim and still lose the next fundraise.
• Ratios that become meaningless on consolidation — the onepager's ROIC of 0.2% and FCF yield of 0.4% are artefacts, not findings.

5. Require named, countable growth engines rather than a market forecast

1. Ask which specific new pool of customers or capital produces the forecast growth, and how big it is today.
2. Prefer engines already visibly working to those still described as opportunities.
3. Check management's target against the analyst estimate; a gap either way is information.
4. Set a checkpoint: what number, in which quarter, would confirm the engine is still running?

• Retail distribution as a growth engine that also changes the liability: individual investors redeem in ways institutions cannot.
• Acquisitions counted as AUM growth. Assets bought are not assets raised, and they are not equally profitable.

6. Publish the case before the trade, and the conclusion for those who will not read it

1. Write the full case first, then a one-page conclusion that stands alone, then the transaction.
2. The conclusion must contain the decision, the price, the size and the limit — everything needed to act or to disagree.
3. Set the limit deliberately relative to the quote and note which you are optimising for: certainty of fill, or price.
4. Record how long the name sat on the shortlist before being bought.

• A conclusion written after the decision was made; the tell is a case with no disconfirming section.
• Subscriber front-running: an alert published before the manager's own execution moves thin names — less of an issue here, at $595m average daily volume.