Actionable insights — Eli Lilly deep dive

1. Split CAPEX into maintenance and growth before calling a business capital-intensive

1. Run the crude test first and record the result honestly: CAPEX as a percentage of sales (threshold 5%) and of operating cash flow (threshold 25%).
2. If it fails, do not stop. Ask whether the spending is replacing worn-out assets or building new ones — the first is a cost of staying in business, the second is an investment decision.
3. Use depreciation and amortisation as the proxy for maintenance CAPEX. Growth CAPEX is then total CAPEX minus D&A.
4. Re-run both ratios on maintenance CAPEX alone. If the business now passes comfortably, the headline number was measuring ambition, not fragility.
5. Then verify the growth spending is earning its keep — ROIC should hold up while the expansion is under way, or the growth CAPEX is destroying value.

• D&A understating true maintenance needs at a company with old assets bought cheaply, or overstating it after a big acquisition loaded with amortisation.
• "Growth" CAPEX that is really catch-up spending after years of underinvestment.

2. Run three valuations that can fail independently, and read the disagreement

1. Method one — compare the forward multiple with the company's own ten-year average. This catches sentiment shifts but says nothing about whether the historic multiple was ever justified.
2. Method two — an earnings growth model: expected EPS growth + dividend yield + the annualised effect of the multiple moving to a stated exit level. This produces an expected return you can argue with input by input.
3. Method three — a reverse DCF: hold the required return at 10% and solve for the growth the current price implies.
4. Record all three verdicts separately. Two of three failing is a pass on the stock; three of three passing is rare and is the signal to act.
5. When they disagree, name the disagreement rather than averaging it — which input is doing the work is the actual finding.

• An exit multiple chosen to produce the desired return; set it before you calculate, not after.
• A ten-year average multiple struck during an exceptional period — for a pharma company, one blockbuster cycle can define the whole history.

3. Judge the implied growth rate against your own forecast, never the consensus

1. Extract the growth rate implied by the current price from the reverse DCF.
2. Write down your own long-term estimate before looking at it, with the reasoning behind it.
3. Note the consensus separately as a third number. Three figures, three sources, kept apart.
4. If the implied rate exceeds your own estimate, you have no margin of safety regardless of what the consensus says — the consensus is not what you are betting.
5. Record the price at which the implied rate would fall to or below your own estimate; that is your buy level, and it is a number rather than a feeling.

• Anchoring your own estimate to the consensus after you have seen it — write yours first.
• Using net income as the FCF proxy (as done here, because of heavy growth investment) without saying so; it changes the implied rate materially.

4. Publish the pass, with the score attached

1. Complete the full analysis even when the answer is going to be no — the work is the asset, not the purchase.
2. Score every dimension separately so the reason for the pass is isolated to specific line items rather than an overall impression.
3. State the verdict in one sentence that separates business from price.
4. File the name on a follow list with the specific condition that would change the answer, so it can be revisited mechanically rather than remembered.

• A follow list with no trigger price — "wait patiently" is only actionable with a number behind it.
• Quality scores drifting up over time as familiarity grows; re-score from the data, not from memory.

5. When the theme is right and the best name is priced for perfection, look across the duopoly

1. Establish that the end-market growth is real and sized — do that work before choosing the vehicle.
2. List the small number of companies that will capture it. In a genuine oligopoly this is usually two or three names.
3. Score them on quality and on price separately. The best business and the best investment are frequently not the same name.
4. Prefer the participant whose price implies a growth rate you actually believe, provided its position in the duopoly is secure.
5. Be explicit that you are trading some business quality for margin of safety, so the trade-off can be revisited if the discount closes.

• The cheaper participant being cheap for a reason — check its own growth forecast (Novo's EPS is modelled flat to 2028 in the August update).
• Duopolies that turn into a single winner; a share shift makes the discount permanent rather than temporary.

6. Date the patent cliff, and treat it as a known scheduled event rather than a risk

1. For any IP-protected business, list the products that matter and the year each loses exclusivity.
2. Estimate the revenue at risk in each year rather than in aggregate — the shape matters more than the total.
3. Check what is scheduled to replace it and at what stage of approval, since the replacement's timing is the real variable.
4. Treat approval of the replacement as a binary event with a date, and decide in advance what you do in each outcome.

• Reformulations and dosage patents extending exclusivity beyond the headline date, which cuts both ways.
• Government price pressure arriving at the same moment as the cliff, compounding it.