Actionable insights — Friday POW!: Incyte (INCY)

1. Read the consensus price target as a disclosure of what the market refuses to pay for

1. Pull the average analyst target and set it beside the current price. Treat the gap not as a forecast of return but as a statement of what is inside the consensus model.
2. When the average target sits essentially at the market price, the model contains no catalyst credit — the sell side has priced the business it can forecast and left the contested asset at zero.
3. Reconstruct that base case explicitly: name the products and revenue lines the target does capture, and confirm they alone justify the current multiple. If they don't, this method does not apply — you are simply paying up.
4. Then find the dispersion. Split the analyst list into the ones near the average and the ones well above it, and identify what the high group is adding. A cluster at a materially higher number, all citing the same asset, is the price of the option quoted by the market.
5. Frame the trade as the difference between the two groups: you are buying the base case at the base-case price and receiving the disputed asset for the spread — which is near zero when the average target equals the price.
6. Sanity-check the direction of the error: a consensus that is too low because it excludes an asset is recoverable; one that is too low because the base business is deteriorating is not.

• The average target starting to move up toward the bull cluster — that is the option being priced in, and the point at which the asymmetry that justified the purchase is spent. Equally: a bull-case cluster that thins out (downgrades from the $150 group) is the analysts who understood the asset walking away from it.

2. Deflate your own headline beat before you quote it

1. When a company posts a large EPS beat, never carry the headline number into the thesis until it has been decomposed.
2. Identify every non-cash, non-recurring or settlement item inside the reported figure — rebate settlements, legal accruals, tax true-ups, write-downs, one-time credits.
3. Recompute the beat against the adjusted consensus, and quote that number as the result. Both figures go in the write-up; the smaller one is the one the thesis is allowed to use.
4. Cross-check the beat against the revenue line — a large EPS beat on in-line revenue is usually a below-the-line item; a beat on 38–40% revenue growth is operational.
5. Take guidance as the tiebreaker: management raising the full-year number is a statement about the run-rate, and it cannot be manufactured by a one-off in the quarter just past.
6. Do the same test on the bad side too — a headline loss caused by an accounting event is not a business event (the same discipline the house applied to GILD's $10.5B GAAP loss).

• A quarter where the adjusted beat shrinks toward zero while the headline stays large — that is the point at which the "inflection" is being carried by accounting rather than by product sales. Check the net sales line and the guidance revision first, before the EPS print.

3. On a cash-rich balance sheet, value the business on enterprise value — and re-run your own yardstick net of cash

1. Take the market cap and subtract net cash (cash less debt) to get the price actually being paid for the operating business. Ignore the headline cap for multiple work.
2. Divide that EV by the guided revenue at the midpoint — not trailing revenue — so the multiple and the growth rate refer to the same period.
3. Set the resulting multiple beside the growth rate of the same line. A low multiple on a high growth rate is the anomaly worth explaining; state it as one sentence ("4.3x EV/revenue on 40% net-sales growth").
4. Re-run the house's preferred ratio the same way. If Price/Sales is the yardstick — because sales are less volatile and harder to fake — then compute EV/Sales, and show both the headline and the net-of-cash figure so the reader sees the adjustment rather than only its result.
5. Ask what the balance sheet buys you beyond the discount: no refinancing risk, a self-funded pipeline, and the capacity to acquire — optionality that a leveraged peer at the same multiple does not have.
6. Then name the reason for the discount in one clause. A cheap multiple you cannot explain is a multiple you have not finished analysing.

• Net cash being spent — a large acquisition converts the discount into goodwill and removes the cushion that made the option affordable. Also watch EV/sales rising without the growth rate rising: that is the market completing the re-rate, and the entry edge closing.

4. Attach the structural bear case to the asset that carries it, in the same paragraph

1. For each major revenue line, write the expiry, contract end or exclusivity date next to the revenue figure. A revenue line without a duration is an unfinished analysis.
2. Size the exposure as a share of total sales, so the worst case is a number rather than an adjective.
3. Separate what is certain from what is contingent in the protection — a compound patent that expires on a date is certain; formulation patents that may extend exclusivity are not, and must be labelled as such.
4. State the offsetting programme as elapsed effort, not intention: what has the company already spent years and dollars building, and is there evidence in current revenue that it is working?
5. Define the falsifiable version of the bear case — not "the cliff is bad," but "the replacement lines grow too slowly to offset the decline" — and identify the quarterly number that tests it.
6. Only then decide whether the discount already in the price is bigger or smaller than the exposure.

• The formulation-patent litigation outcome (it decides whether the cliff is 2028 or 2033), and the quarterly ratio of replacement revenue growth to legacy revenue. Opzelura plus new launches must be adding faster than Jakafi will subtract — that ratio, not the total, is the thesis.

5. Make a binary catalyst survivable — underwrite the platform under it, not the catalyst

1. Before sizing a position around a catalyst, ask the only question that matters on the downside: what is left if it fails?
2. Distinguish a single-asset company (the catalyst is the company) from a multi-franchise platform (the catalyst is an addition). Only the second is a candidate for this trade — the first is a bet regardless of how attractive the data looks.
3. Require the base business to justify the current price on its own. If it does, the catalyst can be sized as free optionality; if it doesn't, you are paying for the catalyst whatever the write-up says.
4. Name the date and venue at which the option resolves, so the position has a known event horizon rather than an open-ended hope.
5. Write both outcomes in advance as one sentence each — the upside ("the commercial opportunity and the acquisition value are both sizable") and the downside ("absorbed from a diversified platform") — before the event, not after.
6. Count the other shots on goal. Multiple late-stage programmes and near-term approvals mean the failure of any one is a setback rather than a thesis break.

• The ESMO oral presentation — and, just as important, whether the stock has already re-rated into it. An option bought for nothing is a different position from the same option after a 20% run; the four pending approvals are the second set of resolution points, and a stumble on any of them "hits a stock priced for continued execution."

6. Validate an extraordinary clinical read against a proven mechanism and a stated base rate

1. Establish the base rate first: what does the current standard of care achieve on the same endpoint, in the same line of treatment? Without it, a response rate is a number with no scale.
2. Check whether the mechanism has already been proven in humans elsewhere — an analogous target validated in another disease converts "does this class work?" into "does this molecule work here?", which is a much smaller question.
3. Size the addressable slice precisely (which mutation, what share of patients), rather than quoting the whole disease.
4. Quote the data with both its endpoints and both its combinations — response rate and disease control, each arm separately. A single cherry-picked figure is not a read.
5. Then attach the trial-design caveats yourself: interim versus final, single-arm versus randomised and placebo-controlled, population quirks, and the absence of durability and long-term safety data.
6. Cite an independent read of the upcoming catalyst (a sell-side or clinical opinion) rather than relying only on the company's framing.
7. Convert the whole thing into a category judgement — lottery ticket or practice-changing — and say which one you are underwriting.

• The ESMO cut: whether the response rates hold at a larger, more mature n, and whether duration of response and survival data accompany them — response rate without durability is the classic way an early oncology read disappoints. Watch too for a competitor's G12D data landing at the same conference.

7. Distinguish an acceptable late entry from an unacceptable one — then price the wait

1. Say out loud that you are late when you are. Locate the breakout that has already happened and date it; an entry thesis that hides the missed signal cannot be checked.
2. Ask the question that decides whether late is fatal: what drove the move? Decompose the advance into earnings growth versus multiple expansion.
3. An advance carried by earnings, with the multiple flat or compressing, leaves the valuation case intact — you are late to the price, not to the value. An advance carried by multiple expansion means the re-rating is the return, and it is already collected.
4. Check the multiple against its own history (five-year P/E and P/S ranges) rather than against peers, to see where in its band the stock is being bought.
5. Measure the distance above the 200-day moving average, and translate it into a specific retracement level rather than a vague warning.
6. Issue a segmented instruction: the level to buy for those adding now, and permission to wait for the more conservative — which is also how the analyst records the risk without abandoning the call.

• The $110 area and the 200-day: a touch that holds converts a wait into an add, while a break below the level that generated the original breakout (the ~$85–90 shelf) would say the earnings-driven read has failed. And watch the multiple: if P/E climbs back toward the 2021 peak on unchanged estimates, the advance has stopped being earnings-driven.

8. Build a sector sleeve one name at a time, and state the view that makes each addition additive

1. When a new pick belongs to a sector you are already accumulating, present it as an addition to that sleeve and list the names already in it with their entry dates. The sequence is the record.
2. State the sector-level view separately from the single-name case, so a reader can accept one without the other — and so the sleeve can be judged as a unit later.
3. Check the new name is additive rather than duplicative: a different sub-industry, a different driver, a different risk. A sleeve of four correlated names is one position with extra fees.
4. Build gradually across months, not in a single allocation, so the sector call is dollar-cost-averaged and each entry gets its own price discipline.
5. Keep the sleeve's thesis falsifiable at the sector level — "healthcare and biotech are in a bullish recovery mode" is testable against sector relative performance, independent of any one stock.

• Whether the sector view is actually working: healthcare and biotech relative to the S&P is the sleeve-level scoreboard, and it should be checked before adding a fifth name. Also watch concentration — four names in one sector is a deliberate overweight, and each subsequent addition raises the cost of being wrong on the sector call rather than on the stock.