Thesis Development

When to use

Use this skill before a market map exists and long before a target is named, when the partnership has an interest ("we like industrial services") but no written claim anyone can argue with. Use it also when a thesis has drifted into defending consecutive passes.

What it does

Produces a short thesis document: sub-sector boundary, the core claim in falsifiable form, where in the EBITDA bridge the return is created, the why-now catalyst, must-be-true conditions with owners, kill criteria agreed in advance, and the entry multiple range supported. It is written to be attacked, not to survey a sector.

Method

Step 1 -- Draw the boundary before writing anything

State the sub-sector precisely enough that a company is either in or out: activity, customer, geography, size band. "Healthcare" is not a boundary; "UK outsourced clinical waste collection, GBP 3-15M EBITDA" is. If it cannot be written as an inclusion test, it is a preference, not a boundary.

Step 2 -- Write the core claim in one falsifiable sentence

Use the form: because [structural condition], [a type of asset] will [measurable outcome] over [period]. Then name the observation that would make it false; if none exists, rewrite it. "The market is growing" fails; "operators below five sites cannot fund the licensing capex and will sell at 6.0-7.0x" passes.

Step 3 -- Locate the return in the EBITDA bridge

Decompose the return into organic growth, margin expansion, buy-and-build arbitrage, deleveraging and exit multiple change, then state which component carries most of it. A thesis that depends on exit multiple expansion is a bet on the market -- hold the exit flat to entry and label expansion upside.

Step 4 -- Establish why now, not why ever

Name the catalyst that makes this a 24-month window, not a permanent condition: a compliance date, a cost curve crossing, a cohort of sponsor-owned assets entering hold years 4-6. "Fragmented and growing" has been true for a decade and is not one.

Step 5 -- Write the must-be-true conditions

Three to five conditions, MECE, each measurable and each falsifiable by a named workstream against a named source. Conditions that cannot be tested in a normal process are underwriting assumptions -- separate them: assumptions get priced, conditions get proved.

Step 6 -- Pre-commit the kill criteria

For each condition, state the observation that kills the thesis, with a threshold and a date. Criteria written before diligence are worth far more than those written during it: once the team has seen the number it wanted, the threshold becomes negotiable.

Step 7 -- Price the thesis

State the entry multiple range the thesis supports and the multiple at which it stops working -- the discipline anchor for every screening decision taken under it.

Inputs

• The sector interest as currently stated, however loose
• Fund mandate: cheque size, EBITDA range, geography, hold, return threshold
• Prior deals reviewed in the sector, including each pass and why
• Observable transaction multiples, platform and bolt-on
• Evidence for the proposed catalyst

Output format

• Sub-sector boundary (inclusion test, one paragraph)
• Core claim (one falsifiable sentence)
• Value creation location (bridge decomposition in prose, never a table)
• Why now (catalyst with date or trigger)
• Must-be-true conditions (3-5, each with owner and test)
• Kill criteria (one per condition, with threshold)
• Supported entry multiple range, and where the thesis fails

Example

Fictional worked example -- European industrial water-treatment servicing.

Core claim: servicing of an installed base is consolidating, and a platform bought at 8.5x can buy sub-scale operators at 6.0x, so arbitrage carries the return, not growth. Platform EBITDA 12 at 8.5x is EV 102; four bolt-ons totalling EBITDA 18 at 6.0x cost 108; group EBITDA 30 on EV 210 is a blended entry of 7.0x, organic growth assumed at 3% and the exit held flat.

Kill criterion: if the last eight completed bolt-ons cleared above 7.5x, they would cost 135, group EV becomes 237 and blended entry 7.9x -- 0.6x against the platform price, which does not fund integration risk. The thesis is then dead, not re-argued.