What it is
You break the target into its constituent businesses or asset classes, value each separately against its own peer set, and show that the total is materially greater than the current market capitalisation. The "gap" is the conglomerate discount — or worse, accounting opacity hiding a crown-jewel asset.
Why it works
- It forces the market to confront mismatched valuation multiples: "this company trades like X but owns Y, which trades at Z."
- It reveals hidden assets the market hasn't priced (real estate, brands, high-margin segments buried in consolidated financials).
- It provides the mechanical path to value realisation: spin, sell, IPO, or restructure — each with a mapped multiple.
When to use it
- ✅ Conglomerate structure (multiple distinct businesses)
- ✅ Hidden real-estate ownership (the McDonald's / Darden / Cheesecake Factory playbook)
- ✅ A high-margin segment buried inside a commodity business (ADT's monitoring, Xerox's services)
- ✅ Financial holding with non-consolidated stakes (Altaba / Yahoo / Sotheby's / Liberty structures)
- ❌ Don't force it on focused single-segment businesses — too little to break apart
Recipe
- Identify 2–5 distinct segments. More than 5 and you look like you're reaching; fewer than 2 and it isn't really a sum-of-parts thesis.
- For each segment:
- Pick a peer set of pure-play public comparables
- Assign a multiple (EV/EBITDA most common; EV/Sales for high-growth; P/E for steady compounders)
- Apply the multiple to the segment's standalone financials
- Sum the parts. Add net debt/cash at the parent level.
- Subtract a haircut for friction — real deals cost 3–7% of equity value in transaction costs, tax leakage, stranded overhead. Include it; it pre-empts management's counter.
- Show the single "big number" — per share — on the cover slide. "Sum-of-parts value: $86/share vs. $48 today."
- Provide the mechanical path: "Separating the real estate into a REIT unlocks $1B; the SoP math assumes this occurs within 24 months."
Headline language that works
- "The market values this company as a [dominant segment] business, but [hidden segment] is worth more than the current market cap."
- "Sum-of-parts value: $X per share, [Y%] above current."
- "Three independent businesses, three different multiples, one market misprice."
Common mistakes
- Cherry-picked multiples. If Segment A trades at 12x and you used 15x, you'll get torn apart. Use the median of the peer set and show the range.
- Ignoring stranded costs. When you spin the real estate, the OpCo loses economies of scale. Model it.
- No mechanical path. A SoP without a specific realisation path (spin / sale / REIT conversion / IPO) is a napkin exercise.
- Peer sets too loose. Each segment's comps must be genuine pure-plays.
- Inflating segment financials. Take them from the 10-K segment reporting, not from your allocation of shared costs.
The "big number" principle
SoP decks live or die by a single number on the cover slide: the delta between "current" and "SoP value". Lead with it in the title, repeat it in the first three slides, close with it on the last.
Exemplars
See examples/by_pattern.json → sum_of_parts.
- Pershing Square · McDonald's (Nov 2005) — the masterclass. Splits MCD into PropCo (real estate, 12–13x EBITDA), FranCo (royalty stream, high-multiple), and McOpCo (restaurant operations, low multiple). Unlocks $45–50/share vs. $33 at the time (+44%). The deck runs 149 pages defending each multiple.
- Starboard · Darden Restaurants (Sep 2014) — 294-page "Transforming Darden" deck. SoP of $67–86/share vs. $48, with proposed REIT separation as the unlock. Won full board turnover.
- Elliott · BHP Billiton (Apr 2017) — Australian dual-listing SoP, arguing unification + US petroleum divestiture unlocks ~$46bn.
- Trian · PepsiCo (Feb 2014) — snacks / beverages SoP, arguing Frito-Lay alone was worth more than PEP's market cap.
- Trian · DuPont (Feb 2015) — 6-segment SoP, "DuPont Can Be Great".
The slide itself
See slides/sum-of-parts-reveal-recipe.md for the canonical 2-slide
structure (build-up table + comparative bar chart).