Slide recipe: the peer-gap chart

The single most-copied slide in activism. This is the exact layout, used with minor variations from Ackman (Canadian Pacific 2012) to Elliott (Phillips 66 2025). Build it this way unless you have a specific reason to deviate.

See also patterns/peer-gap.md for the rhetorical logic. This file covers the physical slide build.

Anatomy

 ┌─────────────────────────────────────────────────────────────────┐
 │                                                                 │
 │   [TITLE — THE TAKEAWAY, NOT THE CHART NAME]                    │
 │   44–52pt serif, left-aligned, max 2 lines                      │
 │                                                                 │
 │   [SUBTITLE — METRIC + PERIOD + ORIENTATION]                    │
 │   16pt sans, var(--text-2) colour                               │
 │                                                                 │
 │  ─────────────────────────────────────────────────────────      │
 │                                                                 │
 │   CN        ██████                      63.2%                   │
 │   KSU       ███████                     66.8%                   │
 │   UNP       ████████                    68.5%   ← Peer median   │
 │   NSC       █████████                   71.1%                   │
 │   CSX       ██████████                  74.3%                   │
 │   ──────────────────────────────── [dotted line at 72]          │
 │   CP        ██████████████████  81.4%  (Target, accent colour)  │
 │                                                                 │
 │                                                                 │
 │   "To match CN's 63%, CP must close 18 percentage points        │
 │    of operating-ratio gap — worth ~$2bn of EBITDA."             │
 │                                                                 │
 │                                                                 │
 │   Source: Company 10-Ks, fiscal 2011. Lower is better.          │
 │                                                                 │
 └─────────────────────────────────────────────────────────────────┘

The six layers, top to bottom

1 · Title slot (the takeaway)

44–52pt serif, left-aligned, max 2 lines.

The title IS the conclusion, not the subject. Never:

❌ "Operating Ratio — Class I Railroads"

Always:

✅ "CP has the worst operating ratio of any Class I railroad."

Use markup to emphasise the single most important word — worst, lowest, highest. Italicising it (in the serif face) draws the eye without shouting.

2 · Subtitle (metric, period, orientation)

16pt sans, muted colour (var(--text-2) equivalent).

Three facts, one line:

Metric name ("Operating Ratio", "EBITDA Margin", "Same-store Sales Growth")
Period ("FY2011", "TTM", "5-year average")
Orientation ("Lower is better" or "Higher is better")

The orientation line is crucial. A chart with CP at the top and peers at the bottom reads opposite ways depending on metric. Tell the reader which direction is bad.

3 · Thin horizontal divider

1px line, var(--border-subtle). Separates the narrative header from the quantitative body. Helps the eye calibrate.

4 · Chart body (bar chart, horizontal)

5–7 bars, sorted ascending (or descending depending on orientation).

Horizontal bars, not vertical. Company names are readable left; vertical labels require rotation and feel rushed.
Sorted by value — the eye should be able to rank by glance.
Peer bars in neutral grey (#7c7a73 or equivalent).
Target bar in accent colour — warm warning (salmon/ember #D97757), not alarm red. Red feels editorialised; warm warning feels analytical.
Data labels right of each bar, monospace, 14pt.
Peer median line — a thin dotted vertical at the median value, labelled "Peer median" or "Industry median".
No gridlines (or a single light one at the median). Gridlines are noise in a sorted bar chart.
No y-axis ticks if names are on the axis. Keep the chart breathing.

5 · The annotation callout

Immediately below the chart. 15–17pt sans or light serif italic.

One sentence tying the visual to the thesis — a translation of the gap into a monetary or strategic consequence.

Examples:

"To match CN's 63%, CP must close 18 percentage points of operating-ratio gap — worth ~$2bn of EBITDA."
"Every peer we studied outperforms on this metric by ≥8 points."
"A 100bps improvement would add $18 per share at current multiples."

This is what the reader screenshots and tweets. Make it quotable.

6 · Source line

9–10pt sans, 50% opacity, bottom-left.

Always:

Source of the data ("Company 10-Ks", "Bloomberg consensus", "FactSet")
Period / as-of date
Orientation reminder for long audiences ("Lower is better")

If the peer set is unusual, add a second source line naming the selection criteria. This pre-empts the "these aren't real peers" PR rebuttal.

Typography cheatsheet

Title           serif, 44–52pt, bold or regular with italic emphasis
Subtitle        sans, 16pt, secondary colour
Axis labels     sans, 13pt, medium weight
Data labels     monospace, 14pt, bold
Annotation      serif italic or sans, 15–17pt
Source          sans, 9–10pt, 50% opacity

Resist the urge to use four typefaces. Serif + sans + mono = three, and three is already at the limit.

Peer-set selection — the defensible version

The slide's credibility lives or dies in the peer set. Rules of thumb:

Use peers the target itself cites in its own 10-K or investor day. This pre-empts "they're not real comps".
Public-company peers only. Private comps are un-verifiable.
Same business model, not just same sector. Class I railroads are a tight peer set because they share gauge, network model, and regulatory regime. "Industrial companies" is not a peer set.
5–7 peers ideal. Fewer looks cherry-picked; more becomes a blur.
Include the target in the sort — don't show it off-chart.

Colour discipline

Peers        #7c7a73   (neutral grey, one bar per peer, all identical)
Target       #D97757   (warm ember — the accent colour)
Median line  #504E48   (dotted, low contrast)
Text         #EDECE8   (near-white on dark) or #1a1a1a (on light)
Annotation   #A9A7A0   (subdued — don't compete with the chart)

Absolutely avoid:

Rainbow legends (a different colour per peer)
3-D bars (instant credibility loss)
Drop shadows, gradients, or bevels
Stock-chart red for the target unless you're a short-seller

Variations

The time-series version

Same logic but across 3–5 years instead of a single point. Shows that the gap is persistent, not a fluke. Useful when:

The target's PR will argue "2023 was a bad year, we're recovering"
The gap is widening (trend is the punchline)

The multi-metric variant

If a single KPI doesn't carry the argument, use 3 KPIs side-by-side. Only use for 3 metrics max — beyond that the reader loses the punch. Each panel is its own horizontal bar chart with consistent styling.

The scatter / quadrant version

For margin × growth or margin × capital intensity. Target ends up alone in the bad quadrant. Works for early-stage activism where the peer grouping is less tight.

Common mistakes

Title describes the chart, not the takeaway. "Operating Ratios" → 3/5. "CP is the worst" → 5/5.
Too many bars. ≥10 peers and the chart becomes a rank list, not a point.
Wrong peer set. Always cite selection criteria in source line.
Hidden baseline. If the bars start at 40% instead of 0%, the gap looks larger than it is. Use full range unless the distortion is genuinely material to the argument.
Missing orientation. Without "Lower is better", the reader has to think. They won't; they'll skim past.

Exemplars

Pershing Square · Canadian Pacific (Feb 2012) — the archetype.
Starboard · Darden Restaurants (Sep 2014) — G&A margin gap vs. casual-dining peers.
Elliott · Arconic (Apr 2017) — 5/5 visual craft. EBITDA margin vs. aerospace-components peers.
Starboard · Autodesk (Aug 2024) — operating-margin gap vs. Adobe and Intuit.
Trian · PepsiCo (Jul 2013) — the classic snacks/beverages peer decomposition.