02 · Enterprise Portfolio Management

Level 3's portfolio module scored and selected a handful of candidate projects against a budget. At enterprise scale — dozens of portfolios, hundreds of initiatives, multiple business units each with their own priorities — the hard problem shifts from "how do we score projects" to "how do we allocate capital and capacity across portfolios that don't trust each other's scoring, and don't want to."

The enterprise portfolio hierarchy

Level	Scope	Owner	Decision made
Enterprise portfolio	All investment across the company	Executive committee / CFO	How much capital goes to each business unit's portfolio
Business unit portfolio	One BU's initiatives	BU portfolio board	Which of the BU's own candidate projects get funded
Program	Related projects within a BU	Program manager	Sequencing and cross-project trade-offs
Project	Single deliverable	Project manager	Execution

The failure mode unique to this scale: a BU's internally-consistent scoring model (module 02, Level 3) ranks its own projects fairly relative to each other, but scores from two different BUs are not comparable unless someone normalises them — a BU that scores generously to protect its own budget will always outrank a BU that scores conservatively, independent of actual value.

Normalising cross-BU scores

BU	Self-reported top project score	BU's average project score (all projects)	Normalised score (top project relative to own baseline)
Retail Ops	88	71	88 − 71 = 17 above own average
Digital	92	85	92 − 85 = 7 above own average
Supply Chain	75	58	75 − 58 = 17 above own average

Comparing raw scores (92 > 88 > 75) would rank Digital's top project first. Comparing each project against its own BU's baseline reveals Retail Ops and Supply Chain's top projects are actually standout bets relative to what those BUs normally propose, while Digital's "92" is only middling by Digital's own generous standards. This relative-ranking technique is the standard fix for cross-BU scoring inflation, and it's the calculation an enterprise PMO runs before capital allocation, not the raw scores BUs submit.

Capital allocation models

Model	How it works	Best for
Zero-based	Every BU re-justifies its full ask from $0 each cycle	Rapidly changing strategy, willingness to disrupt
Incremental	Prior year's allocation ± a negotiated delta	Stable, mature portfolios
Strategic buckets	Fixed % to pre-defined themes (e.g., 40% growth, 35% run, 25% transform) regardless of BU	Enforcing a strategic mix top-down (Level 3, module 02's mix table, at enterprise scale)
Venture-style staged funding	Small initial tranche; more released only after a project clears defined milestones	High-uncertainty bets (new markets, unproven tech)

Worked example: staged funding decision

A digital transformation initiative requests $4,000,000 upfront. The portfolio board instead approves staged funding:

Stage	Funding released	Milestone required to unlock next stage
Stage 1	$600,000	Proof-of-concept validated with 2 pilot customers
Stage 2	$1,400,000	Pilot shows ≥15% efficiency gain, technical architecture approved
Stage 3	$2,000,000	Stage 2 milestone hit on time and within 10% of stage budget

python3 -c "print(600000+1400000+2000000)"

→ $4,000,000 total, matching the original ask — but the enterprise never has more than $600,000 at risk until the concept is proven, and can redirect the remaining $3,400,000 to another initiative at any stage gate if the milestone isn't met. This is the mechanism that lets an enterprise portfolio take on genuinely uncertain bets without betting the full amount on day one.

Enterprise risk aggregation

A risk rated "medium" independently in five different BU portfolios can be a severe enterprise risk if it shares a root cause across all five — the same pattern as Level 3 module 03's correlated-risk example, at portfolio scale.

Risk	BUs affected	Independent BU rating	Aggregated enterprise rating
Single cloud provider outage	Retail Ops, Digital, Supply Chain, Finance	Medium (each BU has "some" mitigation)	Severe — an outage would hit all four simultaneously
Key regulatory change (data residency)	Digital, Supply Chain	Medium	High — same root cause, correlated timing
Senior engineering talent shortage	All five BUs	Low individually	Medium-High in aggregate — they're all competing for the same limited talent pool

An enterprise PMO's distinct value here is exactly this aggregation step: no single BU portfolio board can see that its "medium" cloud-outage risk is one of four identical bets on the same infrastructure.

How It Actually Works

Cross-business-unit score normalization is a z-score standardization problem: if BU-A's project scorers rate everything 6-9 out of 10 (lenient) and BU-B's raters use the full 1-10 range honestly, comparing raw scores cross-BU systematically favors BU-A's mediocre projects over BU-B's good ones. The fix is normalizing each project's score against its own BU's score distribution, z = (score − BU_mean) / BU_stdev, before ranking across the enterprise portfolio — this is the same normalization math underlying standardized testing, applied to project prioritization instead. Staged capital allocation (funding a tranche now, more only if a gate is passed) is, in financial terms, a real option: the enterprise is paying a small premium (the first tranche) for the right, not the obligation, to invest the rest later once uncertainty resolves, which mathematically dominates committing 100% of capital up front whenever the project's true value has significant variance — the option value comes specifically from being able to walk away cheaply if the first stage reveals bad news.

Exercise

Three business units report their top project scores: BU X scores its top project 80 (BU average 68), BU Y scores its top project 95 (BU average 90), BU Z scores its top project 72 (BU average 50).

1. Normalise each BU's top project score against its own baseline and rank the three. State which BU's raw score is most inflated relative to its own typical proposals.
2. Design a 3-stage staged-funding plan for a $2,500,000 initiative, specifying the dollar amount and the milestone required at each stage, verifying the stages sum to $2,500,000 with python3 -c.
3. Propose one risk that could plausibly be rated "low" or "medium" in each of three separate BU portfolios but should be aggregated to "high" or "severe" at the enterprise level — explain the shared root cause.