The methodology for sequencing portfolio investment using Cost of Delay Divided by Duration (CD3) is set out in Appendix C of Evolve: The Operating Model AI Demands (Hodgson, 2026). This companion page presents that methodology as a standalone reference — covering the formula, its variable definitions, and the estimation approaches for organisations where direct revenue attribution is limited or absent. Three worked examples, drawn from different product lifecycle stages, illustrate both the calculation and its most common misapplication.
In Brief
- CD3 — Cost of Delay divided by Duration — sequences portfolio investment by the rate of value creation, not the absolute value of individual products.
- An organisation sequencing by Cost of Delay alone will consistently place long-duration items ahead of shorter items that generate more value per week of team capacity.
- The estimation discipline CD3 requires — defining what each product creates and how long delivery takes — is itself diagnostic: a portfolio that cannot answer these questions is not positioned to sequence correctly.
CD3 sequences by value per week of Duration, not absolute value
CD3 is a ratio: Cost of Delay divided by Duration. Cost of Delay (CoD) is the economic value of delivering a product sooner — or equivalently, the cost of each additional week of waiting. It is expressed in dollars per week. Duration is the elapsed time to complete the work, from commitment to delivery, expressed in the same time unit as CoD.
The resulting CD3 score carries the unit dollars per week per week of duration. The absolute value of any single item matters less than the relative ordering: items with higher CD3 scores are sequenced before items with lower scores. This is the core sequencing heuristic identified by Black (2010) as optimal for product development portfolios under resource constraint. The method works because it quantifies the trade-off between the value a product creates and the time required to create it — a trade-off that informal sequencing processes systematically underweight.
Three variable definitions are critical for correct application.
Cost of Delay is forward-looking. It is the value that will be deferred if delivery of this product is postponed by one additional week — not the budget spent to date, not the revenue earned by the product, not the product’s projected lifetime value. It is the marginal cost of waiting one more week.
Duration is elapsed calendar time: the clock that starts at commitment and stops at delivery, including dependency waits, handoffs, and scheduling constraints. A product requiring four weeks of engineering but six weeks of elapsed time because of integration dependencies carries a Duration of six weeks, not four.
The CD3 score is a rate: value per unit of time per unit of duration. Two products with equal Cost of Delay will produce different CD3 scores when their Duration differs. The shorter product is sequenced first, because it returns value sooner and frees capacity for the next item — a benefit that grows with each additional product in the queue (Black, 2010).
Cost of Delay is estimable in every portfolio, including those without direct revenue
Many government and not-for-profit products do not have direct revenue streams. Cost of Delay remains estimable — the estimation shifts from “what revenue is deferred” to “what value is deferred.”
Four estimation approaches apply to non-revenue products, and the best choice depends on the nature of the value the product creates.
Mission-value estimation identifies the specific outcome the product enables and asks what each week of delay costs in terms of that outcome not being achieved. A capability that reduces the time to process a public service case — from ten days to three, across 500 cases per week — produces a calculable Cost of Delay without any revenue figure: the value of seven days of processing time, multiplied by 500, multiplied by the unit cost the organisation assigns to processing delay. The estimate is approximation; it does not need to be precise, because the CD3 calculation requires only that CoD estimates be consistent in their units and logic across the items being compared.
Risk-reduction valuation applies where the product addresses a regulatory, compliance, or reputational exposure. The Cost of Delay is the exposure accruing each week the product is absent. Actuarial or legal estimates of the exposure provide a floor value. Where the weekly probability of the risk materialising is available, an expected-value calculation applies: Cost of Delay equals the weekly probability of the event multiplied by the cost if it occurs.
Strategic position valuation applies where the product defends or establishes a competitive or strategic position. The Cost of Delay is the erosion of that position per week — assessed through user demand data, competitor capability development, or explicit strategic commitments the organisation has already made and cannot afford to miss.
Proxy signal estimation applies where none of the above produces a direct estimate. Signals from user behaviour provide a basis for relative ordering: volume of the problem the product addresses, severity ratings, frequency of workarounds, and the operational cost those workarounds impose per week. These signals do not produce a precise dollar value, but they enable relative ordering — which is what the CD3 calculation requires. An organisation that can say “this problem costs twice as much per week as that one” has enough information to sequence correctly.
Black (2010) observes that every product carries a Cost of Delay, even when that cost is not immediately visible. Organisations that treat CoD as absent have typically not defined what outcome the product creates. The estimation discipline required by CD3 is itself diagnostic: it forces the portfolio to be described in terms of value created, not activity delivered.
Discovery-phase Duration can be estimated without committed build scope
Discovery-phase products present a higher uncertainty band than products in active development. That uncertainty does not exclude them from the CD3 calculation — it requires that Duration be estimated using methods appropriate to the phase.
Reference class forecasting draws on what comparable discovery efforts took in this organisation. A sample of five to ten prior discovery phases of similar scope produces a reference distribution. The midpoint of that distribution is the Duration estimate. Where prior examples are not available within the organisation, Black (2010) provides benchmark ranges for discovery phases across product types, which serve as a starting point until the organisation builds its own reference data.
Bounded range sizing expresses Duration as a range — for example, four to eight weeks — rather than a point estimate. The midpoint is used for initial sequencing. As the discovery phase progresses and uncertainty reduces, the estimate is updated and the CD3 score recalculated. The initial sequencing decision is revisable; the methodology requires no more precision than the organisation has at the point of decision.
Iteration-based sizing applies where discovery is structured as fixed-length cycles. Three iterations of two weeks each produces a Duration estimate of six weeks. This approach creates an explicit decision point at the end of each iteration: the portfolio review can update the Duration estimate and adjust the product’s sequence position as the discovery phase reveals more about the scope of subsequent work.
The CD3 calculation is a sequencing heuristic, not a delivery commitment. A Duration estimate revised during execution updates the CD3 score for the next sequencing decision. It does not invalidate the prior decision, because that decision was made with the best available information at the time. Refusing to estimate Duration on the grounds that discovery work is uncertain is not a conservative position — it is an abdication of the sequencing decision, which the portfolio will make anyway, implicitly, through default to whatever was started first or whoever asked most recently.
Ordering by Cost of Delay alone produces a different — and worse — result than CD3
The table below shows CD3 sequencing across a portfolio of seven products in a government digital team. All monetary values are notional; the sequencing logic applies regardless of scale.
| Product | Cost of Delay ($/week) | Duration (weeks) | CD3 Score | Sequence |
|---|---|---|---|---|
| A — Claims intake automation | 45,000 | 8 | 5,625 | 3 |
| B — Correspondence management | 30,000 | 4 | 7,500 | 1 |
| C — Self-service appeals portal | 60,000 | 12 | 5,000 | 6 |
| D — Case officer dashboard | 20,000 | 3 | 6,667 | 2 |
| E — Reporting and analytics | 15,000 | 6 | 2,500 | 7 |
| F — Document verification | 50,000 | 10 | 5,000 | 5 |
| G — Notification service | 25,000 | 5 | 5,000 | 4 |
The CD3 sequence is: B (7,500), D (6,667), A (5,625), then the three-way tie at 5,000 resolved by shortest remaining Duration — G (5 weeks), F (10 weeks), C (12 weeks) — then E (2,500).
This sequencing differs significantly from ordering by Cost of Delay alone, which would sequence C, F, A, B, D, G, E. Product C has the highest Cost of Delay at $60,000 per week, but its 12-week Duration means it generates less value per unit of team capacity than Products B, D, and A. Sequencing B and D first — shorter items with high CD3 scores — returns value earlier and reduces the total Cost of Delay accumulated across the portfolio during the sequencing period. Sequencing by Cost of Delay alone is a common error: it optimises for the value of individual items rather than for the portfolio’s aggregate throughput (Black, 2010).
The tie-breaking rule at a score of 5,000 — shortest Duration first — reflects the same logic. When two items generate value at the same rate, completing the shorter one first returns capacity sooner and reduces the cumulative Cost of Delay for everything waiting behind it.
The CD3 calculation applies equally across discovery, active development, and mature products
Example 1: Discovery-phase product
A government agency is planning a new digital self-service portal. No development has started; the team is at problem definition with no committed build scope.
The portal is expected to reduce call centre contact volume by 200 contacts per week. At $40 per contact, that is a Cost of Delay of $8,000 per week. Reference class forecasting of two prior portal discovery efforts in the same organisation puts Duration at six weeks.
CD3 = $8,000 / 6 = $1,333 per week per week of duration.
This score is placed in the portfolio ranking alongside products already in active development. If any product currently in flight carries a CD3 score below $1,333, the discovery work sequences ahead of it — even though the discovery phase has produced no deliverable yet. The CD3 methodology applies equal treatment to discovery and delivery: both are investments of team capacity with a cost if deferred.
Example 2: Active development product
A risk analytics module is six weeks into a projected 12-week delivery. The module enables a risk-identification process currently performed manually at a cost of $20,000 per week in analyst time, which is the Cost of Delay.
For sequencing purposes, Duration is six weeks remaining. The six weeks already elapsed are not included — only the work yet to be done determines the current CD3 score. The elapsed time is sunk.
CD3 = $20,000 / 6 = $3,333.
This score goes into the portfolio ranking against other items competing for team capacity from this point forward. The six weeks already invested are irrelevant to that calculation (Hodgson, 2026).
Example 3: Mature product with enhancement
A mature case management system has an enhancement request for bulk reassignment functionality. The requesting team estimates this saves $5,000 per week in administrative overhead, giving a Cost of Delay of $5,000 per week. Development would take three weeks.
CD3 = $5,000 / 3 = $1,667.
This enhancement sequences after the risk analytics module (CD3 $3,333) and ahead of the discovery portal (CD3 $1,333). That the case management system is mature and the portal is new does not affect the calculation. The CD3 methodology treats every item in the portfolio on the same terms: what value does completing this create, and how long will that take? Lifecycle stage is not a variable (Hodgson, 2026; Black, 2010).
Sunk cost has no role in CD3 sequencing
The CD3 calculation is forward-looking. What has already been invested in a product has no bearing on which product should be sequenced next. Cost of Delay, Duration, and the resulting CD3 score all measure from the present moment forward.
The sunk cost fallacy applied to portfolio sequencing produces a specific and common error: portfolios ordered by “what has already been started” rather than “what creates the most value next.” A product that has consumed significant investment over many months accumulates weight in the portfolio — not because it creates the most value per week of remaining effort, but because the organisation feels committed to protecting what was already spent.
The counter-example makes this concrete. Two products are available for sequencing:
Product A has had $3M invested over 18 months. The team is still active on it. Its current CD3 score is 0.8.
Product B has had no prior investment. Its current CD3 score is 4.2.
A portfolio team reasoning from sunk cost continues Product A. A portfolio team applying CD3 sequences Product B next.
The $3M invested in Product A is gone regardless of what the team decides. That investment does not change what Product B will create per week of Duration. It does not reduce the Cost of Delay that Product B is accumulating while it waits. Continuing Product A because of past investment does not recover the $3M. It delays the realisation of Product B’s value and adds to Product B’s accumulated Cost of Delay by exactly as many weeks as the team spends completing A.
The sequencing question is: given where we are today, which product creates the most value per week of team capacity? Past investment is not part of that question (Hodgson, 2026; Black, 2010).
The same logic applies to staff already assigned to a product in progress. Salaries already paid are sunk; the only decision available is what to apply team capacity to from this point forward. The salaries yet to be paid are part of the Duration estimate — future cost that should inform the Duration figure, not justify the decision to protect past cost.
Organisations that embed sunk cost reasoning into portfolio governance — through stage-gate processes that favour continuation of existing investments, or through a cultural norm that “we finish what we start” — produce portfolios ordered by past decisions rather than forward value. CD3 makes that ordering visible and gives the portfolio team the evidence to make a different argument. Changing the sequencing decision requires more than the calculation; it requires governance conditions that allow the calculation to determine the sequence rather than to be overruled by it.
What this means for senior leaders
- If your portfolio team is sequencing by Cost of Delay alone, it is placing long-duration items ahead of shorter items that generate more value per week of team capacity — the result is a sequence that looks optimal on value but accumulates avoidable delay across the queue.
- Discovery-phase work is not exempt from CD3 prioritisation. A discovery effort with a high CD3 score sequences ahead of in-flight delivery work with a lower score, regardless of which team is already assigned.
- The estimation discipline CD3 requires cannot be skipped: a portfolio that cannot define what each product creates, or how long delivery takes, is not equipped to sequence correctly.
- Sunk cost is not a sequencing signal. Prior investment in a low-CD3 product does not change what a high-CD3 product will create per week — or what it costs to keep that product waiting.
- Portfolio governance that overrides CD3 on the grounds of commitment or continuity produces sequences ordered by past decisions rather than forward value — CD3 makes that trade-off visible and gives teams the evidence to make a different argument.
References
- Black, D. (2010). The principles of product development flow: Second generation lean product development. Crisp Learning.
- Hodgson, M. (2026). Evolve: The Operating Model AI Demands (Appendix C). Zen Ex Machina.
