Evaluate a portfolio of data domains and products

A portfolio turns competing data requests into capacity and investment decisions. Separate eligibility from priority before ranking: a product without a committed consumer, use right or source contract does not compete regardless of attractive value. Compare eligible items through one decision unit covering outcome, total cost, dependency, risk and reversibility. Mandatory constraints stay outside a weighted average, and a score is never a permanent promise; ranking changes when evidence changes.

A portfolio selects interdependent investments under constraint; it is not a cosmetic ranking of domains. It begins with non-tradeable eligibility, then compares value, risk reduction, readiness and cost among eligible choices. Dependencies prevent funding a product that requires a missing capability. Governance approves weights and constraints; management gathers evidence, runs trials and updates the record.

Use a portfolio register with four distinct judgments. Eligibility rejects items missing a consumer, right or source. Value testing links each item to a decision change. Dependency analysis identifies what must precede construction. Capacity allocation then funds eligible items with an explicit reserve. Preserve the rejection reason and fact that restores an item, because an unconditional backlog becomes a hidden commitment.

Supplier is ineligible because the right to create a derived product is undocumented. Customer offers higher benefit but waits for an identity service; product offers faster learning with ready data. A third option funds source repair only. The staged decision funds product for eight weeks, preserves customer for review after the identity decision and does not turn deferral into permanent rejection.

If customer scores 8 for value, 7 for risk reduction and 6 for readiness, its result at 50%, 30% and 20% is 7.3. Product scores 7, 5 and 8, producing 6.6. Raising readiness to 40% while lowering value to 30% narrows the gap. Record scale direction and unrounded inputs; a sensitivity-dependent ranking is not a permanent fact.

Do not monitor a portfolio through delivery alone. Compare realized benefit with actual capacity, the share of items losing eligibility, investment concentration in one domain and time dependencies remain open. An on-time unused product creates no value; a shared repair consuming current capacity but unlocking three items may beat a quick standalone report. Indicators expose that trade-off rather than compressing it.

After the quarter, the portfolio lead compares realized benefit with actual capacity. A product consuming 35% of capacity while changing one decision is not protected by its old score. Alternatives are re-estimated with new information, and the record distinguishes dependency delay, weak adoption and optimistic estimation.

The portfolio decision is which domain receives trial funding next quarter, not which domain is universally most important. Customer data supports churn reduction, product data supports lower stockouts and supplier data supports continuity. Apply eligibility gates before weighting. Supplier depends on third-party data whose contract permits operations but not a derived analytical product, so it is temporarily excluded; a weight cannot purchase a missing right. Customer and product remain. Customer estimates 180,000 fictional units of benefit but needs six months for an identity service. Product can test 110,000 of benefit in eight weeks with lower continuing cost. Define decision value, risk reduction and readiness on scales that point in one direction, retaining raw values before normalization. At 50% value, 30% risk and 20% readiness, customer leads narrowly; raising readiness to 35% puts product first. A ranking reversed by a small weight change is fragile. The unavailable identity dependency supports funding product now, with a customer review after the identity decision. Eight weeks later, measure whether replenishment choices changed and stockouts fell without inflating obsolete stock. If source quality prevents the test, product returns to the readiness gate. The decision record contains gates, raw scores, transformation, weights, sensitivity, dependencies, cost and the fact that would return each option to competition. This prevents the strongest presentation from winning and keeps mandatory constraints outside tradeable scores.

At quarter close, reconstruct the decision from the old record before viewing outcomes: did the evidence available then justify the choice? Update estimates with new learning without judging a sound earlier decision through hindsight. Also record the alternative opportunity displaced by capacity so council sees the cost of continuing, not only the cost of stopping. Portfolio stress test: reduce forecast benefit 30%, increase transition cost 20% and delay a major dependency one month. Identify items whose position changes and require more evidence before commitment. An item remaining eligible and preferred across scenarios is more robust than one leading through a single fragile score. The decision also exposes its concentration cap.