Observed records
Incident dates, linked tickets and severity. Fictional records here; operational evidence in your systems.
Platform investment portfolio
Are we working on the right things?
I bring customer and support evidence into platform priorities, working with leadership and engineering to weigh customer need, business value and delivery tradeoffs. Explore how those decisions connect to delivery and outcome review. See the experience behind this approach
Worked example · Fictional dataHow this works ↓
All portfolio initiatives
Select two or three initiatives to compare at the same horizon.
Remaining investment basis · No portfolio benefit total: customer claims may overlap.
First three visible issues in the current sort direction, for discussion.
Explore a decision
Three entirely fictional stories: customer need, evidence, alternatives and proposed outcome review.
01 / Understand the investment
The reasoning stays visible.
Establish customer and business need, inspect evidence and limitations, evaluate alternatives and potential value, check feasibility and obligations, make a conditional decision, then measure outcomes and revisit.
A high modeled return may justify discovery before a full build. Enabling work may have no standalone revenue. An investment gap may be acceptable for a measurable operational need. Revenue is one decision input alongside customer need, evidence and delivery constraints. Explore the three decision stories ↑
What does an investment gap mean?
A gap means modeled benefits have not covered investment and maintenance by the selected month. A surplus means they exceed those costs in the scenario. Neither is money earned by an incident, actual profit or a cash balance.
Portfolio decisions use remaining investment. Already-spent money is shown separately and is not charged again. Original issue scenarios use their full initial-fix estimate; the scope is always labeled.
All evidence here is fictional. The demonstration does not reproduce a former employer’s records, formula or outcomes.
Estimated business inputs
Staff effort, loaded cost, revenue exposure, implementation cost and elapsed delivery weeks. Reviewed with their owners.
Decision assumptions
Uptake, contribution margin, prevention and cash realization. Adjustable judgments, not observed outcomes. Issue-only weights do not rank new capabilities.
Source records → Unique incidents → Operational impactInvestment comparison / Issue analysis
Compare contribution and cash savings with remaining investment, timing and capacity. Gross incremental revenue is reduced by contribution margin; released capacity is separate. Mandatory and enabling work may have no standalone financial estimate. Shared benefits are not added into a portfolio promise. Issue scores remain separate from feature comparisons. Assigned attrition is an existing-customer loss probability; new-customer loss concerns prospective business. Neither is measured churn. Delivery weeks are elapsed time, not engineer-weeks. Zero assigned exposure does not establish zero indirect customer impact.
Counting and source reconciliation
Map the ITSM problem key and customer-support category to one issue. Shared event references identify one incident; retain the highest linked severity. The observation window includes its start and excludes its end.
Frequency = unique incidents × 30 ÷ observed days. Annualized volume = unique incidents × 365 ÷ observed days.
Operational cost: staff time and direct expense
Cost per incident = staff hours × loaded hourly cost + direct expense. Observed cost = unique incidents × cost per incident. Staff value is an estimate of resources consumed; it is not all cash that can be saved.
Priority score: fixed anchors and adjustable weights
Each factor is capped at 100. Score = Σ(factor × weight) ÷ Σ(weights). Fixed anchors mean filtering does not change an issue’s score. Frequency and cost both reflect recurrence; adjust their weights deliberately.
A high score identifies a candidate for investigation. It does not establish root cause, effort-adjusted return or delivery approval.
Forecast: timing, cash, capacity and customer value
For m months, active benefit fraction = max(0, m ÷ 12 − delivery weeks ÷ 52). Historical incident rates remain constant; benefits accrue linearly after delivery. This planning convention uses 365-day annualized demand and a 52-week delivery year, without seasonality or growth.
Annual cash savings = annualized incidents × prevention share × (direct expense + staff cost × cash-realization share). The remaining avoided staff value is capacity, not additional cash. Staff hours released measure time, not a staffing commitment.
Annual protected revenue = exposed existing annual revenue × assigned attrition probability × relative risk reduction. Recovered new revenue uses the annual opportunity × assigned loss probability × relative risk reduction. A 20% loss probability reduced by 50% becomes 10%, not −30%.
Cumulative net = cash savings + (protected revenue + recovered new revenue) × contribution margin − implementation − maintenance. Benefits and maintenance accrue only after delivery; implementation is counted once, upfront, in every cumulative horizon.
Interpretation: uncertainty, negative net and overlapping benefits
Negative net means modeled benefits do not cover investment and maintenance in that horizon. Released capacity may still support a decision. Customer probabilities are assigned annual revenue scenarios, not measured retention or contracted sales.
Customer cohorts may overlap across issues. Do not sum issue revenue or net values into a company-wide promise. Compare forecasts with outcomes and revisit assumptions.
Feature value, sunk cost and resource tradeoffs
Annual incremental revenue = eligible accounts × incremental adoption share × incremental annual revenue per adopting account. This is above the baseline and assumes the modeled run-rate begins after delivery, without a validated ramp. Evidence grades are qualitative; they are not another probability multiplier.
Investment result = cash savings + margin-adjusted revenue − remaining implementation − ongoing cost in the horizon. Already-spent cost does not change this forward decision. Remaining person-weeks are shown per team; elapsed weeks are not staffing effort or an available-capacity forecast.
Compare obligations, dependencies and unquantified value explicitly. A dependency’s costs are owned once; shared customer benefits may overlap. No company-wide net-value total is claimed.
Fictional feature example: Self-service onboarding
Worked example: Executive reports use stale data
Read the anonymized product experience behind this approach →
02 / Connect evidence across the organization
Bring commercial and operational evidence together.
A first useful version can use approved exports. Repeatable integrations follow once definitions and decisions are trusted.
Service & support records
Incidents, problems, services, severity and account references.
ServiceNowJira Service ManagementZendesk
Business assumptions
Research themes, product usage, incremental revenue, margin and reviewed costs.
Customer research / CRMProduct analyticsFinance reportingReviewed worksheets
Delivery estimates
Current commitments, team effort, remaining spend, dependencies and acceptance.
Existing backlogDelivery tools
Map, reconcile and validate
Approved exports first · APIs or existing data pipelines later
Approved exports first · APIs or existing data pipelines later
Shared initiative and issue evidence
Source traceability · Reviewed links · Exceptions · Visible freshness
Source traceability · Reviewed links · Exceptions · Visible freshness
Priorities and forecasts
Outcomes · Current commitments and alternatives · Explicit investment assumptions
Outcomes · Current commitments and alternatives · Explicit investment assumptions
Optional AI assistance opportunity
Draft source-linked themes and validation questions · Human review · Deterministic calculations remain authoritative
Draft source-linked themes and validation questions · Human review · Deterministic calculations remain authoritative
People review & decide
Product, business and engineering owners review evidence and tradeoffs.
Deliver through existing controls
Use the authorized backlog, change process and acceptance criteria.
↺ Measure and feed learning backCompare adoption, contribution, recurrence and effort with the baseline
Illustrative implementation path. Products are possible sources, not connected integrations or required purchases. Availability depends on edition, configuration and access. This demo uses static fictional data; AI and a new platform purchase are not prerequisites for the method.
First pilot
Approved exports, a field-mapping worksheet and a reviewed issue register. Prove the decision is useful before automating.
Repeatable operation
Approved APIs or existing pipelines, incremental refresh, validation, traceability and visible freshness with named owners.
What needs to be agreed first?
- One service and decision, with a named owner.
- Source ID, occurrence date, issue/service mapping, severity and reviewed incident identity.
- Who supplies and reviews costs, customer exposure and delivery estimates.
- How fresh the evidence needs to be, who reviews priorities and who measures outcomes.
The demo supplies shared event references. Real systems may need a human-reviewed mapping. Similar titles alone do not establish duplicate incidents. Unknown cost or risk is unknown, not zero; ticket counts do not establish loss probabilities.
Agree approved fields, source permissions and retention with the data owners. Use the minimum relevant information. A discovery conversation does not require sending raw company records or credentials.
Technical integration notes and sources
ServiceNow REST access depends on permissions and table configuration. Zendesk offers incremental ticket exports. Jira integration requires appropriate access and handling pagination and rate limits. These are building blocks, not a universal compatibility guarantee.
03 / Establish a portfolio decision cycle
Review the portfolio. Measure what changes.
Select a stage to see the work, owners and evidence needed to move forward. Short and long term describe maturity, not promised dates.
Short term · establish and pilot
Is the evidence useful and repeatable?
Needs work → refine definitions, mapping or assumptions and review again.
Ready ↓ establish repeatable operation.
Long term · operate and improve
↺ Refine the modelCompare forecasts with outcomes and repeat the review
Accountable participants:
Output and completion condition
Agree outcomes and constraints, map current commitments and alternatives, validate benefits and capacity, compare under a common horizon, and decide through existing authority. Deliver, compare outcomes with forecasts, then revise. Evaluate the pilot before expanding. Delivery acceptance alone does not prove business benefit. Compare recurrence and effort with the baseline while considering changes in demand and reporting; before/after movement alone does not establish causation. Keep authorization in existing backlog and change processes.
Where could AI help later?
AI assistance may suggest intake links, synthesize research, draft acceptance tests or summarize outcome measurements. Each initiative identifies proposed inputs, output, reviewer and validation. Illustrated drafts have not been executed. AI cannot turn missing evidence into certainty, authorize commitments or replace deterministic financial calculations; no automatic productivity uplift is assumed.
Your next step
How do I get this into my systems?
Start with one portfolio or domain, its current commitments and the alternatives competing for resources.
Bring customer evidence into platform investment decisions, work with engineering on delivery, and review outcomes. Let’s discuss how that approach could support your platform product team.
Useful starting points: current commitments, competing priorities and the owners of commercial, operational and engineering inputs.