Screen technology alternatives before selecting a supplier

A portfolio committee often receives one polished supplier proposal without a fair comparison with internal development, partnership, process improvement, or continued current practice. The skill here is to form a **comparable option set** before discussing a price ranking. Begin with the desired outcome from lesson 1 and the dependencies from lesson 3. Give every critical requirement an observable acceptance test. “The parts status reaches the planner within five minutes in 95 of 100 trials under a specified load” is testable; “seamless integration” is not. That numerical threshold is a hypothetical service requirement, not a general technical standard.

Use two passes. The first tests non-negotiable conditions: data ownership, export, the organization's minimum security requirements, its applicable contractual obligations, and connection to its request register. A pleasant interface cannot compensate for failure of a genuinely mandatory condition. The second pass compares surviving options on whole-life cost, time to learn, flexibility, dependence on one supplier, and switching risk. Record confidence beside every estimate; an initial quote is not a confirmed operating cost. When evidence about a decisive condition is missing, a proof of concept or a short integration trial can be a better decision than a fabricated final ranking.

PMI's public portfolio standard description supports aligning and grouping components around strategic objectives. It does not prescribe the fictional two-year horizon, these thresholds, or the vendor choice. Those are authored decision examples. [PMI, The Standard for Portfolio Management, 4th ed.](https://www.pmi.org/standards/for-portfolio-management).

**Sensitivity and lock-in.** In the simplified comparison, B's two-year outlay equals 20,000 plus twice the annual license. It reaches C's 99,000 when the annual license is 39,500: (99,000−20,000)/2 = 39,500. That does not prove B is preferable below this figure; capabilities, timing, and risk still differ. It identifies a negotiation and evidence question: does the license rise with users or data volume, and can the team export usable records when leaving? If the annual price rises to 36,000, two-year outlay becomes 92,000, only 7,000 below C before switching cost. A possible 15,000 exit cost could reverse the cash ranking. Because exit is conditional, label it as a scenario rather than silently adding it to a certain bill. Present the base and stressed cases instead of one falsely precise number.

Ask a vendor for evidence, not reassurance: a sample export, documented fields, retrieval time, and a worked termination or deletion path. Test import into an isolated environment before marking “export available” as demonstrated. Data files that cannot recover an initiative's identifier or decision history may satisfy a superficial file-transfer check while failing the operating need. Assign technical and operational acceptance owners. An interface that works in a laboratory does not show that a planner can use its alerts under service pressure. Finally, record what the option comparison omitted, such as another supplier or the strategic value of an internal data foundation, so the committee understands the limits of this trial decision.

**A comparison that does not reward missing data.** Give every option a “must prove before contract” line. In the earlier case, A must demonstrate event export; B needs a live integration and exit test; C needs maintenance and staffing estimates; D needs a measure of continued delay cost. If one cost column holds verified whole-life amounts and another omits operation, a mechanical ranking is misleading. Show a plausible range or material unknown. D requires no new purchase but consumes service capacity or misses benefits; show that separately unless a credible cash value has been derived.

Suppose supplier B advertises a response “within one second” without saying whether the test used ten concurrent requests or one thousand. Specify neutral evidence: one hundred cases covering common and rare codes, a peak load resembling scheduling time, and a timestamp for completed reservation rather than only an interface response. For wrong cases, require a planner warning and no customer promise. Fast display cannot compensate for unsafe reservation. Request contract text giving executable export rights and a usable timetable after termination, not a sales phrase that “you own your data.”

Internal development is not free of supplier dependence either; it relies on staff skill, infrastructure, and perhaps a cloud service. If a key engineer leaves, can another person maintain the algorithm? Document tests, handover knowledge, and code ownership. Partnership can add flexibility or dependence on two parties instead of one. “Build, buy, or partner” are not inherently good or bad labels. Tie each route to the case's risks and evidence. The committee should end with a defined trial or material information request, not merely a favored product name.