A score can make a messy comparison legible. The danger begins when the compressed number starts to look more objective than the judgments and tradeoffs that created it.
In the Costco Singapore market-entry project, our country-screening model produced a clean ranking: Singapore 7.50, South Korea 6.25, Japan 5.35.
That ranking was useful—but it was not the recommendation by itself.
The score compressed many different kinds of evidence
We weighted Benefits at 50%, Costs at 30%, and Risks at 20%. Underneath that were purchasing power, growth, vehicle accessibility, corruption, infrastructure, legal/business conditions, labor/property cost, inflation, property rights, and demographics.
Those inputs were not naturally commensurable. The model made comparison possible by standardizing and weighting them, but the final decimal still depended on choices about criteria, scale direction, normalization, and importance.
The top-ranked country still challenged the business model
Singapore’s purchasing power and growth were attractive. But low vehicle ownership and higher labor/property costs conflict with assumptions embedded in Costco’s classic bulk-warehouse operating model.
That meant the recommendation could not stop at “Singapore wins.” The next question was what had to change—site format, accessibility, product mix, membership pricing, logistics, or operating assumptions—to make the model fit the market.
Weights are governance choices
A 50/30/20 Benefits–Costs–Risks split is not a discovered law. It is an explicit statement about what the decision-maker values. That is a strength when the weights are visible and challengeable; it becomes a weakness when they are presented as objective truth.
Decision models should expose tradeoffs, not hide them
A weighted score is valuable because it structures comparison. It becomes dangerous when the final number is treated as self-executing. The point of the tool was to focus attention on the tradeoffs that needed operational adaptation.
This lesson became even more important in my later research
In the Monte Carlo project, I carried this distinction forward explicitly. AHP weights represented strategic importance; they did not become NPV coefficients. Country variables affected value only through separate economic transmission channels.
That separation prevents a subtle modeling error: turning “management cares a lot about this factor” into “this factor must have a proportionally large causal effect on valuation.” Those are different claims.