The price on the proposal is not the cost of the decision

Every platform selection ends with a number. Sometimes it is the figure on the cover page of the winning proposal, sometimes an NPV in the business case that finally convinced the board. Either way, the number is exact, and exactness feels like truth. That feeling deserves a closer look, because in technology sourcing it misleads more often than it helps. Below are four ways a precise figure can hide an imprecise decision, and the questions that bring the hidden part back into view.

A calculation is only as honest as its inputs

Business cases for new commerce platforms usually lean on projected revenue: better conversion, fewer abandoned carts, bigger baskets. The projections tend to arrive from the vendor's own success stories, and those stories rarely reveal how the numbers were produced. Was there anything to compare against, or did the client simply perform better after the switch than before it? Companies typically replace platforms when results are disappointing, and disappointing results have a habit of improving no matter what you change. Some of the promised effect would have shown up anyway.

And this gap between projection and reality is anything but theoretical. When McKinsey and the University of Oxford examined more than 5,400 IT projects, they found that large technology projects run 45 percent over budget on average while delivering 56 percent less value than predicted. Read that second number again. The typical business case does not miss by a rounding error. It misses by half.

Yet feed assumptions like these into a financial model and it will happily return a figure with two decimal places. The mathematics are flawless. The conclusion is a guess wearing a suit.

The practical remedy is a strict separation. Money you will certainly spend belongs in one column, money you hope to earn in another, and the two should never be allowed to blur into a single confident total.

Razor and blades, platform edition

Some proposals from vendors with closed, exclusive systems come in strikingly cheap. That is rarely an accident. When leaving a vendor means rebuilding half your stack, the entry price can afford to be low, because the real earnings sit in the years when switching is no longer realistic. The bill for the dependency lands two or three budgets later, at a point when nobody remembers the alternatives from the original offer.

Decision makers can sense this trap, even when their templates ignore it. When Flexera asked organisations in its 2026 State of the Cloud research why they spread workloads across multiple providers, the most common answer, given by 67 percent of respondents, was avoiding vendor lock-in. So the concern is mainstream. What remains rare is translating it into a line item during vendor selection, which is exactly where it would do the most good.

For anyone who compares offers on price alone, the consequence is uncomfortable: the method itself rewards closed systems and punishes open ones, even when the open option ends up cheaper across the full lifespan of the platform.

Three questions restore the balance, and none of them appears in a standard template.

• In what format can we take our data out?
• Who owns the integration code once the project ends?
• Can you name clients who moved away from you, and how did that go?

A supplier confident in its product will answer all three. A supplier that hesitates has answered a different question, and that answer also belongs in your risk assessment.

Paying for a door you may never open

Two platforms can carry an identical five year cost and still differ in something that never shows up in a spreadsheet: what happens if the business changes direction. A new market, a marketplace model, an extra sales channel. On one architecture the turn is a project. On the other it is a rebuild.

Finance has long recognised that the ability to change course has a price of its own. People pay for options they never exercise, because holding the choice reduces risk all by itself. The same logic applies to architecture. Flexibility is worth money even in the scenario where you never use it.

There is no need to model this formally. One question per offer does the job: if we needed to turn the ship two years from now, what would that cost here? The answers will separate the offers faster than most line items in the comparison.

Fast enough to outrun its own ageing

A positive NPV tells you an investment pays off within the analysis window. It says nothing about when. And in e-commerce, when is the whole game, because platforms grow old quickly. How quickly? At the moment Digital Commerce 360 ran its survey, 27 percent of brands were already considering a platform migration, and 61 percent of those planned to carry it out within the following year. Picture that for a second: at any given time, roughly one company in four is eyeing the exit. Against that backdrop, assuming five useful years for a commerce platform starts to look generous.

Treat the technology like any other asset that loses value over time and a simple benchmark appears: something built to last five years should earn back around a fifth of its cost each year. Two offers with identical five year returns are therefore not equal. If one delivers steadily while the other saves its payoff for the final stretch, the second one is a bet that nothing important changes in the market before the money arrives. That is not a bet to make casually.

This check does not replace the financial model. It asks a different question. The model asks whether the investment pays. The benchmark asks whether it pays before it expires.

What the board actually needs

None of this means abandoning financial analysis. It means being honest about what each part of it is made of. A board presented with one blended figure cannot tell where the knowledge ends and the bet begins. A board shown certain costs, plausible estimates and hopeful projections side by side can decide, with open eyes, how much risk it wants to carry.

That distinction is the real deliverable of a good vendor comparison. Because the most expensive line in any proposal is the one that was never printed on it.