InsightsPoint of View

The Economics of Adaptability

Why the cost of changing a system belongs in the original investment decision

A large auction yard with premium agricultural equipment, construction loaders, and commercial vehicles arranged in distinct inventory groups.
By
Cam Myers
Content type
Point of View
Reading time
6 min read
Status
Author’s perspective
Topics
Ownership & Economics

In brief

  • The investment comparison includes the cost and disruption of changing a system’s assumptions.
  • Separating responsibilities preserves options only when the connections remain dependable.
  • Flexibility, waiting, and immediate commitment each have costs that require an economic case.

The argument

The cheaper system may be the one you can afford to change.

That sounds obvious when buying equipment. An owner considers replacement parts, service access, and whether a machine can handle the next product line. Yet software decisions often focus on the launch price and the work the system will save under today's conditions. The cost of changing those conditions arrives later, as an unpleasant surprise.

A pricing change requires a development project. A new location needs a duplicate application. Replacing a supplier means rebuilding several integrations. What looked like a modest purchase has become a commitment to a particular way of operating.

The question is how much to spend preserving choices, and which choices are worth preserving. Flexibility has a cost too.

The decision underneath

If this important assumption changes, what else must change with it?

Every design commits to an assumption

Consider a hypothetical regional rental business choosing a quoting system. Today, nearly every customer rents for four weeks. The fastest build stores one four-week price in each product record and uses that figure throughout quoting, payment, and advertising measurement.

That can be a reasonable decision. But suppose the business starts serving contractors whose rentals run for uncertain periods. It now needs minimum commitments, extensions, and different prices for longer use. The original assumption about duration was embedded in several places. Changing it requires coordinated repairs, and the team must check whether old contracts still reconcile correctly.

Another design might keep contract terms and price calculation separate from the customer and booking records. It would take more effort initially. When terms changed, fewer parts would need revision. The financial difference between those choices includes the cost and disruption of making the change, not just the original programming bill.

In software design, David Parnas made an influential argument for organizing components around decisions likely to change, rather than simply dividing up the steps in a process.[1] The useful management implication is practical: identify an uncertain decision before allowing the entire system to depend on its current answer.

This does not require predicting exactly how the business will evolve. It requires recognizing where today's certainty is weak. A customer must remain identifiable across transactions. A pricing policy may change several times a year. Those two requirements do not need the same degree of flexibility.

Preserve the connection that lets a part change

In work with a regional auction operator in the northeastern United States, I separated an advertising classification requirement from the internal inventory taxonomy. Product descriptions supplied information that the broad internal categories did not preserve. A focused application interpreted that information and supplied classifications through a supplemental feed.

The immediate benefit was a route to the required output without making a larger taxonomy initiative a prerequisite. The longer-term design question concerns how much of that route must remain fixed.

The classifier could change because a different model becomes more suitable, costs change, or the advertising platform revises its requirements. Those changes need not require a new receiving procedure or a different way for catalogers to identify inventory. Keeping the classification function separate creates room to evaluate alternatives without reopening every surrounding process.

That room only exists if the connections remain dependable. The business needs consistent product identifiers, a defined output, and a way to check that an assignment describes the correct item. Separate applications with unclear connections can be harder to change than one well-designed system.

The distinction is between separating responsibilities and scattering them. A part is replaceable when another part can perform the required function without forcing the rest of the business to learn its internal workings. Testing, migration, and support still cost money.

This is especially relevant when considering AI. Model availability and performance can change. The durable investment may be the product information, evaluation method, business rules, and integration surrounding the model. Treating one model's current behavior as a permanent foundation can make later improvement unnecessarily expensive.

Flexibility needs an economic case

It is easy to take this argument too far and build elaborate systems for possibilities that never arrive. Every extra interface, configuration option, and supported alternative creates work. A small distributor should not finance a technology estate sized for an imagined multinational operation.

The case for adaptability depends on the consequence of a plausible change, the cost of preparing for it, and how long the business expects to use the system. An uncertain label on an internal report may be cheap to revise later. A choice affecting every customer identifier can be difficult to unwind. Giving both equal attention wastes resources.

For illustration, suppose separating a pricing function adds $6,000 to a build. A plausible change in contract terms could otherwise cost $25,000 in rebuilding, testing, and transition work. Those figures alone do not establish which design is better. If the change is unlikely within the useful life of the system, the simpler build may win. If it is already under discussion and would interrupt bookings, preserving the option becomes more attractive. These are hypothetical amounts, not a reported project return.

The investment also depends on when the business can learn. Robert Pindyck's work on irreversible investment shows why the ability to wait for information can have economic value when a commitment cannot easily be recovered.[2] Applied to an operating project, a limited pilot may let the business learn about customer behavior before committing the entire workflow.

Waiting carries costs as well. A company may forgo useful cash flow, remain burdened by manual work, or miss a market opportunity. Sometimes the sensible choice is to commit now and accept a future replacement bill. The important discipline is to make that tradeoff visible rather than treating either flexibility or speed as an unquestioned virtue.

Being able to change is different from staying available

A system can survive an outage and still be difficult to adapt. Backups and recovery procedures help restore operation. They do not necessarily make it economical to introduce a new service, revise a decision rule, or replace a vendor.

The reverse is also possible. A replaceable application can still fail during a busy trading period. Adaptability and dependable operation deserve separate consideration, even when some design choices help both.

There are limits to reversibility too. Rolling back software does not undo a price already accepted by a customer, recover a wasted delivery, or restore inventory sold under the wrong terms. Preserving technical choices is useful only if the business can also manage the consequences of using them. A small release may limit exposure; it cannot make every action reversible.

For an owner or executive, the revealing question is specific: if this important assumption changes, what else must change with it? Ask the team to trace the answer through records, customer commitments, staff work, interfaces, and costs. The resulting picture often says more about the investment than a long list of advertised features.

The goal is to make a deliberate commitment: stable where consistency creates value, changeable where uncertainty creates a material economic risk. A well-designed system lets the business benefit from what it knows today without making every future decision pay for today's assumptions.

Sources

  1. David L. Parnas. On the Criteria to Be Used in Decomposing Systems into Modules. Carnegie Mellon University technical report, 1971; published in Communications of the ACM, 1972.
  2. Robert S. Pindyck. Irreversibility, Uncertainty, and Investment. Journal of Economic Literature 29, September 1991, pages 1110–1148.

For the leadership team

In practice

  • Identify the uncertain decisions before embedding their current answers across the operation.
  • Compare preparation cost with the likelihood, timing, and consequence of a plausible change.
  • Evaluate adaptability separately from availability and from the reversibility of customer commitments.

Questions worth asking

  1. Which assumptions are stable enough to commit to, and which create material future exposure?
  2. Could a bounded pilot provide useful evidence before a difficult-to-reverse investment?

Weighing this decision now? Tell us what you are considering and we will say plainly whether we can help.

Discuss your business