Cetrix SA All articles
Enterprise Technology

The Cheap Build Isn't Cheap: Scoring Maintainability Before You Sign

Cetrix SA
The Cheap Build Isn't Cheap: Scoring Maintainability Before You Sign

Photo by Photo by Vitaly Gariev on Unsplash on Unsplash

There is a particular kind of enterprise system that everyone in the organization knows about and nobody wants to touch. It works, mostly. It was delivered on time and under budget. It has been running for nine years, and the only person who understood its integration layer left in 2021. Every change request takes four times longer than it should, and every quarterly planning cycle quietly routes around it.

That system was, at the moment of purchase, the cheapest option on the table. This is not a coincidence. It is a predictable outcome of how build decisions are scored.

The quote travels well. The consequences do not.

Initial build quotes are comparable because they describe the same thing: a scope, a timeline, a price, a team. They are seductive for exactly that reason. A chief financial officer can put three numbers side by side and pick the lowest, and the decision will look disciplined in the board pack.

What the quote does not carry is the shape of the system it produces. Two vendors can deliver the same feature set at the same price and leave behind artifacts of radically different durability. One leaves a codebase with tests, coherent boundaries, and a deployment pipeline. The other leaves a working demonstration of what the requirements said in March.

The difference does not appear in year one. It appears in year three, when the business wants a change the architecture did not anticipate, and the estimate comes back at six figures because the only viable path is a rewrite of a component nobody documented.

Where the money actually hides

Total cost of ownership in enterprise software is dominated by change, not by construction. Industry benchmarks have said this for decades and it remains stubbornly true: the majority of a system's lifetime cost arrives after go-live.

For a system built cheaply, that post-launch cost concentrates in four places.

Undocumented customization is the first. Every deviation from the standard configuration is a future puzzle. When the vendor's next major version arrives, someone must determine which of forty bespoke modifications still matter. Without a decision log, that determination is archaeology.

Brittle integration is the second. Integrations assembled quickly tend to assume the other system's behavior rather than contract for it. They fail in ways that are hard to reproduce and expensive to diagnose, usually at the worst possible moment.

Scarce skills are the third and most expensive. A codebase that only one or two people understand creates a labor market of its own. You cannot negotiate with a single point of failure. You pay what it costs, or you wait.

Deferred modernization is the fourth. Systems that are too expensive to change are, by definition, not being changed. They fall behind on security patches, on platform support, and on the integrations the business now needs.

A maintainability scorecard

The remedy is not complicated, though it requires discipline in the procurement phase when everyone is eager to move.

Score candidates on observable evidence rather than assurances. The following five dimensions work well as a structured evaluation, weighted roughly equally.

Dimension What to inspect
Test coverage and quality Proportion of critical paths under automated test; presence of integration tests, not just unit tests
Documentation practice Whether architecture decision records exist in the repository and are dated alongside commits
Deployment reproducibility Whether a new environment can be built from source control without manual steps
Dependency hygiene Age and patch status of third-party libraries; process for reviewing upgrades
Knowledge distribution Number of contributors per module over the past twelve months; count of single-owner components

Each of these can be inspected rather than asserted. Ask for repository access during evaluation, even read-only and time-boxed. Vendors who have done the work will welcome the scrutiny because it differentiates them from competitors who have not. Vendors who resist are telling you something useful.

A practical refinement: judge the codebase by its most recent commit, not its most impressive module. Every team has one beautiful component built by an enthusiastic engineer. The question is what happens in the ordinary parts of the system on an ordinary Tuesday.

Questions that surface the truth quickly

Four questions tend to separate durable engineering organizations from efficient ones.

Ask how a production incident is diagnosed when the original author is unavailable. Ask what happens to a change request that touches the oldest module. Ask how the team decides when to take on technical debt and when to repay it. Ask who reviews pull requests on the least popular service.

The answers reveal operating habits. Habits are what you are actually buying, because the system will keep changing long after the contract ends.

Clauses that keep a bargain from becoming a liability

Contractual protection is possible and rarely requested. Three provisions are worth the negotiation time.

A maintainability warranty, tied to defined artifacts at delivery: documentation in the repository, a deployment pipeline, test coverage thresholds for critical paths, and a dependency inventory. Payment milestones should attach to these, not only to functional acceptance.

A remediation window. Ninety days after go-live during which defects traceable to undocumented or non-standard implementation are fixed at the partner's cost. This aligns incentives efficiently, because it prices the shortcut at the moment the shortcut is chosen.

A transition and knowledge clause. A defined obligation to support handover to another provider, including architecture walkthroughs and access to decision records, with a tail period. This is the clause that prevents lock-in from becoming permanent.

The discipline of looking past the number

The cheapest build is not a fraud. It is an accurate price for a narrower scope of work, where the scope was quietly defined to exclude the things that make a system ownable.

For U.S. enterprises buying engineering capacity at global scale, the correction is a matter of evaluation design. Score maintainability as a first-class criterion, inspect rather than interview, and attach commercial consequences to the artifacts you will need in year four.

The system you buy is not the one delivered at go-live. It is the one your team is still living with in 2033. Price it accordingly.

All Articles

Related Articles

When Houston Sleeps: Why Time-Zone Resilience Is the Metric Your Vendor Evaluation Is Missing

When Houston Sleeps: Why Time-Zone Resilience Is the Metric Your Vendor Evaluation Is Missing

While New York Sleeps: The Compounding Returns of a Continuous Delivery Loop

While New York Sleeps: The Compounding Returns of a Continuous Delivery Loop

Less Is More Resilient: The Engineering Philosophy of Constraint That US Enterprises Need to Adopt

Less Is More Resilient: The Engineering Philosophy of Constraint That US Enterprises Need to Adopt