In December 2022, the Consortium for Information and Software Quality (CISQ), co-sponsored by Synopsys, published its annual Cost of Poor Software Quality in the U.S. report. The headline figure was $2.41 trillion: the total cost of poor software quality across U.S. companies that year. Embedded in that number was a more specific and more actionable figure: $1.52 trillion in accumulated technical debt principal. That is the estimated cost to rework the suboptimal code that exists right now in U.S. company codebases. It was up from $1.31 trillion in the 2020 edition, a 16% increase in two years.
These are aggregate figures, and aggregate figures are easy to dismiss. What makes them relevant to an engineering leader preparing a board presentation is the per-organisation translation. McKinsey research found that enterprises typically allocate 25 to 40 percent of their IT budget to debt maintenance rather than new development. Separately, their analysis found that teams carrying high architectural debt deliver new features 40% slower than comparable teams in lower-debt environments. Developers in high-debt codebases spend approximately one third of their time on debt-related tasks rather than building new functionality.
The challenge with technical debt, and particularly with architectural debt, is that it is invisible to the people who approve remediation budgets. A DORA deployment frequency chart showing decline is visible. An architect's assessment that the service boundary between the billing domain and the identity domain is creating coupling that adds two sprints of effort to every new pricing feature. That is not visible without deliberate measurement.
The most effective approach we have seen for making the business case is a three-metric framework: feature drag, incident amplification, and integration tax.
Feature drag is the additional sprint effort required for new features due to structural constraints. You measure it by asking engineers to estimate what a feature would have taken to build if the relevant architectural constraint did not exist, then tracking the delta over six to twelve sprints. The cumulative drag is your quarterly technical debt servicing cost, expressed in engineering hours with a salary-based dollar value attached.
Incident amplification measures the blast radius of failures that propagate across poorly defined service boundaries. For each production incident, record how many services were affected beyond the one that originated the failure. A healthy microservices architecture has an amplification factor close to one. Monolithic or tightly coupled architectures routinely show factors of five to ten.
Integration tax is the bespoke engineering effort required to connect the system to new partners, data sources, or platforms. It is directly proportional to how far the system's API surface deviates from the contracts it was designed to expose.
Presenting these three metrics in a quarterly architecture review, with trend lines showing deterioration over time, changes the conversation from "we need to pay down tech debt" (abstract, easily deferred) to "our feature drag cost us 340 engineering days last quarter, the equivalent of two senior engineers doing nothing but servicing the interest on decisions made in 2019" (concrete, financeable).
Sources: CISQ "Cost of Poor Software Quality in the U.S.: A 2022 Report," December 2022 (it-cisq.org); McKinsey & Company technical debt research 2022–2024.