Legacy modernization rarely fails because of technology. It fails because of predictable decision-making patterns that take hold long before the first line of new code is written. In this three-part series, Broadpin modernization expert Stephan La Rocca looks at the cognitive biases that derail modernization projects and how an analysis-first approach counters them. Part 1 starts at the beginning: the biases that shape where and how teams begin. The full picture is available in our whitepaper Software Modernization and Its Cognitive Biases.
Why Modernization Projects Fail Before the First Line of Code
This blog was written by Stephan La Rocca (Director of Business Development).
I spent the first half of my career writing new software. The second half has been different: modernizing the software that already exists. These two phases did not merely feel different. They really were different.
New software starts from an open space of possibilities. Modernization starts inside an existing fabric of code, data, habits, tacit knowledge, and earlier decisions. And it carries something else that greenfield projects rarely do: a long history of emotions, assumptions, and unexamined success.
That history is where most modernization projects go wrong. Not because the technology is too hard, but because teams and leaders make predictable decision-making mistakes before the architecture is even drawn. This is the first of three posts about the cognitive biases I see most often in practice.
The Turkey Illusion: "From simple to complex" is not a strategy
The turkey illusion comes from a parable by Bertrand Russell. A turkey is fed every day by the same farmer. With each feeding, its confidence grows that tomorrow will bring more of the same. The evening before slaughter is when that confidence peaks.
I see the same pattern in modernization projects. Teams start with the easiest components, gain confidence sprint by sprint, and build momentum on the assumption that the hard parts will follow the same curve. They do not. Somewhere around month four, a cluster of tightly coupled legacy logic appears that nobody analyzed upfront, and the architecture built on simple cases cannot absorb it.
When someone asks me "Where do we start?", my answer is always the same: with the cluster that gives us the most momentum for the project, not the one that feels safest. That requires knowing what the system actually contains before making the first move.
Confirmation Bias: the legacy system through a pre-colored lens
In software testing, confirmation bias is well known: teams test to prove the code works rather than to find where it breaks. In modernization, the same bias strikes in a more dangerous place. When reverse-engineering the legacy application, it is almost impossible to start with a neutral lens. Every time the thought "I knew it!" occurs while reading existing code, the team has unknowingly stopped analyzing and started confirming.
The result is that the analysis mirrors what the team expected to find rather than what is actually there. Edge cases, implicit business rules, and undocumented dependencies slip through, only to resurface months later when the new system is supposed to go live.
The Bandwagon Effect: "Everyone is using it, so it must be right"
The first question in most modernization projects is: "What will we build on?" And the decision is often made before the question has properly been asked. A technology is trending, a charismatic team lead champions it, or the market narrative says it is the obvious choice.
What gets overlooked is whether that technology actually fits the requirements of the legacy application, requirements that may not even be fully understood yet because a proper analysis has not been done. If modernization starts with a technology decision instead of a scope decision, the project is shaped by preference rather than by evidence.
The common thread
All three of these biases have the same root cause: acting before understanding. Starting with the easy parts instead of the important ones. Analyzing through assumptions instead of data. Choosing technology before scope.
The structural answer is to replace assumptions with a systematic analysis of the existing system before the first architectural decision is made. What does the legacy application actually contain? How large is it, really? Where are the dependencies? What is the implicit logic that nobody documented?
In the next post, I will look at the biases that distort estimation: why teams consistently underestimate the size, effort, and complexity of what they are trying to modernize.
Want to see the full picture?
The biases in this post are only the beginning. Our whitepaper Software Modernization and Its Cognitive Biases covers every bias that shapes how modernization projects start, estimate, and plan, and shows how an analysis-first approach counters each one.