Build or buy? Code or re-use?

Senior IT managers starting a new project often have to answer the question: build or buy? Meaning, should we look for a packaged solution that does mostly what we need, or should we embark on a custom software development project?

Coders and application-level programmers also face a similar problem when building a software product. To get some part of the functionality completed, should we use that framework we read about, or should we roll our own code? If we write our own code, we know we can get everything we need and nothing we don’t – but it could take a lot of time that we may not have. So, how do we decide?

Your project may (and probably does) vary, but I typically base my decision by distinguishing between infrastructure and business logic.

I consider code to be infrastructure-related if it’s related to the technology required to implement the product. On the other hand, business logic is core to the business problem being solved. It is the reason the product is being built.

Think of it this way: a completely non-technical Product Owner wouldn’t care how you solve an infrastructure issue, but would deeply care about how you implement business logic. It’s the easiest way to distinguish between the two types of problems.

Examples of infrastructure issues: do I use a relational or non-relational database? How important are ACID transactions? Which database will I use? Which transactional framework will I use?

Examples of business logic problems: how do I handle an order file sent by an external vendor if there’s an XML syntax error? How important is it to find a partial match for a record if an exact match cannot be found? How do you define partial?

Note that a business logic question could be technical in nature (XML syntax error) but how you choose to solve it is critical to the Product Owner. And a seemingly infrastructure-related question might constitute business logic – for example, if you are a database company building a new product.

After this long preamble, finally my advice: Strongly favor using existing frameworks to solve infrastructure problems, but prefer rolling your own code for business logic problems.

My rationale is simple: you are (or should be) expert in solving the business logic problems, but probably not the infrastructure problems.

If you’re working on a system to match names against a data warehouse of records, your team knows or can figure out all the details of what that involves, because that’s what the system is fundamentally all about. Your Product Owner has a product idea that includes market differentiators and intellectual property, making it very unlikely that an existing matching framework will fulfill all requirements. (If an existing framework does meet all the requirements, why is the product being developed at all?)

Secondly, the worst thing you want to do as a developer is to use an existing business logic framework “to make things simple”, find that it doesn’t handle your Product Owner’s requirements, and then start pushing back on requirements because “our technology platform doesn’t allow X or Y”. For any software developer with professional pride: I’m sorry, but that’s just weak sauce. Again, the whole point of the project is to build a unique product. If you can’t deliver that to the Product Owner, you’re not holding up your end of the bargain.

On the other hand, you are very likely not experts on transactional frameworks, message buses, XML parsing technology, or elastic cloud clusters. Oracle, Microsoft, Amazon, etc., have large expert teams and have put their own intellectual property into their products, making it highly unlikely you’ll be able to build infrastructure that works as reliably and is as bug free.

Sometimes the choice is harder. You need to validate a custom file format. Should you use an existing framework to handle validations or roll your own code? It depends. It may not even be possible to tell when the need arises. You may need to use an existing framework and see how easy it is to extend and adapt. Later, if you find you’re spending more time extending and adapting than rolling your own optimized code, you can change the implementation of your validation subsystem. Such big changes are much easier if you’ve consistently followed Agile engineering practices such as Test Driven Design.

As always, apply a fundamental Agile principle to any such decision: how can I spend my programming time generating the most business value?