Build or Buy
The decision to build a solution vs. buy one is a point of some
contention. There are so many factors that serve as inputs to that
decision - which, themselves, are contentious - making consistent
decision making elusive. Over the years, here’s the set of questions
I’ve come to ask myself when making this decision. This is divided into
two sections. First the question of IF then the question of
HOW
Deciding IF you should build or buy
- Is there anyone in the company who has the time and skill to create
the solution right now?- Yes: Next Question
- No: Buy it
- Is there anyone in the company who can, over time, fix issues that
crop up with a built solution- Yes: Next Question
- No: Buy it
- Is there an existing product that perfectly addresses the problem
being solved and is it affordable?- Yes: Buy it
- No: Next question
- Is the problem core to the primary function of the company? (In
other words, is there a significant competitive advantage to solving
this problem better than competitors)- Yes: Build it
- No: Buy it
So you’ve decided to
Build - now what?
This part is as important as the first decision. It could make the
difference between a long-lived, quality solution and a solution that
imposes a burden on the company for the long term. The goal when
building a solution is to avoid “Zombie Tools” - these are tools that
are barely functional and useful but just enough that it is not possible
to abandon without a significant switching cost. Zombie fans will
appreciate that, like zombies, these solutions are hard to kill.
Here are some tenets that I keep in mind if I’m building an internal
tool.
Use Existing
Company Tools As Much As Possible
If there are tools that are currently being used that provide
some aspect of the desired functionality, use them. This is
only possible if the tools expose an API or some other form of
interoperability that allows for programmatic access to its
functionality.
This is a big one - because it could mean that you are only building
glue code that provides a purpose-built interface to existing
functionality.
Ultimately, if you can create a solution without persisting data in a
new data store, then you have created a much more
maintainable solution.
Using Zapier, IFTTT, Open Source
There are solutions that are purpose-built to be
glue. Zapier is a great example of that. And while it
does cost money, it’s very versatile and can be a single tool providing
multiple solutions. This is especially true if your company has existing
tools that provide some (or most) of the functionality you’re looking
for but falls short of the full solution.
A common example of this is that you have a workflow tool (like Jira)
but need to provide an interface to it that is not as complex and closer
to the “customer” - like Slack. While Jira has a Slack interface it’s
not a great integration (right now). Zapier can help create an alternate
interface to a Jira project that is hyper-focused on the need.
Document Document Document
I cannot stress how important this is. Keep in mind that what is
being created is likely being built by someone as a part-time gig
(probably). The tempation is to build fast and build without thought to
the future. This is a terrible instinct in this circumstance - in fact,
it is these types of projects that need more documentation.
Here’s why: It is likely that whoever works on this next will have as
much time or even less time to maintain or enhance your solution. That
means that the more churn you can remove from their process the
better.
Zombie tools are often very poorly documented. They have gotten to
the point where they work but folks will avoid touching them because
they don’t understand how they work. They will make the decision that
whatever needed to be added or fixed is not worth the risk. That is
probably a good decision. So. Give the gift of confidence to the
future maintainer and document everything about it. At a minimum
document these two things:
- How to get started developing locally (assume that the
maintainer knows nothing about the technology used to
build it) - How to deploy into the production environment (again,
assume that they know nothing about the deployment pipeline - it’s very
possible that however this is being deployed is no longer the way
everything else is being deployed since it was not migrated along with
everything else.)
Unless this tool gets its own team with product manager and dedicated
engineering, this will likely be something that the company
wants to stop supporting in the future. Don’t be offended -
it’s nothing personal but it’s probably not how the company makes money
(directly) so its support will waiver depending on the leader at the
time. It’s natural and understandable.
So that said, help leadership understand why it was created in the
first place and how it can be deprecated. This will do two things:
- Help you justify why this needed to be built
- Help leadership work toward an environment where this isn’t
necessary - either by keeping an eye out for a buy-able solution and/or
putting aside budget for buying it.
There are a few reasons why a tool could/should be deprecated: 1. The
problem it was solving is no longer a real problem 2. The nature of the
problem has changed and there is now a buy-able solution and a budget.
3. The burden of maintenance turned out to be much higher than expected
(this could justify a rewrite - not necessarily deprecation)
Some Other Considerations
Cost of Bought Solution
When buying something, remember that this does NOT mean zero
maintainenace burden. Remember that a bought solution that doesn’t quite
fit the problem will need customization and that customization can
sometimes be messy and difficult to maintain. To the point where it
might be easier to build something that is a better solution and only
slightly more burdensome from a maintenance perspective. In other words,
make sure when buying something that it truly solves the problem (and
remember that a salesman will almost always say “Yes, we can do
that”.)
Test the Solution Before
Buying
As an extension of the previous point, be sure and trial the
solution. DO NOT TRUST SALES REPS about what the product can do. Only
your own staff knows what works and doesn’t. Don’t saddle them with a
crappy solution because some checkboxes looked checked.
It Always Takes Longer To Build Than Planned
Most engineers gravitate to problems like moths to flame - I am
guilty of this. Licking their lips to find an elegant solution to the
problem, they will often say “Oh, I can finish that in a week”. This is
almost certainly not true. It’s not that they’re lying - it’s a case of
their eyes being bigger than their stomach (so to speak). There are
complexities in every solution that are not apparent at first. More
experienced engineers will likely understand this and temper their
responses.
When Building, Do Not Create a Perfect Solution Immediately
Well, you can’t. As with any project, do not plan six months to a
year ahead of time before releasing a solution. Get something out there
quickly and test with a subset of future users. This will help with a
couple of things:
- Quickly determine if a solution is viable.
- Allow the engineer to focus on the features that are actually
needed, not the ones that were thought to be needed.
Conclusion
All of this is to say that the decision to build or buy is highly
specific to a company’s situation, the nature of the problem being
solved and the people available to support the decision. Most people
understand this but how to proceed is not well understood. Use this
article to help guide the decision-making process and remember to make
as few assumptions as possible.