A six-question framework for technical evaluations
When introducing a new tool, programming language, or dependency into your environment, what steps do you take to evaluate it? In this article, I will walk through a six-question framework I use to make these determinations.
What problem am I trying to solve?
We all get caught up in the minutiae of the immediate problem at hand. An honest, critical assessment helps divulge broader root causes and prevents micro-optimizations.
Let's say you are experiencing issues with your configuration management system. Day-to-day operational tasks are taking longer than they should, and working with the language is difficult. A new configuration management system might alleviate these concerns, but make sure to take a broader look at this system's context. Maybe switching from virtual machines to immutable containers eases these issues and more across your environment while being an equivalent amount of work. At this point, you should explore the feasibility of more comprehensive solutions as well. You may decide that this is not a feasible project for the organization at this time due to a lack of organizational knowledge around containers, but conscientiously accepting this tradeoff allows you to put containers on a roadmap for the next quarter.
This intellectual exercise helps you drill down to the root causes and solve core issues, not the symptoms of larger problems. This is not always going to be possible, but be intentional about making this decision.
Does this tool solve that problem?
Now that we have identified the problem, it is time for critical evaluation of both ourselves and the selected tool.
A particular technology might seem appealing because it is new because you read a cool blog post about it or you want to be the one giving a conference talk. Bells and whistles can be nice, but the tool must resolve the core issues you identified in the first question.
What am I giving up?
The tool will, in fact, solve the problem, and we know we are solving the right problem, but what are the tradeoffs?
These considerations can be purely technical. Will the lack of observability tooling prevent efficient debugging in production? Does the closed-source nature of this tool make it more difficult to track down subtle bugs? Is managing yet another dependency worth the operational benefits of using this tool?
Additionally, include the larger organizational, business, and legal contexts that you operate under.
Are you giving up control of a critical business workflow to a third-party vendor? If that vendor doubles their API cost, is that something that your organization can afford and is willing to accept? Are you comfortable with closed-source tooling handling a sensitive bit of proprietary information? Does the software licensing make this difficult to use commercially?
While not simple questions to answer, taking the time to evaluate this upfront will save you a lot of pain later on.
Is the project or vendor healthy?
This question comes with the addendum "for the balance of your requirements." If you only need a tool to get your team over a four to six-month hump until Project X is complete, this question becomes less important. If this is a multi-year commitment and the tool drives a critical business workflow, this is a concern.
When going through this step, make use of all available resources. If the solution is open source, look through the commit history, mailing lists, and forum discussions about that software. Does the community seem to communicate effectively and work well together, or are there obvious rifts between community members? If part of what you are purchasing is a support contract, use that support during the proof-of-concept phase. Does it live up to your expectations? Is the quality of support worth the cost?
Make sure you take a step beyond GitHub stars and forks when evaluating open source tools as well. Something might hit the front page of a news aggregator and receive attention for a few days, but a deeper look might reveal that only a couple of core developers are actually working on a project, and they've had difficulty finding outside contributions. Maybe a tool is open source, but a corporate-funded team drives core development, and support will likely cease if that organization abandons the project. Perhaps the API has changed every six months, causing a lot of pain for folks who have adopted earlier versions.
What are the risks?
As a technologist, you understand that nothing ever goes as planned. Networks go down, drives fail, servers reboot, rows in the data center lose power, entire AWS regions become inaccessible, or BGP hijacks re-route hundreds of terabytes of Internet traffic.
Ask yourself how this tooling could fail and what the impact would be. If you are adding a security vendor product to your CI/CD pipeline, what happens if the vendor goes down?
This brings up both technical and business considerations. Do the CI/CD pipelines simply time out because they can't reach the vendor, or do you have it "fail open" and allow the pipeline to complete with a warning? This is a technical problem but ultimately a business decision. Are you willing to go to production with a change that has bypassed the security scanning in this scenario?
Obviously, this task becomes more difficult as we increase the complexity of the system. Thankfully, sites like (k8s.af)[https://k8s.af] consolidate example outage scenarios. These public postmortems are very helpful for understanding how a piece of software can fail and how to plan for that scenario.
What are the costs?
The primary considerations here are employee time and, if applicable, vendor cost. Is that SaaS app cheaper than more headcount? If you save each developer on the team two hours a day with that new CI/CD tool, does it pay for itself over the next fiscal year?
Granted, not everything has to be a cost-saving proposition. Maybe it won't be cost-neutral if you save the dev team a couple of hours a day, but you're removing a huge blocker in their daily workflow, and they would be much happier for it. That happiness is likely worth the financial cost. Onboarding new developers is costly, so don't underestimate the value of increased retention when making these calculations.
I hope you've found this framework insightful, and I encourage you to incorporate it into your own decision-making processes. There is no one-size-fits-all framework that works for every decision. Don't forget that, sometimes, you might need to go with your gut and make a judgment call. However, having a standardized process like this will help differentiate between those times when you can critically analyze a decision and when you need to make that leap.