‹ Hawkins.io

Accelerate

The Science of Lean Software and DevOps: Building and Scaling High Performing Technology Organizations

By Nicole Forsgren, Jez Humble, Gene Kim

Published: Apr 2018
Page Updated: Mar 2021

I was excited when Accelerates was released. Their last book, The DevOps Handbook, laid out a repeatable path for moving towards the DevOps value stream.

I didn’t know what expect from Accelerate, but I hoped the authors would delight me again. This review includes my thoughts, highlights, and lastly a recommendation on who it’s best for. Skip to the end if you’re just interested in the recommendation.

Background

There has been hand-wavy correlation between DevOps and high performance that tends to come across as marketing fluff. Accelerate’s goal is to replace the fluff with facts by demonstrating the science behind building high performing IT organizations.

High performing means delivering more value more quickly. Or in other words, the ability to ship higher quality changes more frequently. The book covers how teams achieve this from commit to production. The entire software development process is omitted from the book.

Part 1 covers findings. Part 2 documents the science. Part 3 is a case study in transformation from ING. There’a also summaries in the appendixes.

Thoughts

I read Accelerate after the DevOps Handbook. Accelerate felt like a large appendix to the DevOps Handbook. This isn’t surprising given it’s the same authors discussing the same topic with roughly the same data. Generally speaking part one did not contain new conclusions or ground breaking revelations. It felt like Accelerate provided more ammunition for proponents of DevOps practices.

Accelerate puts a finger on measuring software delivery performance. This a fuzzy problem since feel like part of a high performing, but it’s challenging to compare performance across teams for a multitude of reasons. Their measurement works around these problem with directly observable metrics and inferred metrics. They break software delivery performance into four metrics:

  1. Lead Time: time it takes to go from a customer making a request to the request being satisfied.
  2. Deployment Frequency: frequency as a proxy for batch size since it is easy to measure and typically has low variability. In other words: smaller batches correlates with higher deploy frequency.
  3. Mean Time to Restore (MTTR): given software failures are expected, it makes more sense to measure how quickly teams recover from failure.
  4. Change Fail Percentage: a proxy measure for quality throughout the process

I bet it’s intuitive what are better values for each of these metrics. Obviously high performing teams satisfy customer requests faster, while doing it more often, recover from failure faster, and fail less. Accelerate provides values for high, medium, and low performers across all the metrics.

That set of metrics and the discussion around them was worth the cost of admission for me. I’ve shared the metrics but not the discussion of why and how they relate to each other — that’s what the book is for and the author’s do a much better job than I could.

Accelerate covers more than those four metrics. I found the organization structure analysis interesting, especially since it was measurable and correlated with other positive indicators. There’s also information on leadership and employee satisfaction. I didn’t expect the discussion of employee satisfaction, but the data shows that good values on these metrics correlate with less burnout. Anyone (my self included) who’s been through burnout knows that it’s horrible and should be avoided.

I struggled finding satisfaction in my work as an engineering and also creating a fulfilling work environment for my team as an engineering manager. My operating mode was try to ship useful everyday. That way you accomplish something each useful each day. Accelerate’s findings corroborate my approach. It’s nice to know that there’s a better long term solution than providing pizza and beer.

Accelerate addresses the leadership gap in the DevOps Handbook. I didn’t recognize the gap until reading other material. The DevOps Handbook doesn’t address the importance of leadership and organizational support. I was pleasantly surprised to see this discussed, with backing data, in Accelerate.

Highlights & Take-aways

This huge increase in responsiveness does not come at a cost in stability, since these organizations find their updates cause failures at a fraction of the rate of their less-performing peers, and these failures are usually fixed within the hour. Their evidence refutes the bimodal IT notion that you have to choose between speed and stability — instead, speed depends on stability, so good IT practices give you both.

So finally, once and for all, we can all say that you don’t have choose between speed and quality. This is just more ammunition in the fight for moving organizations toward the DevOps value stream. Please take this to your team if you’re still in this fight.

The moral of the story, borne out in the data, is this: improvements in software delivery are possible for every team and in every company, as long as leadership provides consistent support — including time, actions, and resources — demonstrating a true commitment to improvement, and as long as team members commit themselves to the work.

Improving the four metrics is possible for all IT companies including those “stubborn enterprises”. There’s more on this in the book. There’s no technical show stoppers, only people.

It’s interesting to note that having automated tests primarily created and maintained either by QA or an outsourced party is not correlated with IT performance.

Take this one to your team if you’re fighting this battle. I’ve experience this one first hand, so now there’s data on my side. There’s another related point in the book that having an external change approvable board (CAB) negatively correlates with IT performance. The point is things you may think are improving performance are at best creating toil and at worst hurting performance.

It’s possible to achieve these characteristics even with packaged software and “legacy” mainframe systems — and, conversely, employing the latest whizzy microservices architecture deployed on containers is no guarantee of higher performance if you ignore these characteristics.

This passage gave me chuckle (besides when Martin Fowler wrote “bullshit masquerading as science” in the foreword) since teams adopt technologies or architectures to achieve IT performance. In fact, there are no findings of correlations between specific technologies and IT performance. Practices are all the matter which are orthogonal to technologies.

Recommendation

Accelerate is an easy read which you can get through in 4–5 hours. Reading the figures and tables on my Kindle was frustrating to the point that switched to Kindle for MacOS instead. You can discern them on the Kindle but it’s much easier on a better screen. The tables and figures are discussed in the text so you can (like I did) make do without seeing them clearly.

I recommend this book to decision makers wondering how to improve their organization’s IT performance. If you enjoy statistics and correlations then Accelerate makes a strong case for improving IT performance by applying technical and organizational practices. However you’ll be left wanting if you prefer case studies and personal experience. Granted part 3 is dedicated to this, but it’s not interwoven into the findings.

I also recommend this book to anyone trying to convince their leadership to try DevOps practices. The data makes a strong case. Also the gap between high and low performers is widening because the high performance get better and low performers don’t. Push this book up the chain of command sooner than later.

I do not recommend this book if you’ve already decided to move towards DevOps and need a guide on how to get there. The DevOps Handbook is far better suited and draws upon the same data with more first hand experience and implementation patterns.

You’ll probably enjoy Accelerate if you’re passionate software engineering and curious on how to improve the process.