Disclaimer: this Post describes one way not the only way to
approach product metrics for internal teams
As our move from Project to Product gathers pace, it’s important that not only are we introducing a mindset shift and promoting different ways of working, but by doing so we also need to ensure that we are measuring things accordingly, as well as showcasing examples for others to help them on their journey. As Jon Smart points out, there is a tipping point in any approach to change where you start to cross the chasm, with people who are early/late majority wanting to see social proof of the new methods being implemented.
I’ve noticed this becoming increasingly prevalent in training sessions and general coaching conversations, with the shift away from “what does this mean?” or “so who does that role?” to questions such as “so where are we in PwC doing this?” and “do you have a PwC example?”
These are trigger points that things are probably going well, as momentum is gathering and curiosity is growing, but it’s important that you have to hand specific examples in your context to gain buy-in. If you can’t provide ready-made examples from your own organisation then it’s likely your approach to new ways of working will only go so far.
This week I’ve been experimenting around with how we measure the impact and outcomes of one of the Products I’ve taken a Product Manager role on (#EatYourOwnDogFood). Team Health Check is a web app that allows teams to conduct anonymous health checks with regards to their ways of working, using it to identify experiments they want to run to improve areas, or identify trends around things that may or may not be working for them. Our first release of the app took place in December, with some teams adopting it.
In a project model, that would be it and we’d be done, however we know that software being done is like lawn being mowed. If it’s a product, then it should be long-lived, in use and leading to better outcomes. So, with this in mind, we have to incorporate this when it comes to our product metrics we want to track.
Adoption & Usage
One of the first things to measure is adoption. I settled on three main metrics to track for this, the number of teams who have completed a team health check, adoption across different PwC territories and repeat usage by teams.
This way I can see what the adoption has been like in the UK, which is where I’m based and where it’s predominantly marketed, compared to other territories where I make people aware of it but don’t exactly exert myself in promoting it. The hypothesis being you’d expect to see mostly UK teams using it. I also then can get a sense as to the number of teams who have used it (to promote the continued investment in it) and see which teams are repeat users, which I would associate with them seeing the value in it.
Software Delivery Performance
We also want to look at technical metrics, as we want to see how we’re doing from a software delivery performance perspective. In my view, the best source for this are Software Delivery Performance metrics presented each year as part of the State of DevOps/DORA report.
I’m particularly biased towards these as they have been formulated through years of years with thousands of organisations and software professionals, with them proven directly correlate with different levels of software delivery performance. These are actually really hard to track! So I had to get a bit creative with them. For our app we have a specific task in our pipeline associated with a production deployment which thankfully has a timestamp in the Azure DevOps database, as well as a success/failure data point.
Using this we can determine two of those four metrics - Deployment Frequency (for your application how often do you deploy code to production or release it to end users) and Change Failure Rate (what percentage of changes to production or released to users result in degraded service and subsequently require remediation).
So looks like currently we’re a Medium-ish performer for Deployment Frequency / Elite performer in Change Failure Rate, which is ok for what the app is, its size and its purpose. It also prompts some questions around our changes, is our batch (deployment) size too big? Should we in fact be doing even smaller changes more frequently? If we did could that negatively impact change failure rate? How much would it impact it? All good, healthy questions informed by the data.
Feedback
Another important aspect to measure is feedback. The bottom section of the app has a simple Net Promoter Score style question for people completing the form, as well as an optional free text field to offer comments.
Whilst the majority of people leave this blank, it has been useful in identifying themes for features people would like to see, which I do track in a separate page:
Looking at this actually informed our most recent May 20th release, as we revamped the UI, changing the banner image and radio button scale from three buttons to four, as well making the site mobile compatible.
I also visualise the NPS results, which proved for some interesting responses! I’d love to know what typical scores are for measuring NPS of software, but it’s fair to say it was a humbling experience once I gathered the results!
The point of course is that rather than viewing this as your failure, use it to inform what you do next and/or as a counter metric. For me, I’m pleased the adoption numbers are high, but clearly the NPS score shows we have work to do in making it a more enjoyable experience for people completing the form. Are there some hidden themes in the feedback? Are we missing something? Maybe we should do some user interviews? All good questions that the data has informed.
Cost
Finally we look at cost, which is of course extremely important. There are two elements to this, the cost of the people who build and support the software, and any associated cloud costs. At the moment we have an interim solution of an extract of peoples timesheets to give us the people costs per week, which I’ve tweaked for the purpose of this post with some dummy numbers. A gap we still have are the cloud costs, as I’m struggling to pull through the Azure costs into Power BI, but hopefully it’s just user error.
We can then use this to compare the cost vs all other aspects, justifying whether or not the software is worth the continued investment and/or meeting the needs of the organisation.
Overall the end result looks like this:
Like I said, this isn’t intended to be something prescriptive - more that it provides an example of how it can be done and how we are doing it in a particular context for a particular product.
Keen to hear the thoughts of others - what is missing? What would you like to see given the software and its purpose? Anything we should get rid of?
Leave your comments/feedback below.
