Is 9L/100km good performance for a Hilux on dirt roads?
A recent adventure in Australia gave my family and me a life lesson in practising what I preach about software teams using a new metric.
On a three-month road trip, on any given driving day, we optimised for our time to destination, fuel-value for money, or, a new one for me - straight fuel consumption because of range anxiety⛽.
390 km from Derby in WA, in the middle of the Kimberly, staring at an L/100km number, I instantly appreciated our client's dilemma when we introduced them to flow efficiency and their % of wait time.
The number was 9. The dilemma was, 'Is that ok? What's a good L/100km? How can I improve it?' because, at 9, our estimated range was 430km, which was cutting it fine to reach Derby and the next known fuel stop 😬 (on the plus side, at least we knew our destination unlike many delivery teams).
-- cue futile-no-internet speculation on how Toyota calculates the consumption and the remaining amount of fuel --
Toyota, it turns out, did a good job. We got to a dusty Derby with 40km of range left, smiles on our dials, and an appreciation that driving with an efficiency of 9L/100km is...ok 🤷?
We jumped into a big boxy campervan a week later to drive from Broome to Perth. Heading out of town, I changed the dash display to my new friend - fuel consumption. It was 18L/100km 👀👀. Then we turned south into a savage headwind, and consumption hit 24L/100km. 9L in the Hilux was now AMAZING.
We were in rideshare in Freemantel three weeks later - one of those hybrids with all the energy info twirling on the console.
Move over Hilux; the Camry was getting 4.5L 😲
So there it was; I finally had an opinion on fuel efficiency, an L/100km range of 4.5 (desirable) to 24 (undesirable). I now use my opinion to make trade-offs while driving, responding to external forces like weather, and inform my next car purchase.
What's your percentage of wait time?
I realised that this metric learning is very similar to when software teams start using a new metric, and the most common example is when a team or organisation needs to check their flow efficiency - get to know their flow, as Jon Smart puts it.
When we start working with clients, our second favourite metric to sample is the percentage of time work waits while it goes from start to deliver to the customer. We love this number because, like blood pressure, it's a lagging health indicator of many other things in a delivery system. Several upstream activities influence flow efficiency:
Planning. Are we over/undercooking our sizing, estimation, etc?
WIP - Are we trashing too much work in progress?
Dependencies - Are we too dependent on other teams? Misaligned with change controls?
Interruptions - is operational noise or coordinating with others slowing us down?
Hidden work - is other work sneaking into our days?
Unknown capacity - Do we know our actual current capacity? People changes?
Compensating Controls: Are approval or change controls helping or hindering?
External Forces: Is our response to some external pressure like competition or regulation costing us more than we anticipated?
However, there are two problems:
🅰️ Generally speaking, nobody knows their flow percentages
🅱️ Nobody knows what's a desirable percentage - for their context
So, like fuel efficiency, we have to start by simply measuring it - which is a topic in itself. Still, at a high level:
Take a small manual sample or a more extensive export of 3+ months of work items and their status change information - depending on the level of tooling.
Interrogate the workflow steps & controls from 'let's do this' to 'customer has it' and categorise each step or stage into 'active' or 'waiting'.
Finally, talk with people to cross-check the discipline of status changes. For example, often, teams need to update items when they go on hold or get blocked. Agree on a weighting to add to the total wait time to compensate for this, e.g. 1.1
ℹ️ Top Tip: If you use Jira, some free-ish plugins make steps 1 and 2 easy to do. Look for Status Time Reports by BloomPeak and Timepiece by OBSS
You can now do a broadly right and precisely wrong calculation of total time in the Wait status group X 1.1 / Total duration spent on items. Trust me, it's worth the effort to do it at least once a quarter. You present that number to the team(s) or a cross-functional leadership group, and everyone gets a reality check. The conversations that follow are a lot more effective.
Just like I now understand what fuel efficiency means for different vehicles, your organisation can develop a nuanced understanding of flow efficiency for your unique context to help with priorities, choices, and investments - little and large.
What is desirable, for us?
On first inspection, we often find committed work waits at least 40% of the time where the team do ~14-day cycles. We've also seen it as high, at 80%, with work sitting for months on the shelf sometimes. At best, we've seen an 18% wait time in lean conditions, small batches, and continuous delivery.
Over time, the team or organisation develops a sense of what's at least better, if not suitable, flow efficiency - by considering it more deeply and sampling more teams or specific types of work. Like me jumping from a ute to a camper to a sedan, it's not a clean comparison, but you develop that opinion on a metric range from desirable to undesirable.
Relative Metric Value X Contextual Weighting = Signal Strength
This metric range is valuable for applying a weighting or care factor to the metric for your context. While many metrics can be used temporarily as a goal, getting it to a desirable point is. You need to know when it's good enough, and our key constraint(s) have shifted elsewhere. It's time to use those metrics as a guardrail.
Back on the road in Queensland, we were only an hour from a fuel stop. We'd shop around for a relatively reasonable fuel price, but, most importantly, we had to get somewhere each evening in time to settle in.
The Kimberly has few fuel stops, and you can only carry so much, so it's all about range. When you fuel, you top up - almost regardless of the price😢
The rest of WA is ridiculously vast! There were fuel stops, albeit with 300km between them, but there were 1000s of km to go, and we were driving in our home, so it was all about the economy. FYI, 94 Km/hr is the optimal compromise of fuel efficiency and time to destination for a boxy Mercedes Sprinter camper with young children on board.
The Diesel prices in the outback were up to 2.5 times more expensive than say Cairns, but with no other option, you just have to pay the price.
This process of monitoring different metrics as the situation changes is a growing part of our work with software teams. We are helping them establish a relative 0-100 index to measure Was this month better or worse than last month? The index is a suite of balanced metrics rolled into a single monthly score. We include industry measures for things like throughput, quality and reliability, organisation-specific measures for governance, product adoption, commercial feasibility, etc.
The hard part is always getting to that opinionated range for each metric. What's a suitable [% of wait time, lead time, cost to serve] for us?
Just like I now understand what fuel efficiency means for different vehicles, your organisation can develop a nuanced understanding of what flow efficiency means for your unique context. The key is to start measuring, keep monitoring, and adjust as you go."
Reach out to us to discover how we can help your team define and achieve their optimal delivery efficiency, balancing productivity with governance, product success, and team experience
Comments