Josh Arnold operates on both the macro and the micro, seeing both the org-level dynamics and the team-level decision-making on which any company ultimately depends. He has a certain level of renown for his championing of Cost of Delay and CD3 (Cost of Delay Divided by Duration), which serve as a catalyst for teams to more deeply consider their prioritization decisions. CoD and CD3 add a time/value element which is sorely missing in simplified value/effort models.
Like surfing, sometimes you don’t know until you “probe” an idea or opportunity. You just have to put yourself out there and learn the patterns, react to what you see in front of you, have a go — and take your chances when they come.Joshua Arnold
His firm, Black Swan Farming, has worked with a diverse array of companies, including: shipping conglomerate Maersk Line, travel and tourism giant TUI Group, and marketing materials purveyor Vistaprint. “In each case, it was about tilting the playing field in each organization to deliver more value, with faster flow, and better quality. What drives me is a passion for helping people and organizations realize their potential, developing products that customers love and teams love building.”
I sat down with Josh for a chat on what he sees others missing, from how to create guardrails, to assessing the roles of workers, to the role of technology, to balancing business priorities.
A thread that is woven through everything is that it’s never just one thing. “As soon as you place a target on that thing, that’s when it becomes problematic,” he reports. Josh is a big proponent, naturally, of employing CD3, but he says, “don’t follow it blindly.” “Despite what some may see as my emphasis and focus on quantifying value and various aspects of product development this is sometimes misunderstood. I’m not under any illusions that the measures or numbers on their own change a whole lot. They don’t.”
In this interview we cover:
- Balancing the Buckets — TOFU, MOFU, BOFU, and Arnold’s Law
- Short-Termism, Values, Mission, and Guardrails
- The Changing Brief for Engineers
- Programmers and Ethics
- Technology as the Forcing Function
- and Surfing
The interview has been condensed, and the comments have been edited for clarity. The source video chat can be found here.
You’ll see links to specific parts of that video in the summary below. For example:
12:10 People were afraid that Cost of Delay and CD3 would become a “GPS router” leading teams off a cliff. You need values, mission, and vision. These are your guardrails.
Balancing the Buckets — TOFU, MOFU, BOFU, and Arnold’s Law
JA: Break down the mission of a bank to four buckets: increasing revenue, protecting revenue, reducing costs, avoiding costs.
Between the sales and service, they’re getting much better at the sales side of things, and this (service) is still a big hole. This is in their blind spot. They’re using data and analytics around sales side of things now because they realize that they can really juice the top of the funnel (ToFu).
They’re using data and analytics around sales side of things now because they realize that they can really juice the top of the funnel (ToFu).
Whereas the middle of the funnel, the MoFu, and then the bottom of the funnel (BoFu) tends to be the cheaper end of it. You find lots of companies buying Adobe Campaign Suite stuff for optimizing the hell out of the top of the funnel. But then what happens after the click, the service side of things, and that experience beyond that is almost forgotten, or it’s, ” I can’t justify that in terms of my OpEx versus CapEx spend, so therefore I’m just going to focus on this part of the puzzle.”
1:55 Different types of silos: run vs. change, opex vs. capex, service vs. sales. Optimizing the top of the funnel. Juicing TOFU. Quarter to quarter thinking vs. LTV.
It is true that if you stuff the top of the funnel, then even if you have a shitty experience further down, you still get even more people coming end to end, so I don’t think that’s a wrong place to start.
It’s more a question of, “When are they going to get it that they can actually extend that through the whole thing?” And the other thing that I find, especially with banks is that there’s no real understanding of customer lifetime value. So it’s all very quarter-to-quarter, sales oriented, rather than thinking about how we should look after the customer over a longer period of time so that they are loyal to us or they do more business with us, basically.
This is where Conway’s Law doesn’t just apply to your technology. It ends up applying to your portfolio and where you place what you would call “bets.” So you end up placing bets in the places that your organizational chart encourages or discourages. And if you look at the service versus sales split, or an OpEx versus CapEx, or a run versus change sort of view of the world, then your value model tends to follow that, and therefore your portfolio tends to follow that.
I asked them, which should be reasonably attractive to banks, because they’re in the money game. “So what are you doing around the increased revenue? What are you doing to protect revenue? What are you doing to reduce costs, and what are you doing to avoid costs?” And in this view of the world, almost all of their efforts are on increasing revenue, which is this top of the funnel side of things and the change side of things, and the CapEx side of things, and the sales side of things, and anything around protecting revenue, which speaks to customer lifetime value stuff, is like a poor cousin in there.
This bucket here (protecting revenue) is severely underweighted. If the broader economy is in a contraction mode or if there’s been a big financial crisis or something like that, they’re extremely efficient at switching from increasing revenue to, “Oh, whoa, hold up. It’s a shrinking market. Let’s fix all out efforts now on reducing costs.”
I guess I’d say there’s a corollary to Conway’s Law. Let’s call it Arnold’s Law for a second, that basically your portfolio will reflect your organization.
4:54 Conway’s Law, portfolios, and placing bets. Increasing revenue vs. protecting revenue (until there’s a financial crisis, which turns your focus to reducing costs).
Short-Termism, Values, Mission, and Guardrails
JA: Increasing revenue should always be subservient to some higher order bit, to use a Jobsian phrase, and that higher order bit for me would be more mission- and vision-oriented, and it is purposely further away and aspirational in nature.
9:10 80% of companies are missing a value framework. Context will dictate your focus. Layering on Cost of Delay adds an element of time to traditional SaaS frameworks (e.g. HEART, Pirate Metrics).
That’s partially where some of the West Coast American technology companies may have gone a bit awry, in that they’ve discounted the ethics to the point where they’ve made this other thing the most important thing. It means, “We were selling loads of advertising to Russian disinfo campaigns just to make our numbers.”
Increasing revenue should always be subservient to some higher order bet, to use a Jobsian phrase, and that higher order bet for me would be more mission- and vision-oriented, and it is purposely further away and aspirational in nature.
That should not be okay. That’s where the values part comes in. You can’t oversimplify it by just saying there’s this one thing, and even the mission or the vision I’m saying needs to have belts and braces around it to basically to protect us as society.
A classic case would be in Australia recently. The Royal Commission came back with the seizure report, basically just hauling the banking industry over the coals for over incentivizing salespeople. That’s the ‘sales versus service’ trade off here. They were incentivizing the sales people to basically, in effect, mass sell products in order to make their numbers. And I think Dave Snowden, as the curmudgeon that he is, would also say that the problem with that isn’t necessarily the metric, but the target.
As soon as you place a target on that thing, that’s when it becomes problematic, because people will find interesting ways to hit a particular target. I think Troy Magennis would say that it’s good to have two or three different competing matrix so that you don’t end up completely juicing one at the expense of the other; they control one another to some extent.
14:42 Metrics vs. targets. Abuse. Perverse incentives. As soon as you place a target on the thing, you risk going wrong. Try two or three competing metrics.
I think that’s the role of the regulator, which I think has been sorely absent, and we’ve been way too subservient to the mighty market, the invisible hand of the market. And I totally believe that the market is a fabulous thing, but I also think that the regulators have a role in this, which is to basically put some guardrails around those.
The Changing Brief for Engineers
JA: I think the boundaries of what engineers have been allowed to care about have changed massively even in my career. A good analogy for this is the civil engineering world. Go back 30 years, and America was dotted with the US Army Corps of Engineers who had gone on straightening rivers here, there, and everywhere and basically turned them into these concrete straights tailor-made to funnel water from A to B.
32:45 Engineer stereotypes. The boundaries of what engineering has been allowed to care about has changed. Ethics matter.
It was never part of the brief to look after the environment because it just wasn’t considered important, or it was considered a lot less important, or we just didn’t realize the impact it was having. And so you have New Orleans, I think, to some extent as a testament to that sort of level of thinking.
You could not make it through engineering school today without having it absolutely drummed into you how important the environment is. The brief has changed. There’s no way you’d hire a civil engineer today who would come and look at your runoff problem or your flooding problem and wouldn’t consider the environmental impacts of what you’re doing.
Ethics for Programmers
JA: Now, let me bring that to more of a technology focus. I think that computer programmers, once upon a time, were never really asked to consider broader things like usability or the user experience or, dare I say it, the ethics of what they were doing.
But that’s becoming increasingly untenable. I think the next generation of developers and engineers and computer scientists are going to agitate for their permission to do those things. I think you’re getting more and more activist kind of employees. They will say “Hey, no, we’re not okay with the ethics of doing what it is that you’re asking us to do. You want us to build this thing for the military. You want to build this to build this facial recognition system to be used in this particular capacity. We’re not okay with it.”
I think the next generation of developers and engineers and computer scientists are going to agitate for their permission to do those things. I think you’re getting more and more activist kind of employees.
A Facebook engineer might say “You’re asking us as engineers to enable an ad tech system, which enables people to manipulate people’s thoughts and the way their emotions are about different things, and in order to change the way they vote, and we’re not okay with that.”
I don’t think it’s just product. I think it’s actually much broader than that, to think about the wider impacts of society. I think the sphere of influence has changed massively. And part of that is because we’ve learned. We’re learning lessons about what matters and what doesn’t.
Technology as the Forcing Function
JA: Technology is a pretty key constraint, obviously.
24:45 Traveling worldwide. Seeing different orgs at different stages. Tech is a big forcing function. Developing in production. Slightly bat-shit crazy innovations. Tooling changes the game.
I think the organizational design and the way the portfolio works, that’s another kind of key constraint in there. I think the people side of things is actually easier to change than the tech side sometimes.
I get to see lots of different organizations at different stages. The technology that is now becoming more and more mainstream is a pretty big forcing function. People are talking about Kubernetes now in a way that they weren’t, because it didn’t exist, but that sort of opens up a whole different way of working from a technical perspective that we didn’t have before.
Tools can actually play a pretty big role in driving things forward. There’s the new norm, for instance, of CICD and there’s still loads of organizations that do it badly.
Part of what drives it for me is my three guiding principles.
- What are the team doing that they can deliver value early and often?
- How are they optimizing their end-to-end flows?
- How’s that cycle time getting shorter and shorter?
The feedback loops that drive quality are also getting shorter and shorter. If you look back over the last 20 years of software, we’ve seen measurable improvements in all three of them. I think from Amplitude’s perspective, the thing that you guys are key in is the speed of feedback loops with customers and users and that sort of behavioral analytics.
The feedback loops that drive quality are also getting shorter and shorter.
So the tooling that’s now available to us and to teams to make us really closely connected to how people are responding to what we’re doing, changes the game. It really tilts the playing field in a really meaningful way. I think the tooling is a key aspect of that.
JA: I see a lot of parallels with Product Development. The two biggest parallels would be:
First, that the “return on investment” in surfing is highly non-linear and stochastic/random. You can paddle out and around at a spot for hours and the distribution of waves caught, duration of each wave and the dopamine ”return” can be very random and uneven and only obvious in hindsight.
Josh out surfing
Like surfing, sometimes you don’t know until you “probe” an idea or opportunity. You just have to put yourself out there and learn the patterns, react to what you see in front of you, have a go — and take your chances when they come.
The work we did at Maersk showed a similar power-law distribution of value in product development, with some changes worth millions per week and others merely hundreds per week, or nothing at all. Like surfing, sometimes you don’t know until you “probe” an idea or opportunity. You just have to put yourself out there and learn the patterns, react to what you see in front of you, have a go — and take your chances when they come.
And second, whilst the returns are highly random and power law distributed, in surfing you can “tilt the playing field” in your favour. Yes, to newbies or even intermediate surfers it looks chaotic and fully random, but even on the messiest stormy days there are patterns. Lots of them. Swell, tide, wind, crowd, bathymetry. And if you have good fitness and flexibility and knowledge of all of the patterns of weight distribution and transferring weight, compression and extension, and a thousand other variables you can make it look easy. And a lot of that knowledge is tacit. I don’t even fully know what I’m doing out there much of the time. But the more I do it, the more I learn.
So yes, in many ways, learning to surf is very similar to learning how to develop products that customers love and (just as important) developing teams.