Move fast and fix things
Can we work out how?
A couple of weeks ago I wrote about my new job as Chief Practices Officer at Nesta and the Behavioural Insights Team.
The main question I was grappling with at the time is essentially the core objective of my new role:
How do you create the conditions in which a range of practices — from data science to qual/quant research to design and the arts — can collaborate to solve social problems?
In the post I shared an analogy between digital products and social missions. I argued that people who work in a mission-oriented way — on complex problems like climate change or obesity — might find it helpful to apply the methods used in product work.
In response to the post I got lots of thought-provoking feedback. Broadly speaking, people liked the product-mission analogy but people also pointed out differences between the two.
The more I thought about this, the more I thought of that big tech mantra, ‘move fast and break things’, but also its counterpoint in responsible technology work: ‘move slow and fix things’.
It made me wonder if the tensions between products and missions come down to the trade-off between pace and responsibility. Do we have to move slowly if we’re trying to fix big social problems? Or can we move fast, learning as we go?
So, in the spirit of thinking out loud, here are three reflections on that theme, unpacking some differences between missions and products, before wrapping up with a nod to Hannah Arendt.
1. How do we iterate when feedback is slow?
One key difference between missions and products is that with missions the feedback cycles tend to be slow. If you invest in a new dataset, or a parenting intervention, the impact can take years to show.
This is tricky because social issues like obesity and inequalities in early years outcomes sit in very complex environments, so the basic rationale for iterative work is there — you have no choice but to learn as you go.
A big challenge for mission work is therefore how to work iteratively with slow feedback loops. So how can we do that?
Part of the answer is simple: try to develop intermediate outcomes — identify the leading indicators of the change you ultimately want to see, and then aim at these outcomes along the way.
In a similar spirit, you can break up the work you’re doing into smaller chunks. If your goal is to reformulate foods to be less calorific, you might prototype this with one food type as a proof of concept.
You can also apply a principle from product work: test your riskiest assumption first. With Nesta’s work on heat pumps, for example, we’re exploring whether community bulk buy schemes feel like plausible routes to scale, because right now this feels really unclear.
One of our board members, Sarah Hunter, shared a metaphor they use at Google for this: if someone tasked you with getting a monkey to recite Shakespeare from a pedestal, you wouldn’t start by building the pedestal. You’d start by seeing if the monkey was up to the task.
Another approach is to think about metrics in a less technocratic way. I often find myself thinking back to this post by John Cutler, which speaks to how you use metrics in complex environments. Cutler argues that looser webs of metrics can be better for alignment than hierarchical cascades of goals.
I also wonder if we need to be open to subjective ways of assessing our work. What we might call ‘gut checks’. Maybe we can ask questions like:
- In my gut, does this work feel like it could be a game changer?
- Can I see a plausible route by which this work could start a chain reaction, scaling to effect change in the whole system?
A lot of this is about acknowledging the uncertainty of mission work, equipping teams to place bets and to pivot fast in the absence of clarity, dialling up work that is promising and dialling down work that isn’t.
2. Catalysing change
Another distinctive feature of mission work is that it’s often about catalysing a big change or transition in a complicated system.
On social issues this often has the quality of requiring a paradigm shift from one type of system to another, or from one mindset to another, which is an especially hard thing to do.
With net zero, for example, we need to go from an energy system that’s unsustainable to one that’s not, and this requires that the system work to a new logic (generation might be more intermittent, for example, or more distributed, both of which change the type of systems we need).
With obesity, similarly, we need to move from a food environment that’s obesogenic to one that’s conducive to healthy behaviours, which again requires more than tweaks — we need systems-level change.
The word I like here is ‘catalyse’ because it speaks to how any actor in the system is tiny compared to the system as a whole. The only hope of effecting change at the level required is to start some kind of chain reaction, or to help the system reach a positive tipping point.
When we’re doing public policy in a market economy, we can often be more specific than this. The tipping point is often economic in nature. We want to reach the point at which a certain technology or behaviour is profitable or valuable for the typical person.
With heat pumps, for example, we need to get quickly to the point at which installing— and selling— a heat pump is the easy, obvious choice.
That doesn’t mean it’s all about economics. If there’s one thing Elon Musk got right it’s that the main thing holding back electric cars a few years ago was that everyone thought they were lame.
What does this mean for how we should work? For me it goes back to not being too technocratic, or assuming you have control. A system transition can’t be planned and implemented as if you’re building a bridge — you need to learn how the system behaves, having a prod and a poke.
This also speaks to how you define success, appraise options, and evaluate your work. In a sense, when you’re trying to catalyse a transition, it’s not very helpful to ask ‘will this work?’. The honest answer is: we don’t know.
And even when you’re evaluating your efforts afterwards, it’s still not all that helpful to ask: ‘did that work?’ This would be like evaluating whether or not you were right to back a horse in a race based on whether it won.
Colleagues at Nesta have explored these questions in some depth. One conclusion is that, when you’re trying to catalyse systems change, questions of impact — and appraisal and evaluation — are often less about distinguishing failure from success and more about minimising the chance that you’ll fail badly versus failing well.
To fail badly is to know that your mission depends on a far-reaching systems-level transformation, and yet do work that is inherently linear or incremental in its impact.
Or you can fail badly by carrying on with a piece of work when you know, in your gut, that there’s no plausible route to scale. (Or, indeed, you can fail badly by not acting quickly enough to ramp up a piece of work that shows huge promise.)
Last point here: this all ties back to the need for collaboration. Because big social tipping points are rarely (maybe never?) reached thanks to one factor alone.
The decision about whether or not to install a heat pump is a question of economics: how much do they cost? But it’s also a question of technology: are heat pumps small, quiet, and effective? And it’s also (maybe even more so?) a question of service and product design: are heat pumps appealing and are they easy to fit to my house?
In this sense, ‘tipping point’ is the wrong metaphor — it’s more like a combination lock, needing several elements to fall into place.
3. Agile policy work
Finally, systems tipping points — or combination locks — depend a lot on the policy or regulatory environment. Which raises the question: can we also influence policy and regulation in an agile way?
I sometimes hear people say no in response to this question. And one recurring theme in the feedback to my last post was people saying ‘policy influence isn’t a product’, and it can’t be done in an iterative way.
I can see the point here, but I don’t entirely agree. This is a topic I’ve written about before, so I won’t repeat my thoughts at length.
Here’s the case for why I think agile policy work is worth trying. And here’s the more nuanced view, sharing what we learned when we tried this in my old job at Citizens Advice.
TL;DR — in policy work, you probably need a mix of agile/iterative approaches and more traditional methods. As with the ‘missions-products’ analogy more generally, you can’t just take agile methods and apply them unthinkingly; you need to adapt them too.
Why is any of this worth spending time on? Aren’t these questions of method an indulgent distraction from the substance of the work on important issues like obesity or early years learning?
My view is that questions of method deserve far more attention than they get. If we focus only on the substance of our major policy failures — on the what, rather than the how, of a policy challenge like climate change — we’re spending too long on symptoms, not causes.
In my last post I got carried away writing about carpentry and the sublime. My point was that social change is like carpentry because it marries an outcome — a change we’re making in the world, like a chair — with the application of a craft — the fine-honed techniques of woodwork.
Work like this has two types of value: an instrumental value, in that we end up with a chair, and an experiential value, in that we find fulfilment — and we learn — in the practice, or the act, of bringing a change about.
That word, ‘act’, has been on my mind lately because while I was writing this post I got a call from my partner Yunhan. She’d been reading Hannah Arendt (that’s the kind of thing she does) and she was interested in Arendt’s distinction between behaviour and action.
With apologies to Arendt scholars for mangling this, the basic distinction is that behaviour has an autonomic or drifting quality — it’s when we’re pulled along by the systems around us — whereas action is what we do when we assert our moral agency and act with intent.
If we think about the world today, the prevailing mood often seems to be one of fatalism verging on despair — a sense that the world is burning and yet people are behaving, rather than acting.
Maybe we could go so far as to say that this mood of fatalistic paralysis is part of what it feels like to live in the 21st century, in conditions of unfathomable complexity. To live in a world in which our big social problems feel too damn complex to do anything about.
It’s a situation that feels not just despairing but infuriating because, after all, we built the systems that give rise to these outcomes — in a sense we, collectively, are the systems. Yet what we see, time and again, are cases in which the sum of the choices we make is something none of us want: heating the climate, shortening our lives, hindering children’s life chances.
Which is why I’m convinced that this whole question of how we act in the world amidst complexity (or maybe how we act on the world?) is a question of first order significance. In public policy, we should spend more time thinking not about what we do, but about how we do it — from the means by which we understand problems together; to the mechanisms we use to deliberate on our options; to the ways, together, that we act.
As ever, thoughts shared in the spirit of thinking out loud. I try to follow the principle of stating my views clearly (maybe not always successfully!) and holding them loosely. So — very open to feedback and challenge.
To stay in touch with this work, you can follow me on Medium or support my writing on Substack. And for the big picture take on how we adapt the state for a digital age, there’s my book, End State.
