Designing Team Culture 

A leader walks into a town hall and tells the room they want everyone learning Python, that upskilling is the future, the department is going all-in on capability. Someone raises their hand. "Can we get training funding?" Leadership checks the budget. Two seats approved, 150 people waiting.

That's the friction. And that friction, repeated across enough decisions, is what culture actually is.

Most leaders don't see it that way. When people complain about culture, they reach for the hygiene factors: more communication, better coffee, birthday celebrations, quarterly get-togethers. These matter, but they're not what I mean by culture. Culture is a set of behaviours, how people come together around a common goal, and whether leadership means what it says. You can have excellent coffee and a toxic culture. I've seen it.

The mistake most organisations make is treating culture as a programme, something with milestones and colourful slide decks. Programmes feel controllable. Culture isn't controllable. It moves through repeated human behaviour, leadership signals, budget decisions, and the distance between what leaders promise and what the system actually allows.

When leadership closes that gap, behaviour changes. When they don't, you get lip service with a fancy name.

The diagnosis

In 2017, I was leading Strategy, Planning and Performance in Digital Business Services at Stats New Zealand, about 150 people across the digital technology division. Everyone was blaming culture for why teams didn't cooperate, why management behaved one way and staff another. Culture became the explanation for every symptom, which meant nobody was actually diagnosing it.

So I found a model people could understand quickly. William Schneider's culture model maps organisational behaviour across four quadrants: Control, Cultivation, Collaboration, and Competence. Simple enough for a frontline staff member to read, specific enough to point at real behaviour.

We ran a survey across the whole department. The result was clear: a dominant Control culture, with a secondary Cultivation culture sitting beside it. Those two sit directly opposite each other on the model. According to Schneider, opposite cultures are in conflict.

That Python training moment is what that conflict looks like in practice. Leadership says "develop yourself, grow, we want you to be excellent," which is Cultivation talking. The budget says "sorry, two seats," which is Control answering. Staff hears both. They weigh both. What they believe is the one that has money behind it.

Multiply that across every approval chain and governance decision, and you get a department where people have learned not to expect much.

I still remember one town hall where we asked staff openly what was getting in the way. The room went quiet. Then someone said, carefully: "customer comes first." Everyone nodded. What they actually meant was that other business units came first, and the technology teams absorbed the pressure downstream. The phrase sounded principled on paper. In practice, people experienced it as cover for decisions that had already been made above them. That one moment of careful language told me more than any survey could.

Making the gap visible

The standard approach is an annual engagement survey. Stats New Zealand, as a government department, already ran one. The number comes back, the CEO says things are improving, and nothing changes until the same survey runs twelve months later. You can't navigate with a map you only look at once a year.

We used NPS instead, adapted for employee experience. Every month, one anonymous question: "How likely would you be to recommend DBS to a qualified friend or family member as a great place to work?" Score from 0 to 10.

The key was converting the quantitative scale into something qualitative. We presented it to staff as a smiley face scale, each range corresponding to a feeling: 0 to 6 is disengaged or passive, 7 and 8 is indifferent, 9 and 10 is genuinely positive. The scale is deliberately skewed. You have to be certain of what you're feeling to hit a 9 or 10. Passivity doesn't count as engagement.

We published the results on Yammer within a week of every month-end. All staff. The bad months too.

That last part matters more than people assume. Most cultural programmes fail because staff have already survived several of them, and they know the pattern: leaders announce the initiative, run some workshops, measure activity, declare victory, and move on. By the time the next programme starts, the scepticism is already baked in.

Publishing results openly, including comments, including the months where the score dropped, was the signal that something different was happening. Nothing to hide. If something was broken, it was visible to everyone, which meant it was everyone's responsibility, not just a topic for the next leadership offsite.

We had three central line uplifts between April 2018 and February 2021: from -4% to 11%, then to 18%, then to 37.4%. Each uplift represented a new statistical baseline, the average exceeding the previous ceiling long enough to set a new floor.

What changed behaviour at each stage was simpler than any programme document would suggest. People watched leaders respond to the comments, not just the number. They saw the data used to hold leadership accountable as much as staff. I would take the results back to my fellow leaders and reflect it at them directly: here are which teams are struggling, everyone can see it, so what are we going to do? That creates a particular kind of pressure among leaders, the kind where difficult conversations start happening before they become crises.

On the staff side, I watched behaviour start shifting toward the left side of the Schneider model. People began working out loud, owning their own paths, asking questions they'd stopped asking. The environment changed because the trust changed. Those two things move together, and you can't manufacture one without the other.

What failed, and why it matters

One team read Holacracy, came to senior leadership, and we said yes, let's try it.

The idea has genuine merit: self-organising teams, distributed authority, people able to act without waiting for approval. It works somewhere. It didn't work here.

The problem was financial delegation. In a government department, spending authority is tied to your role. You can't extend it informally, and we checked. HR confirmed: 150 team members cannot have financial signing authority, even at $100. Someone is accountable for the budget, and that accountability is structural. The model couldn't survive that constraint.

We told the team plainly: this doesn't work in a government context, and we're not going to pretend otherwise.

Worth noting, though. The experiment taught us where the walls were, and that's real information. Management models designed for startups with flat structures and discretionary spending don't migrate cleanly into organisations with legislative obligations and financial controls. You can want a different culture and still operate inside a system that limits how far certain ideas can travel. Knowing that boundary saved us from chasing the model instead of the behaviour it was trying to produce.

The Covid test

Before 2020, DBS had been running flexible working quietly for a while. Two days from home, compressed hours, alternate Fridays off for some people. The Chief Digital Officer believed the trust was there, and high trust means people can work from wherever they're most effective. We were already on Microsoft Teams when most of the organisation had barely heard of it.

Other business units were unimpressed. I heard "recalcitrant" more than once. We were the department that asked too many questions, didn't follow the standard IT playbook, apparently needed managing.

Then Covid arrived.

The CEO was under pressure. Most of the organisation had no remote capability, no tested platforms, no experience running a distributed team. We had been doing it for two years. We helped onboard the rest of the organisation onto Teams, ran the virtual meeting training, and kept delivering without interruption.

That wasn't luck. The trust built over three years and the investment in tools people called unnecessary right up until the day they were essential: both paid off in a week that most organisations remember as a crisis they barely survived. Our critics went quiet. There's not much to say when a department you called recalcitrant already knows how to do the thing you're scrambling to figure out.

What the numbers confirmed

The Schneider survey in December 2020 confirmed what we had felt for months. The organisation had moved left on the model, toward Collaboration, away from the Control-Cultivation tension that had caused the original friction in 2017.

The NPS score that had started at -14% sat at 50% by February 2021.

The DBS 2.0 Programme formally ended. The culture didn't. It kept going because enough people had experienced what it felt like when leadership actually closed the gap between what it said and what it did. That experience is what people carry forward, not the programme name or the slide deck.

Culture drifts toward entropy when leadership stops paying attention. The work is keeping better behaviour alive long enough for people to believe it's real.

It's worked for me.