Flywheel effect, part 2 and conclusion
In the previous post I explained the flywheel effect and how you, as a director, can use it: get the wheel moving in the first place, choose how and when to push it even more, and keep friction away from it. I also went over three avenues that deal with your people and the information they work with: culture, metrics and visualizing business outcomes. In this post I’ll go over four more, this time around how your org builds and makes decisions, and then wrap things up.
Dogfooding #
Dogfooding is when a team can use their own product to develop the product itself. A good example is Apple, whose people use Macs, Pro Displays, iPhones, AirPods and the rest of the lineup every day, including to design and build the next generations of those very products. When the team uses the product just like a customer, they are in a position to feel the customer’s pain, and this is a powerful tool in shaping the product roadmap to have greater traction with customers.
Most companies aren’t in Apple’s position: if you build medical devices, a logistics platform or an insurance claims system, your engineers won’t use the product to build the product. The idea still applies on a smaller surface, though: wherever your own people need something your customers also use (your API, your dashboards, your documentation, your support flows), have them use the customer-facing version, not a back door.
In addition, there’s an implicit increase in ROI: company resources are put into making a single feature or product, which is used both internally and externally, so you don’t have to invest in making one for the team and one for the customers. In the context of the flywheel, when your teams don’t have to develop and maintain two separate products that differ only by who has access, they can put that time into the next improvement, and they also get faster and tighter feedback loops on how the product is shaped over time.
This is where you come in as a director, because no single team will choose this on its own. A direct database query, an internal admin page or a quick script is always faster than going through the product, so every team, acting reasonably, builds its own shortcut. If you multiply that across your org, you end up maintaining a second product nobody planned. You’re the one who can set the default: before a team builds an internal tool, they first check whether the product can already do it, and if it almost can, you put the gap on the roadmap instead of letting it become a side project. Whatever your own people fix this way, your customers get too.
Automated tests #
In my career I met quite a few people that for some reason played down the importance of automated tests. Going back to the topic of Automate selectively I think this is the clearest winner of the “what should be automated” competition. In a software engineering org your people are going to make a lot of changes to the system over time, so why not give them a tool they can use themselves to verify their new work is doing what needs to be done, while at the same time not breaking what the system has achieved so far? This is even more pronounced in new projects where you have no prior constraints and can set up the testing loop fairly quickly and iterate from then on very cheaply.
Another side effect of automated tests is literally unseen: the bugs that were caught and never deployed, so they never created an issue to customers, and never impacted the business. It’s really hard to measure the financial and reputational harm prevented, had those bugs been deployed, or the time your people did not spend fixing them and could put into improving the system instead.
As a director you won’t be writing these tests, but you can make sure they get written by making automated tests an expectation across all your teams, not just a habit of the ones that happen to care; keeping room for them in the roadmap; and holding the line when a deadline gets close. That last one is the hard part: under pressure, tests are usually the first thing a team trades away, and teams will only stop doing that once they see you won’t sign off on the trade.
The power of the written document #
I love written documents. Spreadsheets and presentations are great tools, but their very nature prevents them from being able to drive a meaningful knowledge discovery journey and discussion. Spreadsheets just show a bunch of data, so it’s hard to take the reader through it to tell a story. Presentations are punchy and can have powerful visualizations, but the format doesn’t allow for a lot of text, and discussion comments fall short in the small vertical real estate. A document enables you to convey a lot of information, at the pace and in the order that you want as the author, and the vertical nature of digital documents is a terrific form factor for discussion comments by collaborators.
Written documents also act as knowledge repositories that can be referenced in future projects or on-boarding phases. They could also play a part when investigating an issue or incident or a customer’s complaint. And in today’s world of LLMs being used everywhere in the workplace they are instrumental in building the model’s business context so it would better answer your business questions.
As a director you can’t be in every room where a decision gets made. When you write a good document, your reasoning gets into those rooms even when you can’t be there because of other commitments. That’s why it’s worth making written documents a standard across your org, for example design docs or decision records.
New technologies - risk and reward #
It’s very tempting to adopt the latest rage and play with new tools. This has been even more true over the past couple of years in the constantly accelerating environment of AI workplace tools. As a director you’d be wise to ensure that you weigh the risks and rewards of new technologies, since if you make the wrong decision, you are likely to pay for it in the future. And if not, your successor will have to pay and “thank” you for it.
As a director you also need to look at the risk across your teams: one team adopting one new tool is a manageable bet, but five teams each adopting their own means five maintenance burdens, five areas of expertise to keep up, and five things that break in different ways. You can’t review every choice yourself, but you can decide how these decisions get made: which ones a team can make on its own, which ones need a formal proposal, and who has to sign off.
What are likely risks in adopting new technologies?
- They are immature and not road-proven, so are more likely than not to change rapidly. If you try to match the rate of change, you are doing less impactful work. If you decide to let the changes go by the wayside, you’re just adding tech debt and rot to your system. Mature technologies are more likely to be stable and better at doing what they are meant to do.
- Addition of security or performance issues you have no idea are there. If these materialize it won’t be a fun day at the office.
- Untenable commitments in the long term. Any additional technology you adopt means you also have to invest some time in on-boarding, training, monitoring and maintenance. Keep in mind that these look very different if the technology in question is a small library or a new database. But any kind of technology will accrue these costs and add tech debt, and you don’t want to reach the day when you’ll be overburdened by debt. Some time from now who will maintain the technology? Who will ensure that training resources are up-to-date? etc.
And again, for the sake of a balanced discussion, there are also benefits to adopting new technologies:
- Software tends to rot over time, be it commercial or open-sourced. Nothing lasts forever and your system will require constant maintenance, that tends to increase in cost over time. There could be a tipping point where adopting a new technology is cheaper for the org, in money or effort, compared to maintaining what is currently in use.
- Enabling new capabilities or improving business outcomes. If a new technology means you can do the same thing for significantly less money overall (taking into account the person-hours required to adopt the technology), then there’s a strong incentive to do it. Likewise, if the technology enables a new capability that strengthens the product, resulting in a positive change to some business outcome, then isn’t it smart to consider doing so?
- It’s fun! We’re engineers and we like to experiment with shiny new toys and build new things. Keeping the spirit of fun in the org is a component in maintaining high morale and professional satisfaction.
You won’t do all of these at once. Pick the one where your org feels the most friction today, give it a push, see what it frees up, and put that time into the next one. That’s the point of a flywheel: the first turns are the hardest, and it gets easier from there.