BEFORE THE CRISIS
Jean Baptiste started at Admiral a few years ago. His job was to design new insurance products, but he knew the company needed to adjust its offering to a market in which younger generations no longer buy a car but rather use a mobility service. Given the circumstances, the insurance world would have to follow suit and offer temporary insurance coverage for whatever scenario young drivers might find themselves in, from learning to drive with a parent to sharing a car or occasionally renting one. Such policies would have a duration that can vary between one hour and 30 days. As the company’s motto states, “At Veygo, our goal is to provide the best insurance options for drivers with no cars.”
Jean Baptiste also meant to demonstrate that insurance contracting and associated interactions can be digitized. Like most insurance companies, Admiral still heavily relies on mail and call centers to interact with its customers or would-be customers. JB was determined to develop a contactless insurance product. That’s when Veygo was born. It was mid-2016.
Four years later, the company is making more than half a million sales per year, customer reviews are very positive, and the assumption that there is a market for digital native on-demand motor insurance is confirmed – at least when it comes to Veygo’s insurance products (temporary cover and learner driver’s car insurances). Over time, the team grew from six to forty-eight people, between Q2 2017 and Q3 2018. It is now 53-strong.
“Lean has really helped me,” Jean Baptiste told me during our gemba walk in January. Jean Baptiste – or JB, as his British colleagues call him – took his first lean steps on his own, reading book after book. At first, he was attracted to the tools (aren’t we all?) and saw Lean Thinking as a sort of Agile 2.0. – a compliment paid to lean, I would say, since most “agilists” tend to consider it a thing of the past. Lean tools, JB reckoned, were going to provide the means to be less people-dependent and to eliminate waste, while helping to sustain a massive flow of new features for the website. Back then, JB’s favorite KPI was the number of releases that could be achieved over a given period of time. As Taiichi Ohno’s Workplace Management reminds us, we are all full of wrong ideas on how things work. And JB is very transparent about the topics he actually changed his mind on. And today he firmly believes that those changes of mindset continue to help him and his teams during the Covid-19 crisis.
The first such change required the help of an external sensei, who showed up at Veygo with a question: do you know your customers?
DIGITAL AND LOCKDOWNS DON’T MEAN NOT TALKING TO CUSTOMERS
At the time, the sensei’s question had come as a blow. In a world of fast growth, digitization, websites, agile and millennials as customers, Veygo’s focus had been on sprints and fast delivery. At the time, the scrum teams were fighting for minimal management interference, so long as they delivered, and JB’s vision for the company was very much a group of technical silos he was trying to supervise and align.
Putting the customer back at the center and asking how to better grasp her needs, or what level of quality she perceived, was mind-boggling at first. How do you actually talk to customers when your entire business model is to design a “buy what you need, no assistance needed” website? The fact that no one had to talk to the customers was considered a success factor. And we of an older generation are not allowed to laugh either: in our time, we thought we could do the same with hypermarkets or shared service centers.
It turned out that five out of six members of the Veygo management team had actually never talked to a customer. Product owners thought out features without actually checking them with real customers. Additionally, if a feature had a poor impact on sales, it was unclear whether the idea was wrong in the first place or whether the execution failed to add value for the customer. “Either way, we were not learning,” JB sighs.
And indeed, I had seen during the gemba walk a major shift from that initial approach, with scripts of customer interviews, a takt time for those interviews, NPS or CES indicators. On the Veygo website, FAQs and Help Centre articles are flourishing, fueled by the conversations the team has with customers. Ever since the lockdown began, the live chat feature has become as essential mode of communication with customers.
Discussing with customers led to some interesting revelations:
- A learner driver insurance had until then been considered a practical, financial feature to cover against risks when kids are building up their driving experience in their parents’ car. As such, the insurance guarantees that, in case of a collision or crash, the parent’s No-Claims Bonus is unaffected. Interviews with parents provided another valuable insight: driving with their kid and passing on their experience was seen by some of them as an unexpected opportunity to spend some time with their teenage sons or daughters (many of them, at that age, tend to avoid family time and distance themselves). This important emotional factor is now being taken into account by Veygo teams.
- Many customers complained that they struggled to find help on the site. As reaching out for a contact in Veygo was not obvious either, they were frustrated or unsure whether the products were right for them. This led the team to develop the Help Centre significantly, dramatically improving the customers’ ability to self-serve.
This new focus on customers has not disappeared with the Coronavirus crisis.
JB’s aim was to sustain business continuity while protecting everyone at Veygo. Preparation for a possible lockdown actually began as soon as the first villages in Northern Italy were quarantined on February 21st. It took Veygo 72 hours to plan and prepare for the switch to a 100% remote operation, with laptops and dedicated VPN for remote access provided to every team member. They also worked on the impact of remote work on each job and studied worst case business scenarios – including a lockdown of the Cardiff areas as a result of a local outbreak, with the rest of UK still operational and busy with its business-as-usual. They were ready by early March and switched to remote working one week before the UK enforced a full lockdown on March 23rd.
The teams were still battling with bad chairs and new remote tools when JB asked them what could be done to help National Health Service workers who may need a temporary insurance cover to go to work. Action was immediate: the decision to offer a 75% discount for NHS members was made on Wednesday, March 18 at 9 PM and the new feature was in production by 4 PM the following day. Not bad for a project driven and executed remotely! Within a week, they had 200 requests to benefit from this offer and many positive comments. This gave more meaning to daily tasks and routines and highlighted the importance of acting as a team. JB tells me of how people exchanged internal messages saying how proud they were of their achievements and to be part of the Veygo venture.
As he reflects on today’s situation, with developers and product managers no longer working in the same space, JB expresses a worry that output (deliver at speed) may once again overtake outcome (deliver value). The activity is lower of course – particularly on the Learner Driver insurance covers – but activity for Temporary covers has only decreased by 30%. And the feature release activity is flowing, unimpaired.
CODERS SHOULDN’T HAVE TO CHOOSE BETWEEN GOOD AND FAST
Initially, quality had not been JB’s prime objective. With Veygo celebrating the number of releases and never talking to customers, questions on current quality issues were spurring frowns, if not laughs, from the teams: “We’ll do quality when we have time” or “80% (or even 50%) of many things is far better than 100% of a few items”.
We get a similar response from teams when we ask them about kaizen – “When we have time,” they often say. Kaizen, like quality, is an investment in the future that Deming had perfectly understood. “The result of long-term relationships is better and better quality, and lower and lower costs,” he said. Deming also recommended we do away with final inspection and develop built-in quality. “No one knows the cost of a defective product – don’t tell me you do. You know the cost of replacing it, but not the cost of a dissatisfied customer.” This is food for thought: with the pandemic, new customer expectations are bound to emerge and those who fail to perceive and meet them will found themselves out of the market really quickly.
Veygo decided to have a go at quality in October 2018. They saw it as a quick bug-cleaning campaign at first; that couldn’t have been further from the truth. The team found they had many customer complaints, but few of them were visible and most were not addressed. Critical functionalities on the website didn’t work: for example, the confirmation of an insurance coverage went in a loop after signing up for it, leaving the customer in doubt as to whether she had actually purchased insurance coverage.
When the team tried to address and fix those issues, it turned out that they did not trust the reported bug backlog, that unit and integration tests were poorly designed (if not absent), and that test programming was an unknown notion. Developing code as quickly as product owners could pile up new features had been their sole motto up to that point.
They sat down and worked on it: developers shouldn’t be asked to choose between good or fast. Therefore, the product development culture was changed to include quality. Bugs were tracked in bug review sessions, a tool was secured to check difficulties encountered by customers on the website (tracking their behavior and highlighting potential frustrations with SessionCam), and developers were trained to develop tests. This quality intervention paid off and the conversion rate shot up.
Another positive side effect is that feature ideas now come from actual user problems: the decisions on new developments are consequently easier to make and more obvious to all, and consensus is reached more easily.
JB can see that new projects today are impaired by the difficulty to assess features with customers: what is the customer outcome? How to define success? “Brainstorming is harder to achieve remotely and, as a consequence, serendipity effects on innovation are less likely. Having a coffee with a colleague to discuss a problem in person is just not the same thing as doing so on video call. This is something we will have to pay attention to in the coming weeks.”
AUTONOMY DOES NOT MEAN INDEPENDENCE
We don’t how long the lockdown will last or whether we are bound to see successive waves of lockdowns throughout 2020, as this will largely depend on how fast we can produce protection and testing for us all. In such a scenario, learning to work remotely effectively becomes imperative. But how do you actually develop team autonomy in such a context?
Veygo’s experience with this over the past couple years can teach us a lot. You don’t grow from six to fifty-three people in two years without a major reflection on the onboarding process and how to organize teamwork. There are several aspects to this: you need to create a reliable management team, align everyone around common goals, and foster collaboration on the delivery flow. This may also mean a complete change of behavior from top management.
Back in January, JB had told me the story of how they created standards for developers. When he started to go and see how developers worked, his actions were perceived as an audit and not much came out of it. Then he thought about developing work standards and tried to enforce them. That did not work either, as the teams became more and more defensive against any intrusion.
This prompted JB to think hard about what autonomy means and to introduce a completely new approach:
- First, clarify the long-term goals the team wishes to achieve on lead-time and quality. One board in the Veygo office displays this very clearly: developers now track the bug backlog (quality), the time-to-market for a new feature (delivery), and the website performance (page load speed and the number of releases per week). Customer satisfaction is analyzed both through interviews and NPS scores at purchase. Feedback from customers is gathered by means of independent customer surveys that regularly yield a high score given the nature of those insurance products. And by the way, customer satisfaction is still high today, in spite of the lockdown.
- Then, introduce a form of one-piece flow (a maximum of two products in any given production cell) with an andon system and a chain of help from management in place. This is something Veygo still struggles with as there is a trade-off to be found between delivering at takt time and autonomy and self-reflection. In this lockdown situation, they have a daily scrum meeting still in place and a form of andon on tickets that are obviously stuck.
- Lastly, JB is learning to develop and nurture the right conditions for kaizen (value for customers, clear goals, andon, dojo). “I have learned the hard way and I picture myself today rather like a gardener: nurturing, feeding and waiting for the seeds to grow and blossom,” he comments. Kaizen is also an opportunity to “adjust” the teams as new potential leaders emerge: personalities geared towards learning, collaboration and self-development typically stay and are encouraged to continue growing, while others may well end up leaving (it has happened). The new PPM – Principal Product Manager – on Temporary coverage and Learner Driver insurance, Dan, is a lean enthusiast and he now steers both the Product (features) and Operations (developments) teams.
During the gemba in January, JB had walked me to a board showing a kaizen approach in six steps led by a scrum team, with support by JB and Simon, the other PPM. It was focused on smoke tests, which preliminarily simulate the would-be customer journey with the new code to detect potential failures. Some 40 hours were lost every week running smoke tests that do not work. The first pass on this kaizen identified parameters to be reworked, much to the relief of developers. Faulty smoke tests went down from 97% to 0.4%. In turn, this revealed new issues such as a slow process to release new features into production: they are targeting a 50% decrease and continue to work on it remotely. But JB thinks more work on this is needed: “Cleaning and sorting is one thing, but we need the scrum teams to further investigate root causes and find ways to prevent recurrence.”
Talk to your customer, work on quality, develop autonomy and teamwork: JB believes these major shifts in his mindset helped him to organize the new remote working setting. “The major advantage of lean, as I see it,” JB concludes with a smile, “is that we are learning much faster: we see our problems faster, understand our misconceptions quicker and change accordingly. That is a major advantage when you are trying to disrupt a market, and it becomes even more crucial when you are facing a crisis such as this one!”