At Cuvva, we treat customer support as one of our core products, rather than just something we have to do. To us, getting support when something isn’t going right, or if you just have a question, is just as important as getting insurance itself. The tooling we have built to enable this is just the start of what will eventually be a groundbreaking end-to-end support and claims platform.
Part of the way we channel that view is by treating all of our internal tools as full products with our own team as the end users. By tools, we mean the set of dashboards and applications that our internal teams use to do their jobs more efficiently.
Tooling as a product
When we started the process of building our internal tooling back at the start of 2017, we knew right from the get-go that it should be more than just an afterthought. In too many tech companies the customer facing aspect of the business gets all of the resource and attention, and support tools are slapped together in a haphazard fashion. Our approach ensures that our employees have a seamless experience, just like our customers.
We’re building the future of insurance, but we don’t want to just do this at the user level. We want to change the way things are done at every level of the stack. We have started from scratch, avoiding the old guard of yearly release cycles, dense and overpopulated displays that create a stressful experience via information overload, and years of legacy to support. Instead, we deploy new features and updates multiple times a day, and thoughtfully design every aspect of our tooling to keep information density to a usable and comfortable level. Below are a couple of screenshots of our Admin dashboard:
This is what we see when we look at a customer’s profile. All the products we offer and various other information are in tabs that are accessible if the employee needs to see it, with badges to indicate if the product/section is used, and a core overview available on the home tab.
This is how we show complex information for our longer term products. Everything that happens is represented as an “event” and they’re all shown in a timeline, enabling our team to see how information has changed over time.
Our Ops Dashboard handles manual tasks that we need to keep on top of. Above is the vehicle photo task, where we manually review every photo of a policy that is taken. Other tasks we have to manage include ‘know your customer’ (KYC) verifications, and urgent police phone calls.
Getting into the mindset
When we say we build our internal tooling as a product, this means that we put the same amount of design resource and product discussion into it as we do to our app. On top of this, similar to how we gather feedback from customers, we engage the Customer Operations (COps) team to make sure we are working on the right balance of features. One very important aspect to this process is making sure that we don’t try to do too much. As we are a startup we don’t have huge resource, so we have to be strategic on what to improve. We decided to go with the methodology early on to only automate processes when they become too unwieldy to do manually.
In addition, we have to make sure features make sense. Adding a certain feature might sound like a good idea, but if it can’t be done in the correct way, we’ll end up with a user experience nightmare, where certain aspects behave in one way, and others in another. Avoiding this early on means we don’t create a confusing product, and can prevent ourselves from digging a hole that we may later struggle to get out of.
But there are times when we need to realise that we shouldn’t be reinventing the wheel. If there is already a product that exists that can serve our needs, we should be investing in that instead of trying to build our own. A prime example of that is a tool we use called Intercom, which is the platform we use for handling all customer interaction. It isn’t perfect and comes with a few fairly major caveats that we have had to build processes around, but overall it’s pretty good for our use-case. The drawbacks of Intercom are heavily outweighed by the benefits of our approach not to build any mini SaaS (Software as a Service) startups inside Cuvva. This could leave us in a situation where our platform is down and customers can’t contact us.
This is Intercom. When a customer messages us, we get to see a brief context of the customer, previous conversations they’ve had with us, and a link directly to their account on our internal platform.
Although the inherent nature of a COps team is that it scales linearly to customer demand, a single Cuvva COp can meet the needs of drastically more customers than any other insurance company. This will continue to improve due to our investment in tooling, and improving our apps to reduce the need for assistance.
Of course, the team is much happier when the tools they use every day are constantly being improved based on their feedback. This happiness trickles down into both being able to retain great staff, and also into customer satisfaction via faster, friendlier, and more efficient support interactions.
The brand and personality of Cuvva is just as present internally as it is externally. The tone of voice is always people-friendly, rather than insurance focused, and we use emojis to convey certain information in a personable fashion across our tooling.
Scaling the team
When I joined Cuvva back at the start of 2017, we had no dedicated COps team (which was fine as we were only catering to about 500 conversations a month), and only a very primitive dashboard for viewing customer data. Today we have a 16 strong team, covering the needs of 35,000 people each week, and extensive tooling that enables a real person to reply within 1 minute – 24 hours a day, every single day of the year. No annoying “AI” chatbots, no queues and we don’t do anything to avoid messages – there’s a button which goes directly to chat on every screen in the app.
Initially, we set what we thought was quite an optimistic target: a 5 min median response time between 6am-1am every day. Somewhat surprisingly, we achieved this almost immediately. We gradually raised our internal expectations, first bringing the response time down to about 90 seconds, then expanding to 24-hour coverage.
Now, our median response time is about 45 seconds. Some customers are suspicious that they might not be talking to real people, or that there’s an automated system providing the fast responses. (Yes, they are exclusively human!) So we’ve decided to stop here for now.
As we’ve grown, we’ve found that we’re able to consistently improve the quality of service. This really surprised us, as it seems to be quite different from the teething issues similar companies have experienced while scaling. We aren’t exactly sure why we haven’t had similar difficulties, but we think it might be partially linked to how we manage COps’ time.
Our COps are heavily over-allocated to allow for spikes in demand, then when it’s quieter, they can work on larger/more intensive behind the scenes work (manual tasks like identity verification) which are less time sensitive. Unlike most companies, we don’t have “levels” of support. Every COp is highly trained, so customers aren’t left waiting for a specialist to become available to give them the information they need. Our COps do still specialise, but this doesn’t have any relation to their ability to handle support queries – specialisations only apply to the COps’ involvement in project work and more complex tasks.
This is a bit of a behemoth of a topic, and there is a lot more analysis that could be done on the work we’ve done, so I think we’ll most likely write another blog post on it in the near future.
We’re very happy with the progress we’ve made, but we are by no means finished. The bar is always moving, so we can’t get complacent. A big push for us as we go into the new year is automating more and more of the manual tasks COps currently do, which will improve our ability to keep the quality and level of reviewing consistent with increasing scale.
With the upcoming release of our next product, travel insurance ✈️, this will be the first product where we will be handling the start of the claim process (also known as “FNOL”, first notification of loss). This is our first foray into handling claims, and we’re keen to see if we can streamline the process of passing the information smoothly from the customer right through to the claims team. Eventually this will become complete end-to-end claims management. Every party involved in the claim from Cuvva, to the underwriter, to the customer, will all have access to a single source of truth where they can see all the information about the claim and its status.
We’re a tech-first company, and we want our technology to be great at every level. If you want to help us on our mission of making internal tooling great – we’re hiring! Head over to our jobs page to see if we’re looking for someone like you. If none of the jobs fit your skill set but you’d still love to work with us, send an email to firstname.lastname@example.org telling us why you think you’d be a great fit 🚀