From 0 to 2M€ ARR in less than a year: what we’ve learnt at Onepilot?
The genesis of Onepilot 🐣
A few months ago, on a sunny lockdown day, we — yet to be cofounders — were discussing business ideas: we had all experienced some pains managing customer care in our previous experiences and hadn’t found any proper solution to solve our tickets. Of course, there were many great ticketing tools and bots out there but we were wondering why no one had built a seamless outsourcing solution that could actually solve tickets anytime.
As the discussion went on, our belief grew that being tech-driven could improve the state of the art of the industry and set new standards where outsourcing rhymes with quality, while maintaining competitive costs and operating locally. That’s how we got the first idea to build Onepilot, the next-gen customer care outsourcing solution. Since the beginning, we always had this idea that we wouldn’t want to be like traditional outsourcers measuring success by their workforce size and not efficiency or ticketing tools and bots out there providing great value but still limited as they’re not solving tickets.
However, we needed to start somewhere to test the idea; and we decided to publish a sketch of our inexisting dashboard on a Facebook group called “French startups” asking for beta testers. We had immediate good traction, receiving dozens of comments and DMs and a few weeks later our first client called us and said “Can you launch next week?”. Needless to say that we didn’t have any agent nor any tech but we still decided to take on the opportunity. We quickly realized how big the pain of customer care could be for businesses, in fact it’s a growing pain of 50 millions tickets, daily, in Europe.
Our initial sketch for Onepilot
Our initial “tech stack” was pretty simple: it was made of a Google Doc for our knowledge base and a Google Sheets for the billing tracking. We called our younger brothers and sisters and asked them to help find initial agents to take shifts: without knowing it we were already shaping a key foundation of our model with agents being freelancers, working from home on their own schedule. A few days later, we were solving our first tickets. Not very tech-driven, you would say…. As you can guess, the productivity of our agents was really low (at around 4 tickets per hour on average) and the cost structure was quite ineffective.
The reality of our initial tech..!
Initial tractions 🧨
Time flew by and without realizing it we had onboarded our first 10 clients and celebrated hitting 5K€ MRR just a few weeks after our soft launch, which was a really impressive milestone for us. This is also when we decided to quit our previous jobs and run Onepilot full time.
As we grew, things were becoming messy, with agents’ onboardings being quite time-consuming, clients onboarding and management taking a lot of dedication. We had already replaced our Google Docs & Sheets by a beta of the Onepilot app, with very basic features. But this wasn’t enough to keep up with the pace at which we were growing and this is when we realized that we needed to build a strong roadmap and laser focus on its execution. In April 2021 we raised our seed round, led by GFC, to fuel our growth and hire our team. From that point on, a third of our hires were dedicated to our tech team.
Cofounder, Pierre, running first TV interviews after our seed round
As we kept growing and onboarding more clients, we faced a dilemma: (A) slowing down our growth and focusing on our tech for a few weeks to make sure everything stands or (B) keeping up with our growth and deploying tech releases gradually with operations fixing gaps in the meantime. Of course, we went with option B and that’s precisely what allowed us to focus on the right things.
We shipped our second iteration of our platform with two core basic features: an in-house knowledge base editor (now available on our CSO for testing) enabling agents with fast access to knowledge, as well as users & tickets management tools. This was the start of our journey to reduce the workload on our team and give the right tools to our agents. Agents’ productivity was immediately doped by an incremental productivity of +1.5 tickets per hour.
Reducing human dependencies to scale faster 🚀
As we reached 100,000 tickets solved, in July 2021, we were in a good place to understand what was needed to boost the productivity of our agents and reduce human dependencies while improving quality. We divided agent productivity in four key factors:
Productivity factor #1: agent utilisation
As some of our clients are too small to dedicate an agent, we pool clients on the same shift. That initial manual work has quickly been replaced by a self-staffing module we developed internally. Technically, it was a bit challenging as it needs to sync slots between the agents and the clients with various needs. We wanted to achieve three things here:
- Modularity to fit all the needs of our clients and agents
- No admin interactions and no manual work
- Agents utilization maximisation to remove schedule inefficiencies
Practically, we built a client grouping system based on tickets forecast that allowed trained agents to self pick shifts on these groups. The schedule became just easier for the ops team, freeing up about 6 hours of work weekly per teammate and doping the productivity of our agents by +0.9 tickets per hour by maximizing their utilization.
Productivity factor #2: efficient knowledge access
Shortly after we released our first iteration of our knowledge base, we started working on a V2 which brought significant improvements. Through critical path analysis, we researched the way our agents accessed knowledge and tried to remove all frictions. For example, we realized that many agents would spend time updating macros from “Good morning” to “Good evening”. These now update by themselves based on the local timezone allowing agents to save on average 6 seconds per ticket.
Our new knowledge base editor allows agents to access relevant knowledge in a matter of seconds, with efficient keyboard shortcuts à la Superhuman. From a technical perspective we also upgraded our editor to another tech stack, React, and brought significant UI improvements.
Productivity factor #3: tools centralization & unification
Agents waste a lot of time navigating between tools. An average Onepilot client uses 4 tools: ticketing tool, back-office tool, payment tool and typically WMS or any industry-specific tool. As Onepilot agents often shift on 3–4 clients at the same time, you can imagine how easy it can be to get lost within so many tabs open. This is why we’re centralizing and unifying all these tools in a single page.
We started with integrating ticketing tools on what we call our ticket center. This allows us to route tickets to the agent with the adequate training at the right time and to control the load of the agents. Going a step further, we’re kickstarting a wider sprint to integrate back-office tools such as e-commerce platforms, payment platforms or warehouse management systems. Brick by brick, we’re integrating and unifying various tools starting with the most frequently used one and progressively working on the less common tools.
We’re just at the beginning of our journey with these integrations, but imagine giving superpowers to agents by allowing them to track an order, report an issue with the transporter, reship an order and offer a promo code, all that in a few clicks straight from Onepilot. Magical, no? Well, thanks to open APIs and a high concentration of popular tools, we’re able to bring this superpower to our agents, allowing us to significantly improve productivity going forward.
Productivity factor #4: suggestion engine
Pre-solving tickets to simplify agents’ work is the foundation of our strategy. Our suggestion engine squad is focusing on preparing an answer for our agent so they just have to validate the answer, do extra customization or execute the process which would need human interaction.
The advantage of Onepilot is that we’re processing thousands of tickets each day and get a fair amount of data to parse actual tickets to anchors of our knowledge base enabling us to perfect our model day after day.
Fast problem identification allowed us to fix gaps before they became bottlenecks 👩🚒
We scaled fast and that meant that we needed to onboard more agents. Although fewer agents are needed now for 100 tickets than a few months ago, our hyperscaling required to be organized with the way the supply-side of our business was scaling. Transparent collaboration between the product, tech, ops and sales team allowed us to identify quickly upcoming bottlenecks and fix gaps before it was too late.
An evolving compensation structure for our agents
As agent utilization improved with our self staffing module and smart client grouping, we switched our model from paying agents hourly to a per ticket basis. This allowed us to directly align our incentives and boost the efficiency of our agents. But you will tell us, isn’t that encouraging agents to deliver inferior quality to be faster and cash in more money? You’re right and this is exactly why we incorporated a high quality dimension in the compensation: in fact, more than a third of the agent pay is directly aligned to quality, guaranteeing us a higher level of quality.
Larger clients mean new challenges
As we onboarded larger clients like traditional retailers or well-known unicorns, it was time to step up with cybersecurity. This is why we built a cybersec squad focusing on DevOps and we’ve been shipping important features like virtual machines, advanced agents monitoring systems and a dedicated agent VPN to ensure a high level of privacy and security to our clients.
Scaling agents onboarding
By now, the productivity of our agents has been growing by more than 2x to around 9 tickets per hour, which means that we need less agents per 100 tickets now than a few months ago. However, as you can imagine, scaling from 0 to 2M€ ARR means that we needed more agents than before — leading to 120+ active agents now.
When we launched, onboarding a new agent would take the ops team around 15h of manual work: setting up the agent profile, doing initial training, explaining how Onepilot worked etc. Clearly, this quickly turned out to be not scalable and we have been rethinking everything here: now agents apply and are automatically pre-qualified based on key success criteria that we identified. Most of the training is now automated with interactive e-learning embedded in the platform followed by strict testing procedures. New agent onboardings now require less than 30 minutes of human interaction, which is 30x less time than before!
Scaling clients onboarding and management
On the client side, when we were doing the first clients onboarding, it would take us about 7 to 10 days to run the full onboarding. Having close to 100 clients now, that would have been totally unmanageable; we therefore took the same approach of analyzing the work done by our team to simplify & automate as much as possible.
We just shipped our customer self onboarding flow, where customers can connect their tools in a few clicks, import their knowledge base from template libraries embedded on the Onepilot platform and fast-track their onboardings. This for example results in a massive upside in efficiency, with some onboardings now taking as little as 2h — great upside from 7 to 10 days!
As per clients management, we quickly got to understand that in order to scale, we needed to maximize self-service. And to do that, we needed to build a great product allowing clients to manage everything from their dashboard. We also quickly realized that our clients actually preferred being able to manage their knowledge base, schedule, Quality Checks and much more by themselves, without the hassle of taking an appointment with their Account Manager or writing an email. As a result, self-service coupled with automations, for example QC automation with sentiment analysis, allows account managers to manage 3x more clients now than 9 months ago while raising the level of satisfaction for our clients.
What’s next for Onepilot? 💪
Going from 0 to 2M€ ARR in less than a year was unexpected for us. It went fast and came with great challenges that forced us to ship features, fast & furious. In fact, we have done close to 5,000 commits in this little amount of time; some devs would tell you this is crazy — because it is — but this is exactly how we kept the house standing. We’re currently in the middle of our first phase, where we keep learning from our agents step by step, day after day, to make their work easier, bring them more automation and build efficiencies at various levels. Target: 15 tickets per hour by end of year 2022 with incredible quality level crafted locally and reaching 10M€+ ARR.
Once we reach this milestone, we’ll be ready to kickstart our second phase: where agents supervise our tech. Agents may need to keep executing some remaining manual actions that aren’t automated or unified yet; however, our main goal today is to go from tech-enabled to tech-powered, human-monitored. By achieving a high level of productivity and reaching 25 to 30 tickets per hour and maintaining a good level of pay to our agents, we’d be able to decrease prices for our customers and compete with offshore outsourcing and of course increase margins to SaaS levels, while managing significantly less agents and reducing dependencies.
Find why our clients’ users love Onepilot here. ❤️
Don’t miss any news! Subscribe to our newsletter
Get our advices and latest news directly in your inbox!