Automatisms with impact the journey of a team obsessed with delivering value
And you are?
I’ll start by introducing myself.
My name is Mariana Gomes, I have co-founded two startups before joining Cabify, where I have started as a Product Marketing Manager over a year now. In February of this year, I had the opportunity to become a Product Manager for the Support team, and I love my decision of taking it ever since.
Now, don’t get me wrong — my entire experience at Cabify so far has been an amazing and rewarding learning experience. I just think I’m loving to be a PM more because I have finally found a job that I actually have fun doing.
I will tell you one example where me and several other teams have worked together from the problem identification and quantification, to the scope definition and solution, launch and results analysis, which are now being used to enlarge the scope of this initiative.
Automating customer service — how it all started before a clear why
The on-boarding process as a PM of Support was, let’s say, challenging. There was a lot going on, initiatives to launch, features being launched, and in the middle of the ups and downs of all that, a lot of decisions to be made about the future focus of this team.
One of those potential initiatives was to somehow automate the resolution of certain customer support tickets in order to save Customer Excellence Agents’ time (and costs) on topics that weren’t actual problems.
It was clear that there was a relevant amount of tickets that were submitted by users with either general questions or reporting apparent problems based on a misunderstanding of how Cabify works, especially regarding how pricing is applied.
At the same time, there were different opinions on what was the way to go: some teams were of the opinion that we should invest right away on a virtual assistant with an interactional interface, while other teams were more inclined to test an internally developed model to automatically identify the topic of a certain ticket, reducing costs of assignment to the CEX agents.
After several meetings of a back and forth exchange of opinions, I realised that there was no obvious way to go for anyone, and for me that only means one thing: let data decide the way to go.
The analysis that allowed us to simplify
Especially when working with cross functional teams with several team members, I can’t reinforce this more: question everything. This is something that the Head of my area, Marco Canton, had told me right away, and I am gradually realising how important this piece of advice is the more time I have as a PM at a company with such diverse profiles.
At the moment we were starting doing a more detailed analysis, the outcome of a previous meeting was more inclined to replicating a virtual assistant that was being tested to deal with other types of contacts automatically. The only two conditions that this virtual assistant had were:
For the time being, the model was only trained to deal with contacts with 12 words or less;
As the way the virtual assistant was built at that time was based on a decision tree of potential outcomes, the content intent couldn’t be managed in a too complex way, as that would mean higher costs in covering all the possible variants of interaction between the user and this virtual assistant (or alternatively, worse results!).
It became pretty clear that we needed to understand if this was even possible with the contact categories Cabify had at the moment. Taking those conditions into account, I started looking at the different categories of tickets the company has had so far, specifically to the following metrics:
Number of words, as we wanted to prioritise categories with less word count (and never more than 12);
Number of replies, as less replies would mean less complexity in solving a certain category of tickets, and therefore higher likelihood of having a successful first trial;
Difference from the time a user received a first response from Customer Excellence agents to the time a ticket was considered fully solved, as a complement of complexity measurement, as the number of replies could not be entirely representative of complexity on its own.
After this analysis, there was good news and bad news. First of all, the number of contacts received from users with 12 words or less were frankly low. In fact, only 1 out of around 40 categories had an average number of words below that condition. So going forward with the virtual assistant was simply not an option at that point.
However, when looking closer to the number of replies per category, the average was actually 1 for some of them, and the median was 1 to 39 out of 40 of these categories! Meaning that we were actually solving more than half of tickets with just one answer! So why would we even invest in a virtual assistant with an interactional interface in the first place?
Of course, other things have to be taken into account, such as the user experience and satisfaction on the entire process. But for this first step, with the only goal of reducing time spent by agents on resolving tickets that do not correspond to actual problems — which as you may remember, was the objective of this initiative — that was enough to decide.
The way to go? We needed to identify which categories could be solved with just one answer, so we could move forward with a first pilot as fast as possible, delivering value quickly while measuring the impact of having automatic answers for certain categories.
Solution, Launch and Results
Although several types of categories were taken into consideration for this first pilot, the most important ones were related to when users had doubts about the price charged being different from the estimated.
At Cabify, we aim for transparency at every touch point with the user, but we also understand that in the rush of moving around the city some things may not be super clear at first, so we completely understand why sometimes users contact us with doubts on why a certain price was charged.
The interesting part of it is that around 7% of the contacts we receive on that category are actually regarding journeys that have the same price as the one that was estimated!
So we started this pilot precisely with those cases: exposing the estimated and real price, and, if applicable, other fees related with the abnormal waiting times from the Driver, or tolls — stating exactly why they were applied and where the user was communicated that information in the first place.
With that information, every time a user would submit a ticket on certain categories related to journeys with specific conditions, he/ she would get an automatic message explaining exactly what had happened, and reinforcing that the price was actually the same as the estimated.
And guess what? The reopen rate of those tickets — meaning, the percentage of users that answered back when receiving this automatic message — was pretty similar to the one of manually managed tickets in the same category, which was a great result as a first attempt!
More than that, the NPS was actually higher for these automatic tickets, when compared to the manually managed ones — a great indicator of how users value a quick response in these situations.
What’s so special about the Support team
As a Product Manager, “it’s not your job to choose the best option, it’s your job to create a compelling option” (source: Product Talk).
I feel that I’ve been doing precisely that, essentially working as a filter between the business and engineering disciplines that are part of the Support squad. But I can only afford to do it because I know there’s a stellar team designing and implementing 99% of the outputs we deliver — effectively and consistently.
Because of that, I left this section for some of them to tell us the secrets to creating such a special team as the one we have in Support.
“For my part I have the opportunity to tell that I feel like part of a group of friends, that we built a project together that we love.
We care professionally and personally and that can be seen in the way we trust each other.
The key in this team is that we can share our ideas and they are all valued equally, making all of us important. Personally, that makes me proud of myself and my team.”
“I like to think that we’re driven by the KISS and YAGNI principles (which are kind of a continuation of the “question everything” idea). From our point of view it’s very easy because the prioritisation based on business value has already been done, so our focus is clear. The question is: how do we solve the one problem that we need to handle right now? The response usually is: try to deliver to our users the simplest solution that could possibly work.
Do not think about this in terms of laziness: it’s all about productivity. We’re never delivering temporary solutions or quick and dirty hacks, it’s just that we’re not spending weeks or months in something until we’re sure that’s what our users need. Once a solution is delivered, we measure, and then we evolve it as needed.
Regarding the team, one key aspect is motivation. We’re a multidisciplinary team with different interests, experience and skill levels. Just try to assign the right task to the right team member, and they’ll enjoy their own work. Again, having a clear view on who is going to use this particular feature, and why it is important, is essential to keep motivation up. Very often we’re seeing team members choosing their own tasks, eager to see them working in production in a few days.”
“As Support Team manager, I would say that the most special thing in our team are the people who form it. Yes, it sounds a bit mainstream, but I repeat it again: The most important thing are the PEOPLE and the relationships between us. Productivity, results, efficiency.. Everything else comes later, and will come, trust me.
We have some rituals that have helped us to build strong relationships between us and to create our own culture:
Daily meetings. The daily meeting that you already know. :)
Team meetings, weekly.
Brainstorming. Every once in a while, we hold an open thinking session where everyone can share their thoughts.
Yes, these meetings are not so special. I guess that you all have something similar in your teams. The key thing here is a concept that, at least for me, is the most important one: People First.
In almost every team meeting, we laugh. Sometimes one shares a joke, someone shares some story about his/ her kids… or because of some silly stuff of day-to-day work. But we always try to have a smile on our faces after the meeting is done. When we talk about serious stuff, we get serious, but the start/end of the meeting is very informal and relaxed. As a manager, I try to make that happen.
Cameras are always on, so we can see each other’s faces. Not to see if you have clothes on (well, that too), but to understand the mood of everyone.
Open Agendas, for everyone. All of us can add a point to the agenda and talk about it. Everyone has the space to talk and to be listened to.
We were a 50% remote team before COVID-19 hit us, and now, we are all remote. Thinking in People first is the thing that makes Support Team special, but not only our team, the whole company, as that principle is one of the most important ones in Cabify.”
Did you feel identified with this text? Check out opportunities to join the Cabify team here.
We are hiring!
Join our team and help us transform our cities with sustainable mobility.
Check out the open positions we have in .