6x Improved customer activation rate, 50% better CTIs, and 50% lower Sign-up costs. Find out more about Swvl’s hyperlocal marketing infrastructure!
As Swvl started expanding its network across different markets, we knew that Facebook advertising was the next logical step in our digital acquisition efforts. It’s scalable and it’s fast. You can reach thousands of people with the click of a button, and even if you don’t yet know who your customers are, you can be certain that they will be using Facebook.
The results of our first digital campaign on Facebook indicated we had made the right call. Our ads were successfully driving massive traffic to sign up on our platform. Despite this, we noticed that these sign ups were not yielding the expected uptick in bookings and transactions. Even with the right audiences, creative messaging, and an attractive value proposition (safe, reliable, air conditioned bus rides at affordable prices), not enough users were taking their first trips. There seemed to be a leak in the funnel that needed investigation.
When Covid hit, and lockdowns forced our operations to a halt, we were presented with an opportunity for introspection. We knew we had to take this time to figure out how to come back better than before.
Our approach so far had been to grow at all costs — to create supply first and let demand come. When we relaunched our markets, we decided to turn this approach on its head. Instead of launching our entire network across the city on day one, we were going to identify pockets of demand, and strategically create supply to fulfil the demand that already existed.
Our network’s utilisation dramatically improved — buses were no longer driving dead kilometres and our seats were being booked as quickly as they became available. While this was a great move for the business, it presented a new set of challenges for user acquisition.
More Isn’t Always Better
One of the underlying assumptions of Facebook advertising is that every Facebook user is a potential customer. It doesn’t matter who the audience is, or where they live — the goal of a marketer is to generate a user’s interest in the product, to make it seem like it’s an offer that can’t be refused.
But for an early stage ride sharing business that is slowly expanding its network in a post pandemic world, this assumption broke down.
We could still advertise to the whole city and create appeal for the product, but if we were not yet operational in the area the user was signing up from, we would be selling umbrellas in a desert. Not only would this be a lot of advertising dollars wasted, it would also result in a terrible user experience.
Finding the North Star
It was clear that any solution we came up with needed to encapsulate the inherent understanding that our product was different. It didn’t come to a user’s door step like a food delivery service or a ride hailing cab, and our user would not be persuaded to walk beyond a certain distance no matter how attractive the proposition.
As we explored our data, we realised that Walk to Station (WTS) was, in fact, the single most important factor in determining not only whether a user took their first trip with us after signing up, but also whether or not they would go on to become a high value user. The closer the user to a station, the higher the probability of them converting.
Optimising for walk to station with a route network that is constantly in flux meant that our acquisition campaigns needs to achieve three objectives:
- Geolocation Based Identification: Only users who live within a certain radius of our stations should receive ads. This would help reduce costs, and increase the likelihood of turning an audience member into a customer.
- Creative Localisation: Ads should indicate which station is closest to the user seeing them. This would reduce the perceived distance between a user and the swvl station nearest to them.
- Dynamic Expansion: Ads should automatically expand their reach across the city as more routes and stations get added. This would allow ads to pan the city in tandem with our network.
The Solution
We quickly realised that the tool best equipped to fulfil these needs and build Swvl’s hyper local marketing infrastructure was Smartly.io.
Using Smartly.io’s solutions required two steps:
- Setting up a real time connection that could feed the geo-coordinates of all our active stations to Smartly.io along with other relevant information like station name, district name, pricing, and promos. This was done with a python script that refreshed the data on a daily basis.
- Identifying the right level of granularity for the campaign structure -too small meant over a thousand ads with not enough return, but too large meant not being able to localise our messaging in a way that an end user would find engaging and relevant.
We decided to target all users within a certain radius of our station coordinates, and bucket creatives into districts. This meant that a user would see one out of 50 potential ad creatives with the district name on it, depending on the district they are in, but only if they were within a walking radius of a Swvl station.
The final tool in the shed was Smartly.io’s Predictive Budget Allocation (PBA). PBA utilises machine learning to systematically increase and decrease ad set level budgets based on their performance. With this optimisation algorithm in place, the high performing districts would automatically get more budgets and enable us to unlock more scale with better efficiency.
The Results:
Overnight, we saw huge gains in our upper funnel metrics, and within a week’s time, we started seeing significant improvement in the lower funnel metrics as well.
Our CTIs (Click to Installs rate), CTRs (Click through rates) and CPIs (Cost per Install) improved by 50%, 20% and 46% respectively. On the lower funnel side, our CPSU (Cost per sign up) and CPFR (Cost per First Ride) improved by 45% and 15% respectively.
Within only 12 weeks, we saw a 6x uptick in the number of people taking their first trip within a week of signing up. With these kinds of results on our very try, we knew we had hit a goldmine. It was time to bring more sophisticated use cases to life.
In the five months since our first hyperlocal campaign, we have developed a lot more control over our campaigns. 90% of our Facebook acquisition spend is now powered with automated ads that have enabled us to:
- Launch digital acquisition campaigns for new markets without waiting for the network to hit a minimum threshold
- Deploy budgets in a non linear manner so allocation can be optimised for network density and demand
- Scale up advertising in micro pockets where we saw user convert faster
- Advertise relevant destinations within a city based on the user’s pick up location
- Market different products to the same set of users based on the product’s coverage
The list doesn’t end here. Every week we continue to experiment with new and novel applications of this tool, and find ourselves able to leverage more and more of its capabilities. Setting up this approach does come with its challenges — it is a steep learning curve, and can be accompanied by a spike in CPMs. But at Swvl, we believe that making customers your end destination and starting point will always be worth it.
#1BillionBookings
Authors: Mehr Javed & Mukund Kabra
Campaign Execution: Mohamed Ayman Fathy, Reem Ragab & Mehr Javed
Creatives: Hazem Samra, Ahmed El Raggal, Mostafa & Omar
