How a chatbot was able to deliver an ice-cream seller a sweet ROAS of 44X

8. 10. 2018

For this Milanese ice-cream seller client of ours, the existing way of doing things was tedious and time-consuming. Here is what his everyday work life looked like:

– Publish a local Facebook sponsored post targeting cities and surrounding areas he could deliver the ice-cream
– Received orders via WhatsApp
– Chat 1 to 1 with every single client
– Jump on the car to deliver the order to the customer.

No wonder he was tired of this process! Immediately I saw room to automate the existing process while increasing the number of orders he received.

The solution

There were parts in his existing processes that could be automated through the use of a chatbot. We introduced the idea of experimenting with a chatbot. “It would be like having an extra employee” this is what break the last hesitations the client had and allowed us to get started with the strategy and implementation.

Setting up the chatbot

We started by analysing a dozen order-conversations that had taken place on WhatsApp to try and analyse them and gain as much insight as possible. Based on those conversations, we created frequently asked questions to use in our bot. Besides that, the bot would guide the user through various steps of product selection and collecting delivery information.

We built flows that allow us to:

  1. Automate the upselling process. Based on the number of people the user orders ice- cream for, the Bot automatically proposes, the most suitable container size for the customer. This technique encourages the purchase of the bigger ice cream container and maximises profit.
  2. Through structured flows, we aimed to “recover” the users who have responded negatively to an Upsell by displaying attractive images of cones or biscuits at the right time.
  3. Automate the order process. For our client, it was as if he had an extra employee to process orders on Messenger.
  4. Set up a post-delivery sequence.
    a) As soon as the order was received, the user could share the purchase with their friends. This could help us spread organically.
    b) A couple of hours after the delivery, an automatic message with the direct link to the page (FB or Tripadvisor) will be sent, requiring a short review.
    c) Half an hour from the purchase, the user will be asked if they want to stay updated with our promotions and news.

In addition to the flows above, we used Zapier and a Google Sheet report to record statistics on deliveries automatically. This allows to refine the offer and to have metrics on the performance of the service about the costs of our Facebook Ads as we could study the relationship between delivery turnover and budget for FB Ads.

Also, we created fields to track the number of purchases made by the user. This data can be used for loyalty campaigns and awards for the most loyal customers.

Bot – user communication

The user can access the bot through an advertisement that sends them to the messenger page or “independently” via “Send a message” button on the Facebook Page). By doing so, they will be automatically greeted by the chatbot and interact with it.

Promoting the bot

For promoting the bot, we worked alongside their Communication Agency that provided us with engaging posts both creative and copy-wise.  To ensure that the audience will interact with our bot, we came up with a contest idea. We would ask the audience to help us name out Chatbot, and the winner would be awarded 50 ice creams.

The Facebook campaigns we created

For our prospecting campaigns, we used local targeting relying on ZIP codes of the cities the client was able to deliver the ice cream to. In our posts, we asked them to comment the word “ICECREAM” to access the chatbot sequence.

For the remarketing part, we targeted people who interacted with the page or the post.

Utilising offline channels

To increase the hype about this “New in the city”, I contacted the local magazines to write articles about the client. These helped us with

– Awareness
– Social Proof (articles were shared and sponsored through remarketing campaigns).

Results

The “Find a name” contest post garnered 422 comments with name suggestions. The client chose the three names he liked the most. Due to the high exposure he got, he decided to award the winner 50 ice creams and, additionally two runner-ups with 20 ice creams each. These ice-creams could be redeemed via a card.

Our initial budget of 226,74€ generated 4097€ in sales. The number of orders that came in as a result of our campaign was so great that the client had a difficult time “fulfilling” the orders. This forced us to pause our campaigns, and for the remainder of the summer period, the client worked only with the “organic” traffic coming in. The orders accrued after pausing the campaigns resulted in a total of more than 11k euros, so far, in sales for our client.

Additionally, we achieved:

  • Brought customers with an average purchase of 18€.
  • Out of the people who purchased, we were able to recover 6% of users and convinced them to buy a package of cones or biscuits.
  • 5% of users that purchased ice cream shared their purchase. This exposed us to another, new organic audience.
  • Increased our subscriber base at zero cost to 718 (since everyone that interacted with our bot was added to the subscription list) and out of them, 350 bought at least once.
  • 20% of users have signed up to be contacted for commercial purposes.
  • 15 Facebook Page reviews and 25 reviews on TripAdvisor.
  • 4,400€ came from repeated purchases of the most loyal customers.
Stefano Pisoni on LinkedinStefano Pisoni on Wordpress
I’m Stefano Pisoni, a 31yo Chief Growth Officer at JotURL.com, the all-in-one suite for marketing links (1#f Product of the Day on Product Hunt) and CMO at Lhubagency.com

Comments

Latest posts

how-a-chatbot-was-able-to-deliver-an-ice-cream-seller-a-sweet-roas-of-44x