Experiment Results

[Feb 19] Facebook Messenger ChatBot Experiment Results - Fuel Fiji

Experiment Design Post: Facebook Messenger ChatBot Experiment

Lean Phase: Channels

Assumption: That Facebook’s third-party ChatBot software is a viable platform to automate the ordering process away from PEV’s in-country operations


Orange Light

Number of total ChatBot interactions:


Number of complete Chatbot interactions:


Level of Satisfaction from complete interactions:

All complete interactions had a 5 out of 5 satisfaction rating

Level of assistance required in complete interactions:

No assistance needed for any of the complete interactions

Cost of ChatBot program per month:

$0 FJD

Validated Learning:

Though our results didn’t meet the success metric we selected we still see that the Facebook third party ChatBot software is a potential platform to automate the ordering process away from PEV’s in-country operations.

Page Change

An important note concerning this experiment is that the page was set up with a false Facebook Account under the pseudonym ‘Peter Buka’ in a previous month. This account was shut down by Facebook for being fake and we lost access to the ‘Project Buka’ Facebook page during the time we were collecting data. Therefore a new project Facebook page was created with members of the team as admin. This meant that all likes on the page and previous interactions were lost for the time being. Days later Peter Buka was returned to us but we decided to carry on with the new page to prevent incidents such as that repeating.

Initially it was assumed that the ChatBot would be a way to engage PEV with customers wishing to purchase Buka Stoves. Whilst conducting the experiment and developing the finer points of our business model it became more clear that this would be a tool for Sales Agents to make orders with the ChatBot and stoves be distributed beyond this.

In terms of engaging customers autonomously, it was decided that the ChatBot should simply direct them to a local retailer stocking the stove rather than accept a direct order. The Sales agents seemed to take the role of engaging customers directly, particularly customers with limited or no access to retailers stocking the Buka Stoves.

 Next Move: 

Optimise and extend

The Chatbot was optimised during the experiment to such that sales agents could autonomously communicate their order amount, contact details, any comments and pickup location (information is deposited in the Facebook Ordering System spreadsheet automatically). Then an invoice is created within the drive to be approved and sent back to a sales agent to be paid. Following payment, pickup details would be shared. Based on these alterations, the ChatBot should be tested as a viable distribution channel. Also we feel this should be tested over a longer period of time for a better grasp of the capabilities of our Facebook page.

For more information on how to edit the chatbot and related software for optimisation, see the Facebook Ordering System SOP.

Depending on the outcome of work during today and tomorrow, a future experiment will be posted concerning the testing of the Facebook Messenger ChatBot

edited on 25th February 2019, 00:02 by Georgia Perazzolo

Haziq Ahmed 2 months ago

Status label added: Experiment Results

Reply 0

AbdulKaleem 2 weeks ago

This experiment is really interesting for me because I was searching about the messenger chat-bot and now I want to publish an article for my website thesis writing help in Dubai who provide these type of unique knowledge.

Reply 0