Flybuys - Automating Campaign Management and Performance Measurement

bdforbes

Team members:

• Ben Forbes, Product Owner, Data Platform
• Louis du Plessis, Data Scientist

Country: Australia

Organization: Flybuys (Loyalty Pacific Pty Ltd)

Flybuys is a rewards program that allows you to collect points with our partners for doing the things you already do – like the weekly shop, or filling up your tank at the servo. Once you’ve got a stash of points, you can redeem them for all kinds of things, like money off your shop, a much-needed getaway, gift cards or a pressie from the Rewards Store.

Awards Categories:

• Best Acceleration Use Case
• Best Data Democratization Program

Business Challenge:

Flybuys, Australia’s most loved loyalty program, has a data product in market known as Unpacked by Flybuys where we work with external clients (brands and advertising agencies) to help define audiences for their digital marketing campaigns, leveraging the rich Flybuys dataset of member attributes and consumption behaviour.

We are able to help these clients target high propensity members, so that these campaigns are more relevant to the audience and give a greater return on marketing investment for our clients. We are also able to measure participation and sales uplift following these campaigns, so that the client can have confidence that their marketing budget has been well spent.

This product offering has been very successful in market, and Flybuys are looking to scale it up. This presents some challenges in terms of operational efficiency, given the effort required by the Unpacked team to work with clients to define campaigns and measure their performance. Each new campaign can have very different and bespoke requirements, in terms of campaign objectives, targeting parameters, and measurement metrics. This requires a great deal of effort to set up and leads to operational inefficiencies.

The way that we translate the clients’ business requirements into data problems can involve complex methodologies and require time from expert data practitioners, but we would ideally like to standardise and automate as much of this as possible so that the experts can focus on higher value work such as improving the offering itself, and the more business-as-usual or operational aspects of the product can be handed off to juniors.

Business Solution:

Dataiku provided a broad range of capabilities that assisted in addressing our scaling challenge with Unpacked.

First and foremost, the use of the Flow to visually lay out the structure of the project, making sequencing and dependencies clear, was instrumental in fostering collaboration between team members and helping to quickly gain a shared understanding of the key steps involved. It also made it easy to inspect intermediate steps and outputs to help debug issues more quickly.

Adapting to ever changing and bespoke campaign requirements from our clients prompted us to leverage a mixture of SQL and Python code to get the flexibility we needed. This is made very easy within a Dataiku project, where different recipes in the Flow can be in different programming languages, with the implementation detail hidden from the user. This allows different personas to use the language they are comfortable with, intermixed within a single project.

The ability to create embedded web apps using Dash allowed the more complex machinery to be developed behind the scenes by an expert practitioner, exposing a simplified interface that a non-expert user could easily operate.

A number of automation Scenarios can be triggered from this user interface, including dataset refreshes and generation of the final output datasets, which are sent straight through to operational systems.

The Dashboard feature is also leveraged to provide basic reporting and validation checks for campaign analysts.

The entire end-to-end analytical workflow is developed, deployed and operated entirely within the single Dataiku platform, which removes friction points of having to switch between disparate tools and environments for different stages of the campaign lifecycle.

Day-to-day Change:

The effort required for day-to-day campaign work has been drastically reduced due to the standardisation and automation of analytics work using Dataiku. See below for further details.

Business Area Enhanced: Analytics

Use Case Stage: Built & Functional

Value Generated:

The effort required for day-to-day campaign work has been drastically reduced due to the standardisation and automation of analytics work using Dataiku. Previously, refreshing targeting segments could take weeks or even months of tedious, manual work inspecting each refresh script. This has now been reduced to an automated, overnight job.

Apart from efficiency, this has led to more consistent results that can be automatically validated to prevent errors. This reduction in effort has freed up the team to spend their time more productively, including the higher value work of doing more to improve the produce offering and better service the clients.

Furthermore, a more consistent approach that removes the need for manual intervention reduces the risk of error and ensures that we are delivering a high quality product to out clients. The efficiency gains found in standardising and automating the analytics pipelines in Dataiku has made scaling up to more clients much easier – analysts don’t spend all their time doing manual analytics work and can do more insightful work for clients leading to improved revenue for Flybuys.

Value Brought by Dataiku:

Dataiku specifically has provided a coherent, end-to-end platform that allows team members with different personas to work closely together using the types of tools and workflows that they are comfortable with.

Complex data transformation pipelines with multiple automation scenarios have been authored through a slick visual interface which strikes a good balance between visual clarity and flexibility. User interfaces embedded within the platform have allowed the efforts of expert practitioners to be packaged up and re-used by non-experts to accelerate the day-to-day delivery of value.

Value Type:

• Reduce cost
• Reduce risk
• Save time

JamesO

amazing work team!

RyanMorris

Fantastic, well done