Key Results
- Accelerated plans by six months
- Fivetran is simple to use ‘out of the box’
- More accurate and complete data
- Faster time to reports
- Opened the door to data science
Data Stack
- Pipeline: Fivetran
- Sources: Shopify, Zendesk, ReCharge, PostgreSQL
- Destination: Google BigQuery
- Business Intelligence: Metabase
Fivetran has massively delivered and got us to where we want to be six months quicker than we would have without it.
- James McNeilis, Lead Data Scientist, KatKin
KatKin is the world's first DTC subscription service based in the UK personalising fresh nutrition and wellness for pets. KatKin is vertically integrated with its own in-house manufacturing and supply chain management – and uses technology, data and the best possible pet nutrition expertise for personalisation.
The Challenge: Manual Pipelines and Missing Data
When lead data scientist James McNeilis joined the company, his first job was to centralise data and make it more accessible to the business. The firm’s engineers had built ETL pipelines to pull the data into a PostgreSQL database from multiple applications -- a manual process that was unreliable and prone to errors. He was constantly going back to the data engineering team to clean up and prepare the data -- far from ideal for a web-based start-up that identified accurate and timely customer insights as vital for growth.
A second problem was that product engineers mainly focused on data that improved the application. Information on a customer’s payment was collected, for example, but crucial data on when the next order was scheduled could be missing. Valuable insights would only be achieved if more data was made available more quickly.
The plan was to move transactional data generated by a number of applications into a data warehouse, Google BigQuery, to create a single source of truth. “I needed to focus on doing the analysis and answer questions with data rather than work out how to access it,” said McNeilis. “It became very clear that we were going to have to use a connector solution like Fivetran.”
The Solution: Fast and Simple Setup
McNeilis had previous experience with Stitch Data, but he struggled to make the pipelines work ‘out of the box’ and his issues were compounded as he had difficulty contacting the company for support. By contrast, he signed up for a free Fivetran trial, set up the ReCharge connector with ease and it worked perfectly. McNeilis mentioned:
It was super simple to set up; you don't need a data background to be able to do it. You can point and click on the interface and within an hour you will have your data source and destination synced. You really remove the need for a data engineer.
Customer conversion and retention were the initial focus. KatKin set up pipelines for three sources of transactional data and refreshed them every six hours. Customer service soon began using the data to monitor their performance, and the product team was able to glean insights to influence recipe choices.
McNeilis plans to use more advanced data science to inform strategic business decisions. He is currently running A/B testing on different sign-up methods, and experimenting with the first recipes they present to prospective customers to see if they can increase conversion rates. KatKin also plans to use a Fivetran connector to extract data from Xero, the company’s accounting system, to analyse order values and other financial elements.
The Results: Accelerating the Transition to Advanced Analytics
McNeilis talks about two tiers of Fivetran benefits – what it’s doing for the business and how it has enabled him to do more in his role.
Fivetran has massively delivered and got us to where we want to be six months quicker than we would have without it,” he said. And it’s made my life much easier. I haven't had to worry about building data pipelines, the reliability of the data or things breaking. Everything works straight out of the box.
As an online subscription business providing a high-quality product, data around customer interactions is critical. The first step was enabling fast access to more accurate transactional data to analyse retention and conversion rates. What used to take a member of the management team two hours every Monday morning is now accomplished in an instant.
A retention curve that had to be calculated manually with spreadsheets using data that was almost immediately out of date, has been replaced by an automated process that runs reports in real-time. “We had people spending hours manually assembling reports – that just went overnight when we started using Fivetran.”
He praises Fivetran for excellent service, the comfort of knowing that there is always someone available if there is an issue – a customer support feature notably absent from his Stitch Data experience.
We haven't had any problems with Fivetran, but it's a nice feeling to know they’re there if we need them
KatKin’s company founders quickly bought into Fivetran. They always knew that data would be pivotal for building the business, which is why they recruited data scientists in the first place. “It’s a no-brainer for them. They hired me to use my time analysing the data, not moving it,” said McNeilis. "With the help of Fivetran, we now know what recipes are popular and how often orders are renewed, which is informing the business in multiple ways.”
The Future
In the next six to nine months, McNeilis expects to start to use the data to personalize KatKin's customer experience. An AI-powered recommendation engine that can suggest treats unique for each customer's cat is already in the pipeline. With the availability of clean and reliable data, customer segmentation models and predictive models to catch customers before they churn are also on the horizon. The data science journey at KatKin is only just beginning.
About Fivetran: Our fully managed pipelines and automated connectors for databases, applications, events, files and functions help extract value more quickly from data, accelerating the path to becoming a data-driven business and achieving competitive advantage through insight-based decision making.
.png)
.svg)
.svg)
.svg)