Healthcare Staffing Company saves $1M in marketing spend with Hightouch, Snowflake, and Fivetran
The Results
- Over $1M saved annually in marketing spend
- $520,000 per year in cost savings
- 360% YOY growth between 2020 and 2021
- Engineering time reduced from weeks to days
- Controlled experiments per year increased from ~6 to 50+
- Increased new nurse retention rate by 70%
- Reduced canceled shifts from new nurses by 25%
- Prevented COVID-19 transmission among nurses and patients
- Eliminated CSV files
The Challenge
- Improving query performance, establishing a single source of truth, and unlocking data access
- Delivering personalized experiences to thousands of users to drive engagement and ongoing satisfaction
- Building a data stack that could meet the elastic and growing requirements of various business teams
- Optimizing facility/nurse matching based on changing demands and covid symptoms
The first five to ten shifts for a new customer are extremely important. A single bad experience can result in losing a customer indefinitely. Since the availability and market demand of nurses are constantly changing, prices and shifts need to update in real-time within the app. Through Covid, data accuracy was paramount.
We faced a real risk of propagating Covid to the most vulnerable people if our data was not trustworthy. Hightouch, Snowflake, and Fivetran are absolutely vital to our core data operations.
VP of Data Science
•
Healthcare Staffing Company
The Solution
Snowflake
With a decentralized data stack built on top of various MySQL databases, the data team was struggling with analytics performance. A single query could create production-level incidents or even bring down the company’s reporting tool.
We’d try to run these really big, expensive queries, and our engineering team would just look at us and wonder why we were sucking all the life out of our databases.
VP of Data Science
•
Healthcare Staffing Company
They needed a centralized platform to consolidate all of its disparate data sources into a single location. In search of a scalable solution that could meet the company's needs, the team implemented Snowflake.
There were some tables that we couldn’t even touch in MySQL. With Snowflake, it’s never been a problem. We don’t even have to upsize our compute because Snowflake is so performant.
VP of Data Science
•
Healthcare Staffing Company
With Snowflake, the data team is no longer at risk of damaging application databases and the analytics team can run large queries without butting heads against the engineering team. Snowflake saves the engineers countless maintenance hours–which translates into an estimated $520,000 in cost savings annually.
Using dbt with Snowflake, the engineers have also been able to create an analytics backend that is searchable, tested, thoroughly documented, and obeys a strict set of naming conventions so that data is easy for end users to find and understand. Every dashboard and algorithm goes through the dbt curated tables so that data-quality issues are monitored and easy to remediate.
The fact that all of this data is readily available in Snowflake, means that the company can generate optimal prices and shifts based on changing market dynamics in real-time and this is only possible with Snowflake.
Fivetran
However, there are some datasets that are really tricky due to third-party API constraints. Rather than spending countless hours trying to figure out how to pipe this data into Snowflake, the team uses Fivetran’s fully managed connectors to ingest data from Marketo, Pendo, and SurveyMonkey.
Fivetran not only saves the engineers months of engineering time but it also makes it easy for the data team to join various tables and datasets together. With Fivetran, the data can scale as needed every time the company adds a new data source.
It could take weeks or months just to get data into Snowflake and we’d spend even more time later on anytime an API changed. With Fivetran, we have access to our data immediately, and we don’t need to have senior engineers managing in-house pipelines. That’s a huge cost saving for us.
VP of Data Science
•
Healthcare Staffing Company
Hightouch
With so much detailed customer data now available in Snowflake, the business teams wanted to activate that data in downstream business tools, build personalized experiences, and mitigate the risk of Covid for both nurses and facilities. Sending data downstream, however, was challenging, and building Reverse ETL pipelines was no easy feat.
We used to have a pod of integration engineers whose sole focus was piping data out of our MySQL databases to all of our various destinations. Now we’re able to use Hightouch for 90% of our use cases, and this has freed up our integration team to focus on other things.
VP of Data Science
•
Healthcare Staffing Company
Before Hightouch, every time the marketing team had a question, they had to request a manual CSV file. Sending personalized emails in Marketo was basically impossible and retargeting users in ad platforms was a nightmare. Using Hightouch Audiences, a visual segmentation builder, the business teams are able to self-serve and build/manage cohorts in a non-technical way.
One-third of our Hightouch users are on business teams like marketing, support, and sales. They’re able to build audiences without ever having to interact with the data team. Hightouch is powering audiences for emails, texts, in-app messages, and even ad retargeting across Google and Facebook.
VP of Data Science
•
Healthcare Staffing Company
The app is the lifeblood of the business, so building personalized experiences for each individual is key to keeping users on the app. Using Hightouch to power popups and surveys has helped to increase controlled experiments by 700 percent and encourage more nurses to complete shifts.
We ran an experiment around first shift bonuses and followed different test groups through the application process to see how much the bonus influenced shift completion rate. Without Hightouch this experiment cost us $1.3M. With Hightouch, we’re able to intelligently target applicants and achieve the same incremental gains at an annual cost of $260k.
VP of Data Science
•
Healthcare Staffing Company
The company also uses Hightouch to power real-time Slack notifications to notify channel members so they can intervene faster.
Before Hightouch, we had all of this janky code for Slack alerts that would break frequently, and only one person could fix it. Now we spin up a Slack alert in minutes, and it never breaks.
VP of Data Science
•
Healthcare Staffing Company
Hightouch has eliminated the manual process of uploading CSVs to Salesforce. Hightouch also allows the company to get the most out of their predictive algorithms. The data team maintains models for predicting customer lifetime value, cancellation events, churn likelihood, and local demand and supply metrics.
It’s really easy to build bridges to nowhere in data science. We have a model to predict customer churn. That’s great, but now what? Hightouch makes it easy for us to take action based on these models. We send a churn-prediction score to Marketo, Pendo, and Salesforce, and teams throughout the business take action.
VP of Data Science
•
Healthcare Staffing Company
What’s Next?
The company is always working to improve and maintain their data infrastructure. The past year has paved the way for some huge improvements, but there’s always more to do. Thanks to Hightouch, Snowflake, and Fivetran, this healthcare company is able to surface more insights, and the company has many green-field projects slotted to mitigate clinical risk, optimize shift-related marketing, and make it easier for nurses to find shifts that match their interest.
Hightouch, Snowflake, and Fivetran are a force multiplier. Everyone is more effective with our new data stack. Word got out that we can send data from our data platform to just about anywhere, and our business stakeholders have lined up.
VP of Data Science
•
Healthcare Staffing Company