Synopsis
Fastloop built Enhanced Lifetime Value (“eLTV”) models – a net-new data solution that helped Hootsuite identify new targeting capabilities for better customer experiences and personalized offers. This was built leveraging first party data and learnings across a breadth of Google advertising platforms.
Highlights
New lifecycle metrics now live in Google Analytics
New customer opportunities identified with machine learning techniques
Generated 10+% increase in ROAs with improved bidding strategy
Challenges
Marketers are always looking for an edge. Hootsuite wanted to see how they can spend advertising money more efficiently by targeting customers with the highest lifetime value. Additionally, the company is keen to incorporate ML/AI solutions that would help with:
Core challenges include:
Teams at Hootsuite were spending a significant amount of time pushing their eLTV data to Google Ads as part of a highly manual process.
Integrating additional data sources to existing models would help understand which customers moved from pre-trial, to trial to having high lifetime value.
Hootsuite sought highly transparent machine learning models that were understandable to the business, enabling fast time to insights.
Solution
Fastloop expanded Hootsuite’s input data to include digital engagement data from Google Analytics and built a machine learning approach to eLTV for customers from across the sales journey. This drastically reduced the manual effort previously required to ingest eLTV outputs into activation channels as part of the ETL process.
Key components of the solution:
Google Analytics and application data used to empower eLTV models
Hit-level user engagement data is collected and ingested into BigQuery with Google Analytics’ BigQuery Export functionality.
BigQuery analyzes 12 months of Google Analytics data
By showing how data was queried and analyzed in mere seconds, Fastloop demonstrated the high performance and capabilities of BigQuery.
Review and improve existing LTV models with Google Colab
Understanding the existing model and identifying areas of improvement using newly accessible data in BigQuery Machine Learning.
Three models created to understand customer journey
By creating models for pre-trial, trial and post-trial customers, trends were identified. Notably, post-trial customers of a target eLTV could be extrapolated back to determine predictive cues of valuable customers at the pre-trial level.
Build momentum for marketing to further explore data
Pulling the enriched dataset into Looker for hands-on analytics enables additional insights through custom visualizations.
Automate pushing of eLTV data
By building an automated process of pushing user audience segments back into Google Analytics 360 with ML scoring, audience activation was made possible based on the eLTV models.
Improved targeting, bidding and personalization
New retargeting opportunities surfaced based on eLTV across Pre-Trial, Trial and Post-Trial customers; new applications for improving customer experiences and tailoring offers based on segment cues (e.g., LTV, company size); and enhanced personas based on segments and better measurement.
Custom Looker dashboards
All metrics backed by data models have been made available via Looker. This allows teams at Hootsuite to freely explore data and get insights around KPIs.
Results and Impact
After pioneering a new customer metric using fresh data and machine learning capabilities and surfacing key business insights through Looker, Marketing Operations at Hootsuite have become more efficient.
Through integrations, data activation, and orchestration across Google’s suite of tools, Fastloop helped Marketing Operations at Hootsuite unlock efficiencies and reduce unwanted manual tasks.
Future Outlook
The new eLTV models are a foundation for continuous improvement. As various teams at Hootsuite explored additional models, Fastloop ensured they were prepared to navigate the data-driven solutions delivered and implement further improvements.
Machine Learning Monitoring Dashboards
Created dashboards to help show how future retraining or changes to the models would impact key model performance metrics. These can also identify when improvements may be required.
Post-Trial Model Ready to Include Additional Data
By including more granular revenue data and creating a Churn Model, the team will have the capability to predict remaining payments, create more accurate target eLTV scores and improve the training on all models.