MLBuilds with your dataHigh complexityLevel 2 Refined bespoke
Predicted LTV (pLTV)
A machine-learning model that estimates each customer's expected value over the next 12 months.
What each customer is likely to be worth over the next 12 months, so you know who to invest in.
Who it's for
Businesses making acquisition or retention investment decisions who need a forward value estimate, not just historic spend.
The engagement
What you get
- Predicted-LTV ML model and scoring pipeline
- pLTV attribute written to Braze
- Investment-tier recipes tying spend to predicted value
- Model monitoring and periodic retraining
What Fuse does
- Assemble spend and purchase features and agree the prediction horizon
- Train and validate the pLTV model
- Deploy scoring and write the pLTV attribute into Braze
- Design investment-tier recipes tying spend to predicted value
- Monitor and retrain on an agreed cadence
What we need from you
Data & access
- Purchase and spend history per customer
Your responsibilities
- Provide purchase and spend history per customer
- Confirm the value definition (gross, net, margin) with finance
- Provide Braze workspace access and a scoring feed
Outcomes and proof
- A forward-looking value dial replacing static value tiers
- Marketing spend concentrated on high-predicted-value customers
- A basis for value-based bidding and acquisition targets
Assumptions
- Enough spend history exists to model forward value
- The value definition is agreed and stable
Out of scope
- Margin or cost modelling beyond the agreed value definition
- Acquisition-channel media buying