Pendo helped Emburse's Customer Success team move from reactive firefighting to proactive, intelligence-driven account management.
3.1×
Faster time-to-action on at-risk accounts
224
At-risk accounts identified automatically
0
Additional CSMs hired
Emburse's Customer Success team had more data than they knew what to do with. That was the problem.
By the time a renewal was at risk, the customer was already frustrated. Data existed — NPS, health scores, product usage, billing signals — but it lived in four different places. Stitching it together manually meant risk was visible only after customers churned. CSMs were firefighting instead of planning.
NPS, health scores, product usage, and billing signals all lived in separate tools. No single view existed to connect the dots.
Stitching signals together manually meant churn risk became visible only after it was already too late to intervene.
Without prioritization intelligence, every account felt equally urgent — which meant no account got the right level of attention.
Leadership had no forward visibility into which accounts were at risk, making revenue forecasting a guessing game.
“I was spending my days explaining surprises to leadership instead of preventing them.”
Kelly Causey, VP of Customer Success @ Emburse
Before any tooling or model-building, Emburse made a philosophical shift. They established an internal standard: if a customer is heading toward churn, it should never feel sudden to the team.
That sounds obvious, but operationalizing it forced them to ask an uncomfortable question: Why were they still surprised when a customer churned?
The answer came down to a lack of connected, actionable signals. No one was doing anything wrong — the system just wasn't built to see ahead.
Emburse's
Advice
Define what "no surprises" means for your team. What does early churn risk look like? What would have to be true for your CSMs to know about it 90 days before renewal instead of 9? Determine this and work backward to fix it.
Like most CS teams, Emburse had health scores. But health scores are a snapshot: they tell you how a customer is doing right now, not where they're going.
Kelly's team reframed the question, driving her team's work from "How is this customer feeling right now?" to "What is likely to happen next?". Pendo Predict is built around that second question, modeling trajectory instead of just current state.
That shift changed everything downstream: how they built their model, what signals they prioritized, and what actions they tied to each output.
Emburse's
Advice
Audit how your team is currently using health scores. Are they triggering action, or just living in a dashboard? If a CSM has to log in somewhere to check health, it's not working at scale.
Emburse had NPS, CES, product usage, and billing data — yet none of it talked to each other. Their first real focus was creating a unified customer journey model that ingested all of it.
Their initial health score models ran on about 15–20 signals. Once they rebuilt with a predictive model that could ingest hundreds of variables across Pendo usage, CRM, support, and marketing data, they scaled to approximately 700 predictors.
They didn't need more data. They needed connected data modeled toward a specific outcome.
“We had enough data. We just needed certainty earlier.”
Vikas Sharma, Director of Digital CX Intelligence @ Emburse
Emburse's
Advice
List every signal you have on customer health across all systems. Then ask: which of these are actually connected? Which live in silos? A unified view requires intentional integration. Start with the highest-signal data closest to churn.
Before building any data model or committing to Predict, Emburse defined what "good" looked like. They evaluated every predictor against four criteria:
Accuracy and breadth — how many predictors could they actually use to build a meaningful model?
Were the signals transparent, coherent, and consistent across systems? Did they align with existing health scores?
Did the predictions fit how the CS team actually runs plays today?
Could a CSM look at an output and know immediately what to do? Or would they need to interpret it first?
Emburse's
Advice
Run your current signals through this same filter. You may have 30 data points feeding a health score, but are they all passing the actionability test? Cut what doesn't tell the CSM what to do next.
This is where most teams drop the ball. They build a great model and then surface it in a separate tool — but CSMs don't go there, and those insights die buried.
To fix this, Emburse used Predict to push all risk scores directly into Salesforce alongside account information, health scores, and renewal data. CSMs got one unified view of a customer's health (and why) — no context switching required.
“If someone has to log into a separate system to check health, that's not going to work at scale. We needed it inside the workflow.”
Kelly Causey, VP of Customer Success at Emburse
Once Emburse had reliable predictions, they built a routing model using Pendo to determine which accounts need human intervention and which ones don't.
High risk → human CSM engagement, executive involvement if needed
Medium risk → assisted digital motion (guided in-app experience, targeted outreach)
Low risk or expansion signal → automated play, or proactive growth conversation
Emburse's
Advice
Define your routing logic before you have the model. What does a high-risk account look like? What's the right motion for each tier? When you have the predictions, you should be ready to act immediately.
A predictive model is only as good as the feedback it gets. With Predict, Emburse built in a qualification step: when a CSM sees a flagged risk, they confirm whether it's real. That feedback continuously retrains the model.
They also tied outcomes to business metrics — like average customer health before and after, upsell rates, and renewal predictability — to measure real impact over time.
“Standardize before you automate. Define the actions, define the guardrails, then build the feedback loop. That's what keeps you improving.”
Vikas Sharma, Director of Digital CX Intelligence at Emburse
Emburse's
Advice
Build a simple qualification workflow for your CSMs. Even a thumbs-up/thumbs-down on a predicted risk gives the model something to learn from.
Emburse didn't just improve a metric — they changed how their entire CS team operates.
Cost to serve (down from 3–4% ARR)
By replacing manual triage with predictive routing, Emburse cut their cost to serve nearly in half — without reducing team size.
Predictive signals in the model
Starting from just ~20 signals, the team scaled to 700+ predictors — turning fragmented data into a unified early-warning system.
New CSMs hired to handle the volume
Intelligent prioritization let the existing team cover more accounts with the same headcount — no additional hiring required.
Days of forward visibility on renewals
Instead of reacting to churn signals at 30 days, Emburse's team now sees risk coming 90 days out — enough time to actually intervene.
“If intelligence isn't changing what you do, it's just noise.”
Kelly Causey
VP of Customer Success @ Emburse
Earlier than you think, and the trigger isn't an ARR number.
The underlying work — defining what "healthy" looks like, mapping the customer journey, and capturing intelligence at every stage — is worth doing regardless of where you are in your maturity.
