Expansion Revenue Prediction
“Which accounts will upgrade?”
Book a demo and get a free trial of the full platform: research agent, fine-tune capabilities, and forward-deployed engineer support.
By submitting, you accept the Terms and Privacy Policy.
Loved by data scientists, ML engineers & CXOs at
A real-world example
Which accounts will upgrade?
Top SaaS companies achieve 130%+ net revenue retention through expansion. But CSMs waste 60% of upsell outreach on accounts that are not ready to expand, creating friction. A $200M ARR company where expansion outreach has 5% success rate leaves $30M in accessible expansion revenue on the table. The expansion signal sits in the intersection of seat utilization rates, feature adoption velocity, billing trends, and champion engagement patterns.
Quick answer
Expansion revenue prediction works best when you connect seat utilization, cross-department adoption, billing overage trends, and peer-account expansion patterns in a relational model. Accounts at 85%+ seat utilization where 3+ departments have adopted the product and where similar accounts recently expanded convert at 10x the base rate. SAP SALT benchmark shows relational graph ML achieves 91% accuracy vs 75% for XGBoost on flat tables.
Approaches compared
4 ways to solve this problem
1. Seat utilization thresholds
Flag accounts approaching their seat limit (>80% utilization) and route them to CSMs for expansion conversations.
Best for
Catches the most obvious expansion signals. Simple to implement with billing data alone.
Watch out for
Reaches only the 15-20% of accounts at capacity. Misses accounts that are ready to upgrade to a higher tier or add new product modules without being at seat capacity.
2. CSM-driven pipeline
CSMs identify expansion opportunities through quarterly business reviews, relationship signals, and manual account analysis.
Best for
Captures qualitative signals (executive excitement, budget discussions, new initiatives) that quantitative models cannot see.
Watch out for
Does not scale. CSMs managing 50-80 accounts prioritize renewals over expansion. Subjective judgment misses data-driven signals like cross-department adoption velocity.
3. XGBoost on account metrics
Train a classifier on account-level features: seat utilization, feature adoption count, NPS score, ticket volume, and billing history.
Best for
Good baseline that combines multiple signals into a single score. Works when you have enough historical expansion events to train on.
Watch out for
Treats each account independently. Cannot detect peer-account expansion waves: when similar-sized accounts in the same vertical expand, peer accounts follow within a quarter. This cross-account signal is invisible to flat models.
4. KumoRFM (relational graph ML)
Connect accounts, users, usage metrics, billing data, and feature adoption into a graph. The GNN learns cross-department adoption velocity, billing overage trajectories, and peer-account expansion patterns.
Best for
Highest conversion rate. Captures the peer-account expansion signal and cross-department adoption depth that CSMs and flat models miss.
Watch out for
Requires user-level data with department tags and enough expansion history for training. New products without expansion history need alternative approaches.
Key metric: SAP SALT benchmark: relational graph ML achieves 91% accuracy vs 75% for XGBoost on flat tables in expansion propensity tasks.
Why relational data changes the answer
Expansion is a multi-table signal. Seat utilization tells you the account is growing, but the reason it is growing lives in cross-department adoption (user table), billing overages (billing table), feature adoption velocity (feature events), and peer-account behavior (cross-account patterns). A flat table with 'seat_utilization = 95%' does not tell you that three new departments started using the product last month, that API usage grew 18% MoM, and that four of six similar accounts recently expanded.
Relational models connect these tables and learn that the combination of high utilization, cross-department spread, rising overage charges, and peer-account expansion creates a 10x conversion signal. SAP SALT benchmark shows relational graph ML achieves 91% accuracy vs 75% for XGBoost on flat tables for propensity tasks. For a $200M ARR company, focusing expansion outreach on the top-20% of accounts identified by the model captures $30M in revenue that scattered outreach leaves on the table.
Identifying expansion accounts from seat utilization alone is like identifying a city ready for a new store location by counting current foot traffic. You miss that three new office buildings just opened nearby (cross-department adoption), the parking lot is overflowing (billing overages), and your competitor just opened a location across the street because they saw the same growth signals (peer-account expansion). The full picture requires connecting multiple data sources.
How KumoRFM solves this
Graph-learned product intelligence across your entire account base
Kumo connects accounts, users, usage metrics, billing data, and feature adoption sequences into a graph where expansion signals propagate through the account network. It learns that accounts at 85%+ seat utilization, where 3+ departments have adopted the API integration, and where the finance admin has viewed the billing portal 4+ times in 30 days expand at 10x the base rate. The model captures cross-account expansion patterns: when similar-sized accounts in the same vertical expand, peer accounts follow within a quarter.
From data to predictions
See the full pipeline in action
Connect your tables, write a PQL query, and get predictions with built-in explainability — all in minutes, not months.
Your data
The relational tables Kumo learns from
ACCOUNTS
| account_id | name | arr | seats_purchased | plan_tier |
|---|---|---|---|---|
| ACC101 | DataFlow Inc | $96,000 | 40 | Growth |
| ACC102 | RetailCo | $180,000 | 80 | Enterprise |
| ACC103 | Startup Labs | $18,000 | 10 | Starter |
USERS
| user_id | account_id | department | last_login | role |
|---|---|---|---|---|
| U101 | ACC101 | Engineering | 2025-03-02 | Admin |
| U102 | ACC101 | Marketing | 2025-03-01 | User |
| U103 | ACC101 | Sales | 2025-03-02 | User |
USAGE_METRICS
| metric_id | account_id | month | active_seats | api_calls |
|---|---|---|---|---|
| UM01 | ACC101 | 2025-02 | 38 | 45,000 |
| UM02 | ACC101 | 2025-01 | 35 | 38,000 |
| UM03 | ACC102 | 2025-02 | 62 | 12,000 |
BILLING
| billing_id | account_id | date | amount | overage |
|---|---|---|---|---|
| BL01 | ACC101 | 2025-02-01 | $8,000 | $450 |
| BL02 | ACC101 | 2025-01-01 | $8,000 | $200 |
| BL03 | ACC102 | 2025-02-01 | $15,000 | $0 |
FEATURE_ADOPTION
| adoption_id | account_id | feature | first_used | monthly_events |
|---|---|---|---|---|
| FA01 | ACC101 | API v2 | 2025-01-10 | 15,000 |
| FA02 | ACC101 | SSO | 2024-12-01 | 800 |
| FA03 | ACC101 | Custom reports | 2025-02-15 | 120 |
Write your PQL query
Describe what to predict in 2–3 lines — Kumo handles the rest
PREDICT BOOL(ACCOUNTS.EXPANSION_EVENT, 0, 90, days) FOR EACH ACCOUNTS.ACCOUNT_ID WHERE ACCOUNTS.PLAN_TIER != 'Enterprise'
Prediction output
Every entity gets a score, updated continuously
| ACCOUNT_ID | CURRENT_ARR | SEAT_UTIL | EXPANSION_PROB_90D |
|---|---|---|---|
| ACC101 | $96,000 | 95% | 0.82 |
| ACC102 | $180,000 | 78% | 0.28 |
| ACC103 | $18,000 | 60% | 0.05 |
Understand why
Every prediction includes feature attributions — no black boxes
Account ACC101 -- DataFlow Inc, $96K ARR
Predicted: 82% expansion probability within 90 days
Top contributing features
Seat utilization trend
95% and rising
29% attribution
Cross-department adoption
3 departments active
23% attribution
API usage growth (MoM)
+18% increase
20% attribution
Overage charges (last 60d)
$650 total
16% attribution
Peer account expansion rate
4 of 6 expanded
12% attribution
Feature attributions are computed automatically for every prediction. No separate tooling required. Learn more about Kumo explainability
PQL Documentation
Learn the Predictive Query Language — SQL-like syntax for defining any prediction task in 2–3 lines.
Python SDK
Integrate Kumo predictions into your pipelines. Train, evaluate, and deploy models programmatically.
Explainability Docs
Understand feature attributions, model evaluation metrics, and how to build trust with stakeholders.
Frequently asked questions
Common questions about expansion revenue prediction
How do you predict which B2B SaaS accounts will expand?
Connect seat utilization, cross-department adoption, billing data, feature usage, and peer-account patterns in a relational model. The strongest expansion signals are cross-department adoption (3+ departments active), seat utilization above 85%, and rising billing overages. When combined with peer-account expansion patterns, the model identifies accounts ready to expand with 10x better accuracy than manual CSM assessment.
What is net revenue retention and how does ML improve it?
NRR measures revenue from existing customers including expansion and contraction. Top SaaS companies achieve 130%+ NRR. ML improves NRR by identifying expansion-ready accounts and routing CSMs to the right conversations at the right time. Targeted outreach based on graph ML converts at 10x the rate of untargeted outreach, turning CSM time into revenue.
What data signals predict SaaS account expansion?
Five signals matter most: seat utilization (>85%), cross-department adoption (3+ departments active), feature adoption velocity (new features being adopted monthly), billing overages (consistently exceeding current tier limits), and peer-account behavior (similar accounts expanding recently). The peer-account signal is only visible with graph ML across the account base.
How do you avoid wasting CSM time on non-expansion accounts?
CSMs waste 60% of upsell outreach on accounts that are not ready. Graph ML identifies the 20% of accounts most likely to expand with 10x better precision. By focusing CSM expansion conversations on model-identified accounts, conversion rates jump from 5% to 15-20%, and CSMs spend their time on accounts where expansion conversations are welcome.
What is the ROI of expansion revenue prediction?
A $200M ARR company where expansion outreach has 5% success rate leaves $30M in accessible expansion revenue on the table. Graph ML focuses outreach on the top-20% of accounts, capturing that $30M with 10x better conversion rates. The model also prevents expansion fatigue by not approaching accounts that are not ready.
Bottom line: A $200M ARR SaaS company that focuses expansion outreach on the 20% of accounts most likely to upgrade captures $30M in additional revenue with 10x better conversion rates. Kumo identifies expansion-ready accounts through seat utilization, cross-department adoption, and peer-account signals that manual health scoring misses.
Related use cases
Explore more B2B SaaS use cases
Topics covered
One Platform. One Model. Infinite Predictions.
KumoRFM
Relational Foundation Model
Turn structured relational data into predictions in seconds. KumoRFM delivers zero-shot predictions that rival months of traditional data science. No training, feature engineering, or infrastructure required. Just connect your data and start predicting.
For critical use cases, fine-tune KumoRFM on your data using the Kumo platform and Research Agent for 30%+ higher accuracy than traditional models.
Book a demo and get a free trial of the full platform: research agent, fine-tune capabilities, and forward-deployed engineer support.
