Network Capacity Prediction
“Which cell towers will exceed capacity?”
Book a demo and get a free trial of the full platform: data science 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 cell towers will exceed capacity?
A single cell tower outage affects 2,000-10,000 subscribers and costs $50K-$200K in service credits and churn. Carriers spend $8B annually on network upgrades, but 30% of CapEx goes to towers that did not actually need it while capacity-strained towers go unaddressed. Traffic patterns shift with events, construction, and subscriber mobility in ways that static coverage models cannot predict.
How KumoRFM solves this
Graph-learned network intelligence across your entire subscriber base
Kumo connects towers, cells, traffic records, subscribers, and events into a spatial-temporal graph. It learns that when a nearby tower goes into maintenance, traffic redistributes predictably based on subscriber density and handoff patterns. The model captures event-driven spikes (stadium, concert venue), seasonal shifts, and cascading congestion effects across adjacent cells that isolated time-series models miss.
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
TOWERS
| tower_id | location | type | max_capacity_mbps | sectors |
|---|---|---|---|---|
| TWR001 | 40.7128,-74.006 | Macro | 10000 | 3 |
| TWR002 | 40.7589,-73.985 | Small cell | 2000 | 1 |
| TWR003 | 40.7484,-73.985 | Macro | 10000 | 3 |
CELLS
| cell_id | tower_id | band | technology | current_load_pct |
|---|---|---|---|---|
| C001 | TWR001 | n71 | 5G | 72% |
| C002 | TWR002 | B66 | LTE | 91% |
| C003 | TWR003 | n41 | 5G | 45% |
TRAFFIC_RECORDS
| record_id | cell_id | timestamp | throughput_mbps | connected_users |
|---|---|---|---|---|
| TR01 | C001 | 2025-03-02 18:00 | 7200 | 3400 |
| TR02 | C002 | 2025-03-02 18:00 | 1850 | 890 |
| TR03 | C003 | 2025-03-02 18:00 | 4500 | 2100 |
SUBSCRIBERS
| subscriber_id | home_tower | plan | avg_data_gb_day |
|---|---|---|---|
| SUB101 | TWR001 | Unlimited Plus | 1.2 |
| SUB102 | TWR002 | Basic 5GB | 0.3 |
| SUB103 | TWR001 | Unlimited Plus | 2.1 |
EVENTS
| event_id | type | location | date | expected_attendees |
|---|---|---|---|---|
| EVT01 | Concert | 40.7505,-73.993 | 2025-03-15 | 20000 |
| EVT02 | Construction | 40.7128,-74.005 | 2025-03-10 | N/A |
Write your PQL query
Describe what to predict in 2–3 lines — Kumo handles the rest
PREDICT MAX(TRAFFIC_RECORDS.THROUGHPUT_MBPS, 0, 24, hours) FOR EACH CELLS.CELL_ID WHERE CELLS.CURRENT_LOAD_PCT > 50
Prediction output
Every entity gets a score, updated continuously
| CELL_ID | TOWER_ID | CURRENT_LOAD | PREDICTED_PEAK_LOAD_24H |
|---|---|---|---|
| C001 | TWR001 | 72% | 88% |
| C002 | TWR002 | 91% | 105% (OVERFLOW) |
| C003 | TWR003 | 45% | 62% |
Understand why
Every prediction includes feature attributions — no black boxes
Cell C002 -- TWR002, Small cell, LTE
Predicted: 105% predicted peak load (overflow)
Top contributing features
Current load vs capacity
91% utilized
28% attribution
Nearby event (concert, 0.3mi)
20,000 attendees
25% attribution
Adjacent tower maintenance
TWR004 offline
20% attribution
Historical Friday peak multiplier
1.35x
15% attribution
Subscriber growth (last 30d)
+8% in area
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.
Bottom line: A carrier with 50,000 cell sites that predicts capacity overflows 24 hours in advance prevents $20M in annual service credits and targeted churn. Kumo captures event-driven traffic spikes, maintenance cascades, and spatial congestion propagation that isolated tower-level forecasting misses.
Related use cases
Explore more telecom use cases
Topics covered
One Platform. One Model. Predict Instantly.
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 Data Science Agent for 30%+ higher accuracy than traditional models.
Book a demo and get a free trial of the full platform: data science agent, fine-tune capabilities, and forward-deployed engineer support.
