Your MNO cost, before the invoice arrives.
Invoice Prediction builds a daily running accrual of your projected MNO wholesale cost — broken down by plan, usage type, and pool allocation. 18+ days before the invoice lands. Under 5% error rate.
Daily
Update cadence
18+
Days ahead
<5%
Invoice prediction error
⚠ Up 4.1% vs April variance within threshold
$48,240
Your single largest cost line. Invisible until it arrives.
For most MVNO operators, the MNO wholesale invoice is the largest variable cost in the P&L — and the one with the least advance visibility. Invoice Prediction changes that.
Invoice Prediction AI 
What Invoice Prediction actually delivers.
Not a report. A continuous, updating view of your wholesale cost — with alerts that trigger when the pattern changes.
Daily projected invoice
A running total of your projected MNO wholesale cost for the current billing cycle — updated with every CDR batch. Broken down by data, voice, SMS, roaming, and overage. Per plan code and usage type.
Variance alerts
When the projected invoice moves outside your defined threshold — up or down — an alert fires immediately. You define the threshold. The alert identifies the plan code or usage category driving the variance and its estimated dollar impact.
Cycle-over-cycle comparison
Projected cost for the current cycle compared against the previous three billing cycles — by category. Identify trends before they compound. Seasonal patterns, subscriber mix changes, and MNO rate impacts visible as they develop.
Under 5% error From billing cycle three onwards.
Four streams of intelligence, updated continuously. Each one addresses a gap that your existing platform doesn't surface.
Tested on held-out billing cycles — months the model hasn't seen during training. The 5% error rate means a $50,000 projected invoice is accurate to within $2,500. Typically achieved by billing cycle 3 and improves as the model accumulates more of your specific CDR history and MNO rate patterns.
The model trains on your CDR data specifically — not industry averages or anonymised operator data. Your MNO rate card, your plan mix, your subscriber behaviour profile are what determine the accuracy.
Cycle 1
First prediction. Directionally accurate, wider error range as the model is still learning your CDR patterns.
Cycle 2
Significant accuracy improvement. Rate card calibration complete. Plan code mapping validated against actual invoice.
Cycle 3
Target <5% MAPE reached. Roaming and overage variance are the most unpredictable elements and become more consistent with sufficient history.
Ongoing
Model retrains monthly. Rate card changes, subscriber mix shifts, and new plan codes are absorbed automatically.
Which operators Invoice Prediction
applies to.
Invoice Prediction is built for the direct MNO wholesale relationship. If your operation sits on an MVNA, CDR Intelligence is the module that applies to your margin — with different inputs and different outputs.
Your own wholesale contract with
T-Mobile, AT&T, or Verizon
Invoice Prediction is built specifically for this operator model. You receive a direct MNO wholesale invoice monthly. That invoice is your largest variable cost and the one with the least advance visibility. Invoice Prediction closes that gap — daily projected cost, per plan code, 18+ days ahead.
Direct MNO operators
Operating on MVNA pricing - per-line
or per-usage
If your wholesale relationship is with an MVNA rather than a direct MNO, Invoice Prediction is still relevant to you — the MVNA absorbs the MNO invoice and bills you based on pool usage.
