For an MVNO, a shared data pool can be an attractive product. Instead of giving every subscriber a separate data allowance, the operator allows a group of users to consume from a common allowance. One family member might use 40 GB while another uses 5 GB, but both draw from the same pool.
The commercial challenge is deciding how large that pool should be, how much capacity to purchase from the host MNO, and what happens when usage is significantly different from expectations.
This is where AI can help MVNOs optimize shared data pools, reduce wholesale overage costs, and improve the profitability of data plans. Rather than relying only on historical averages and fixed provisioning rules, an operator can use usage patterns to forecast demand, identify inefficient allocations, and adjust its commercial strategy before costs get out of control.
Research into MVNO data-plan optimization has already explored how subscriber-level usage prediction can inform the amount of wholesale capacity an operator purchases. The same principle provides a useful foundation for AI-assisted shared-pool optimization.
Why Shared Data Pools Are Difficult to Manage
Shared data plans create a mismatch between how operators purchase capacity and how subscribers consume it. An MVNO might purchase a wholesale data allowance for a family plan, business account or group of IoT devices. The operator then sells access to that pool at a retail price, expecting that not every subscriber will consume their full individual allowance.
The economics depend on actual usage distribution. If the operator provisions too much capacity, customers may pay for data they never use, while the MVNO carries unnecessary wholesale costs. If the pool is too small, subscribers can hit their limits early, incur overage charges or experience throttling, depending on the tariff.
Consider a business account with 100 SIMs and a shared allowance of 1 TB per month. Total usage might usually sit comfortably below the limit, but a few heavy users could push the account close to its threshold. An average based on previous months might miss a sudden change in employee behavior or an increase in connected-device traffic.
The challenge isn’t simply predicting total consumption. It’s understanding who consumes the data, when they consume it, and how that behavior affects the economics of the shared pool.
How AI Can Forecast Shared Data Consumption
AI models can analyze historical usage records, subscriber activity, tariff details, seasonal patterns and recent changes in consumption to estimate future demand. Instead of forecasting only the total allowance required by a customer, the model can estimate usage at several levels: individual subscriber, group, tariff, enterprise account, and overall MVNO portfolio. This helps operators distinguish between predictable consumption and accounts that are likely to experience unusual demand.
For example, an MVNO might identify a business account whose data consumption has increased steadily over several weeks. The system could forecast when the account is likely to exhaust its shared allowance and alert the commercial team before the limit is reached.
The forecast should include uncertainty rather than presenting one number as a guaranteed outcome. A pool expected to consume 900 GB might reasonably need a different operational response if the likely range is narrow than if usage could plausibly exceed 1.2 TB. That distinction helps the operator balance wholesale costs against the risk of service disruption or unexpected charges.
Optimize Pool Sizes Instead of Overprovisioning
Many operators manage shared pools using fixed thresholds or conservative assumptions. These rules are easy to implement, but they don’t always reflect actual consumption patterns. AI can help estimate how much capacity each pool needs by considering historical demand, growth trends, usage variability and the cost of exceeding the allowance.
A low-variance pool with stable consumption may need less safety margin than a pool serving customers whose usage changes dramatically from month to month. An optimization engine can compare alternative pool sizes and estimate their commercial impact. It can then recommend whether the operator should retain the current allowance, increase it, reduce it or move the customer to a different plan.
TelcoEdge Inc. describes an AI Pool Calculation capability on its platform that is designed to calculate data and voice pool sizes across plans and rebalance shared pools. This is a vendor-described capability; operators should validate its performance against their own usage records and wholesale agreements before relying on automated recommendations.
Identify Heavy Users Before They Affect Everyone
Shared pools can become difficult to manage when a small number of subscribers account for a disproportionate share of consumption. An AI model can identify these patterns and distinguish between normal heavy usage and behavior that deserves investigation.
For example, a business account might have 200 SIMs, with most consuming between 2 GB and 5 GB per month. If several devices suddenly begin consuming 30 GB each, the system could investigate whether the increase reflects legitimate business activity, a software update, a device configuration problem or unexpected usage.
The operator can then decide whether the account needs a larger pool, a different tariff or additional usage controls. This approach is more useful than applying the same restriction to every subscriber. It gives the MVNO a chance to address the underlying problem while preserving the customer experience for the rest of the group.
Connect Pool Optimization With Real-Time Charging
Forecasting is only one part of the solution. The billing and charging platform must also be able to apply the relevant rules when usage changes. For prepaid or tightly controlled data plans, the operator may need to monitor pool consumption continuously and take action as the allowance approaches its limit. For postpaid business accounts, it may be sufficient to generate alerts and calculate expected overage costs.
These decisions should be based on the commercial agreement and service requirements, not on the AI model alone. 3GPP charging specifications distinguish online charging, which can influence authorization for network usage, from offline charging, where usage information is processed after consumption. Both approaches can be relevant to an MVNO’s charging architecture.
Optiva’s Charging Engine, for example, describes real-time charging, policy integration, and open APIs for integration with other systems. Such capabilities are relevant when an operator wants to connect consumption forecasts with charging and policy workflows.
Why Wholesale Costs Must Be Part of the Model
A shared pool can look healthy from the retail side while still putting pressure on the MVNO’s margins. Suppose an MVNO sells a business a fixed-price data package. The customer stays within its retail allowance, but its consumption pattern drives higher wholesale costs than expected under the MNO agreement. The MVNO may face declining margins even though the customer hasn’t technically exceeded the plan.
An effective optimization model therefore needs to consider both sides of the commercial equation: retail revenue and wholesale cost. This is particularly important when wholesale agreements contain volume tiers, overage rates, usage commitments, or other commercial conditions.
The operator should be able to evaluate whether changing the pool size, adjusting the tariff or moving a customer to another package would improve profitability without creating an unacceptable customer impact. Charging and monetization capabilities relevant to these broader commercial requirements, while Telgoo5 operates in the telecom billing and charging market. Their suitability for a particular optimization workflow depends on the available interfaces, data model, and charging configuration.
What Data Does the AI Need?
AI pool optimization is only as reliable as the information behind it. An MVNO should ideally combine usage records, subscriber and account relationships, tariff configuration, pool limits, historical overages, wholesale pricing, and relevant service events. It also needs consistent definitions for what counts as consumed, reserved, rated, and billable data.
If the charging platform and analytics environment disagree about current pool consumption, the model can produce recommendations that are commercially incorrect.
A practical implementation should therefore begin with data validation and reconciliation before introducing automated pool changes.
MATRIXX Software has focused on real-time charging and monetization technology and is now part of Amdocs following its acquisition. For operators evaluating charging platforms and analytics integrations, the key consideration is whether the architecture supports accurate consumption data, flexible charging rules, and controlled execution.
How MVNOs Should Introduce AI Pool Optimization
The safest approach is to begin with recommendations rather than fully automated changes. An MVNO can first use AI to forecast demand, identify underused pools and flag accounts likely to exceed their allowances. The commercial team can compare these recommendations with actual invoices, wholesale costs and customer behavior.
Once the model demonstrates reliable performance, the operator can automate low-risk actions, such as issuing alerts or recommending a better-fit tariff. Changes to purchased wholesale capacity, customer allowances or network policy should remain subject to appropriate business rules and approval.
Performance should be measured through practical metrics, including forecast error, pool utilization, wholesale overage costs, gross margin, customer complaints and the number of unnecessary plan changes. These measurements help establish whether AI is producing real commercial value rather than simply generating more analytics.
The Real Opportunity Is Better Pool Economics
AI can help MVNOs move beyond fixed data allowances and simple historical averages toward more informed decisions about shared data pools.
By forecasting consumption, identifying heavy users, estimating wholesale exposure and recommending better pool sizes, operators can improve capacity planning without treating every subscriber or enterprise account the same way.
The technology won’t eliminate unpredictable usage, and it can’t compensate for poor charging data or unclear wholesale agreements. It can, however, give the operator more time to respond before a consumption problem becomes a margin problem.
For MVNOs, the goal isn’t to create the largest possible shared data pool. It’s to maintain enough capacity to serve customers reliably without paying for more wholesale data than the business needs. That’s where AI-assisted pool optimization can make a measurable difference.