The network has to learn the crowd

If infrastructure takes years to build but human behavior can change overnight, success should be measured by a network’s ability to adapt in the moment.

Mobile customer experience used to be measured by what reached the consumer’s device, from signal strength and the network generation available to how quickly a 2-minute video could download.

While these metrics are still useful, they are no longer enough on their own to measure success. Coverage remains a precondition, since without it there is no workaround. But the way people use digital services, and connect with each other through them, now asks more of the network.

Alongside the question of where to expand the network, service providers need to ask how to invest in intelligence that extracts more from the network they already have – and compete on experience rather than just rollout and price.

The upload economy puts more pressure on the uplink

The upload economy describes how people and businesses use digital services to create and send data, as well as consume it. Whereas mobile networks were built around an asymmetric model that favored downloads, today’s use cases generate far more upstream traffic.

Uplink is often discussed as an AI story, but the demand behind it is older. As early as 2024, a GSMA Intelligence survey found that 43% of operators ranked improved uplink performance first among their technological priorities for 5G-Advanced, and another 17% ranked it second.

This demand comes largely from everyday services. Social media, for example, has turned a generation of users into content creators, while cloud services constantly ask devices to synchronize and mobile payments need authorizations and confirmations almost in real time.

And while many of these interactions carry very little data, they matter a great deal to the subscriber. So when the uplink is congested, a failed upload or payment is much more visible than its data volume would suggest.

How a Tier One operator adapts its network to live events

Amdocs has worked with a U.S. Tier One operator to differentiate the experience its subscribers get at live events. The solution adds a Customer Experience (CX) layer on top of the Self-Organizing Network (SON) platform and tailors connectivity to live-event conditions.

The operator’s network faced four challenges at live events:

• Extreme device density: Thousands of devices were connecting from a small area.
• Changing resource profiles: Demand in a full venue was very different from the area’s normal baseline.
• Unpredictable traffic patterns: Demand could be estimated, but no one could know exactly what would happen or when. A goal, half-time or the final whistle could rapidly change how thousands of people used the network.
• Greater uplink demand: Fans created and shared large volumes of content during live events. Enterprise traffic, such as payment authorization and broadcast data transfer, was also concentrated in the venue.

For the operator, these same conditions also represented an opportunity. Because subscribers expected to be connected wherever they were, tailoring connectivity to what they were doing during the event could give them a reason to choose the operator over its competitors.

Today, the solution works in three phases:

• Before the event, the CX layer uses public data, historical traffic patterns and analysis of social media signals to prepare the network, for example by adjusting antenna tilt or reserving spare backhaul capacity.
• During the event, crowd-aware optimization runs as a closed loop in real time. Live telemetry feeds a continuously updated demand model, which issues SON parameter updates at intervals of under a minute.
• After the event, the platform rebalances capacity across neighboring sites as the crowd leaves, then gradually returns the stadium’s network configuration to normal.

When an event ends, thousands of people leave the venue at once and use location-sharing and ride-hailing apps on the way out. Demand can move from ordinary to extreme for a few hours, then fall away again.

Each phase responds to what people are doing at that point in the event. This level of specificity is how the operator differentiates a service that would otherwise compete on price.

The same approach could work well beyond live events. With the SON platform as a constant base, a service provider could build a catalog of CX layers, each designed for a different scenario and ready to deploy when needed. The result would be location-specific differentiation across the whole network.

Networks must adapt faster than they can be built

Service providers will increasingly compete on how well they can tailor services to a specific time and place. New use cases make it easier for subscribers to notice that demand has never been uniform, because each one asks something different of the network. For example, a video upload depends on uplink capacity, a live stream on low jitter and an AI application on data moving reliably in both directions.

Physical infrastructure will never change as quickly as the applications running across it, so service providers need an operating model that lets the network adapt to demand without waiting for the next infrastructure cycle.

In Amdocs’ model, infrastructure provides the capacity, the SON platform optimizes how it is used and the CX layer tailors connectivity to each scenario. Automation lets all three respond together as conditions change. The result gives service providers a way to win subscribers on the quality of the experience they deliver and to get more from the network they have already built. The best place to start is the scenario where demand is hardest to predict and subscribers notice failure first, since any improvement there will be the most visible.