For nearly three decades, connecting a branch office to the corporate network meant one thing: ordering an MPLS circuit. It was the reliable, predictable choice. The private toll motorway of enterprise traffic, where every packet travelled in a guaranteed lane. However, the question businesses ask themselves today is not whether they should still use that motorway, but which traffic still deserves to use it, and which traffic is better off on the open road.

From gold standard to bottleneck

MPLS earned its reputation in an honest format. It delivered guaranteed quality of service, predictable latency and the type of stability that mission-critical applications were built around. The problem was never reliability; it was everything else. MPLS circuits are expensive, rigid, and painfully slow to turn up. Provisioning a new site can take from 30 to 90 days. Meanwhile, the traffic itself changed shape. A decade ago most applications lived in the corporate data centre. Today the majority sit in the cloud, delivered as SaaS. Forcing that cloud-bound traffic to travel back to the data centre over MPLS, only to be handed off to the internet from there. A practice known as backhauling which adds cost, latency and complexity without any real benefit. The result is a paradox: companies pay premium prices to route their most ordinary traffic the least efficient way.

This does not mean MPLS should disappear. The honest question has simply evolved from “whether MPLS should or should not be used, to “which traffic earns a place on that MPLS motorway, and which does not?”

Migration is not all-or-nothing

In practice, most enterprises converge on a similar balance: roughly 20% of traffic stays on MPLS. For example, ERP systems, voice, regulated applications, etc remains on MPLS. Whilst the other 80% moves to broadband and 5G under SD-WAN control. Getting there is a journey, not a switch. A sound migration runs in phases. Firstly an audit of existing circuits and traffic, then a pilot in a handful of low-risk sites over 30 to 60 days, followed by a wave-by-wave rollout aligned with the expiry dates of existing MPLS contracts, and finally the gradual decommissioning of circuits no longer carrying critical load.

Some sites keep MPLS for good reasons. Regulatory compliance in banking or healthcare, locations without quality broadband, or applications acutely sensitive to latency. And there is one mistake worth underlining. Never try to tear down MPLS the same day SD-WAN goes live. A period of parallel coexistence is not optional, but it is the safety net that makes the whole transition survivable.

SD-WAN as traffic control for one network & many roads

What makes the data multi-transport model work, is the intelligence sitting on top of it. SD-WAN classifies traffic and decides, in real time, which path each flow should take. MPLS for the critical and the regulated data traffic, broadband for everyday SaaS, and 5G as added capacity or as the primary link for sites where fiber never arrived.

This is application-aware routing. Rather than relying on static rules, SD-WAN continuously measures latency, jitter and packet loss across every available path and reroutes automatically when a link degrades. The outcome is a hybrid network optimized for both cost and performance, where each transport type has a clearly defined role instead of competing for the same job.

SD-Branch platform, beyond the WAN

Yet SD-WAN governs only the transport, and a branch is far more than its uplinks. This is where SD-Branch takes the next step, bringing all the office network equipment together. Routers, LAN switches and Wi-Fi access points, alongside the SD-WAN itself, under a single, centrally managed system. Rather than configuring each device on its own, an administrator provisions, updates and monitors the entire branch from one console, and pushes the same policies consistently across every site. The result is operational simplicity at scale: dozens or hundreds of offices behaving as one coherent, remotely managed fabric instead of a patchwork of devices from different vendors.

Lower CapEx and OpEx

The financial case is rarely subtle. Reclassifying traffic away from MPLS onto broadband and 5G can cut bandwidth costs by 60 to 70%. Just as important are the hidden costs that quietly disappear: cross-vendor licensing and on-site technical visits. Zero-Touch Provisioning (ZTP) sharpens the savings further. Because equipment configures itself once plugged in, so non-technical staff can install it. This means that no engineer has to fly out to a remote site simply to configure a router online.

From contracted SLAs to AI-driven visibility

There is a catch that deserves honesty. With pure MPLS, quality was the operator’s problem; the SLA was a contract that let the enterprise “pay not to think.” In a multi-transport data world, that responsibility shifts back to the organization itself. Cellular and 5G links are inherently variable. They are subject to cell congestion, radio conditions and time-of-day fluctuations. And additionally broadband offers no guarantees either. Actively monitoring every link therefore stops being a nicety and becomes essential.

This is where AI changes the equation. Predictive analysis can spot recurring degradation patterns. A 5G link that congests at the same hours each day and shifts critical traffic away before users ever feel it. Trend analysis can flag a broadband line slowly deteriorating, or reveal that 5G coverage at a given site is now strong enough to promote it from backup to primary. Monitoring, in other words, turns the multi-transport model from a cost-saving compromise into a genuine advantage. Spending less time than you did on MPLS, and gaining continuous, intelligent control over a network that used to be a black box.

Choosing the right road for every flow

The branch office is being quietly redesigned. MPLS is no longer the automatic choice. It is one road among several, reserved for the traffic that truly needs it. While broadband and 5G carry the rest under the watchful orchestration of SD-WAN. Done in phases, with eyes open, the payoff is a network that costs less and understands itself better.

At Teldat, this is the reality our SD-WAN and SD-Branch solutions are built for: multi-transport branch architectures where MPLS, broadband and 5G coexist, and where centrally managed, application-aware orchestration turns mixed connectivity into dependable, observable infrastructure.