SD-WAN is usually sold on circuit economics: replace expensive private connectivity with cheaper broadband and route intelligently between paths. That case is often real. It is also the least durable part of the value, because circuit pricing moves and the saving gets absorbed.
Where the lasting value sits
For an organization running dozens or hundreds of sites, the persistent benefit is operational. Configuration, policy and security posture become centrally defined and consistently applied rather than being the accumulated result of whoever configured each branch and when.
That changes two things materially. New site deployment goes from a scheduled engineering visit to shipping a device that configures itself on connection. And a policy change — a new application priority, a security rule, a routing preference — applies across the estate in one operation rather than a hundred.
For retail, that is the difference between opening a store in days and opening it in weeks. For banking, it is the difference between being able to evidence consistent branch configuration and hoping it is consistent.
What tends to be underestimated
Application visibility. Once traffic is classified per application per site, most organizations discover their assumptions about what the network carries were substantially wrong — and act on it. That insight is often worth more than the circuit saving.
Underlay dependency. SD-WAN manages multiple paths intelligently; it does not create capacity that is not there. Where local broadband is genuinely poor, the technology mitigates rather than solves. Survey the underlay before designing on the assumption of two viable paths per site.
Security architecture. Breaking out internet traffic locally at each site removes the backhaul cost and removes the central inspection point with it. That is a deliberate trade requiring a deliberate answer — local inspection, cloud-delivered security, or accepting reduced inspection at branches. It should be decided during design, not discovered in an audit.
A sequencing note
Pilot on sites that are representative rather than convenient — including at least one with the worst connectivity in the estate. A pilot on your three best-connected branches will succeed and will teach you very little about the rollout.