SD-WAN vs MPLS: A Practical Decision Framework

Most SD-WAN conversations start with the technology. The better ones start with the business: what the network has to do, where it hurts today and what a change is worth.

After several network transformation programmes, these are the questions I now ask before any architecture or vendor discussion.

1. What problem are we actually solving?

"Move to SD-WAN" is a solution, not a problem. The common real drivers are:

• Cost — MPLS circuits are expensive, especially for small sites
• Agility — new sites take months to connect
• Cloud performance — traffic to SaaS and public cloud is backhauled through a data centre
• Resilience — single-circuit sites go down with every carrier fault
• Lifecycle — routers are end-of-life and need replacing anyway

Each driver leads to a different design and a different business case. If the honest answer is "lifecycle", the decision is really about what you replace the routers with — and SD-WAN may simply be the sensible default.

2. What does the traffic look like?

Pull a few weeks of flow data before deciding anything. You want to know:

3. Keep, replace or blend the underlay?

SD-WAN is an overlay. The underlay can be MPLS, broadband, dedicated internet, 4G/5G — or a mix. Three common patterns:

1. Hybrid — keep MPLS at critical sites, add broadband everywhere
2. Internet-first — dual internet circuits, MPLS retired over time
3. Tiered — site design depends on criticality (e.g. Gold / Silver / Bronze)

Tiered designs are usually the most realistic for large estates. They also make the business case easier to defend, because savings come from the long tail of smaller sites.

4. Who owns security?

Local internet breakout changes your security model. Decide early whether security is enforced on-box, in a cloud security service, or both — and who operates it. This is where many SD-WAN projects get stuck in approvals.

5. How will you migrate?

The migration plan matters as much as the target design:

• Pilot at a few representative sites (not just the easy ones)
• Define success criteria before the pilot starts
• Group the rest into waves by region, site type or circuit renewal date
• Agree rollback criteria for every cutover

A simple scoring approach

When I need to compare options for leadership, I use a weighted score rather than a feature list:

Criterion            Weight   Option A   Option B
---------------------------------------------------
Total cost (5 yrs)     30%       4          3
Resilience             25%       3          5
Cloud performance      20%       5          4
Operational effort     15%       3          4
Migration risk         10%       4          3
---------------------------------------------------
Weighted score                  3.85       3.90

The numbers matter less than the conversation they create. Agreeing the weights is often where the real decision gets made.

The takeaway

SD-WAN is a good answer to many enterprise WAN problems — but only once you're clear which problem you're solving. Start with drivers and data, design the underlay by site tier, settle security ownership early and invest in the migration plan.

Have a WAN decision in front of you? I'm happy to talk it through — get in touch.