May 2026 · 12 min read

SD-WAN vs MPLS — a 2026 decision guide for multi-site mid-market

If you have 5–25 sites and an MPLS contract coming up for renewal, this is the most consequential network decision you'll make in the next decade. Here's the framework we use — stripped of the carrier marketing pitch and the SD-WAN vendor marketing pitch.

The setup most mid-market companies are in right now

A typical mid-market IT footprint we audit looks like this: 8–15 locations, a primary carrier providing MPLS or Ethernet private line between them, a secondary connection (sometimes business broadband, sometimes a backup DIA circuit) at each site, and total WAN spend somewhere between $8k and $25k/month.

The MPLS contract was probably signed somewhere between 2017 and 2021. At the time, MPLS was still the only reasonable answer for a multi-site company that cared about application performance. SD-WAN was newer and the operational story wasn't fully baked.

In 2026 that's no longer true. SD-WAN is mature. The math is usually significantly better. But "usually" is doing a lot of work in that sentence, so let's break it down.

The honest cost math

MPLS pricing varies wildly by region, carrier, and site type, but a reasonable benchmark for a 50Mbps MPLS port at a typical mid-market site is $750–1,400/month. At 10 sites that's $7,500–14,000/month before any added services or DIA fallback.

SD-WAN replacement using business broadband or fiber DIA at each site typically runs $200–600/month per site for the underlay connectivity, plus an SD-WAN overlay license ($25–75/site/month depending on platform) and (optionally) a managed service from the carrier or VAR.

Net at 10 sites: typically 40–60% cost reduction on like-for-like capacity, often with 5–10× the bandwidth at each site because you're moving from 50Mbps MPLS to 500Mbps–1Gbps broadband.

Caveat: this math gets worse for rural or hard-to-serve sites where commodity broadband isn't available. If three of your fifteen locations are at the end of an old copper run with no fiber and no cable, your "SD-WAN cost savings" can compress to 10–15% because you're still paying carrier prices for those last-mile circuits.

When MPLS is still the right answer

We don't recommend SD-WAN universally. There are four scenarios where the MPLS renewal is still the right call:

  1. Heavy real-time application use across sites with no cloud equivalent. Older on-prem voice systems, certain ERP architectures, latency-sensitive manufacturing control systems. If your traffic profile is dominated by one of these and you don't have a near-term plan to move it to cloud, MPLS QoS still wins.
  2. Compliance or regulatory frameworks that require private connectivity. Some PCI scope reduction strategies, certain healthcare networks, and government contracts still meaningfully prefer (or require) private WAN architecture.
  3. Geography that doesn't have broadband alternatives. If five of your ten sites are rural and only have DSL or fixed wireless options, the SD-WAN performance story breaks down. MPLS sites in those locations may be your only path to reliable application performance.
  4. Active project conflict. SD-WAN migration is a 4–9 month project when done well. If your team is mid-ERP-replacement or mid-merger-integration, layering a network transformation on top is asking for failure.

The hybrid option people forget

Most multi-site mid-market companies don't need a binary choice. The right answer is often SD-WAN with selective MPLS hold-over for specific sites or applications. You keep MPLS at the 2–3 sites where it's irreplaceable (data center, regional HQ, rural plant) and migrate everywhere else to SD-WAN over commodity transport.

Carriers will not lead with this option. They make more on the MPLS-only renewal and more on the SD-WAN-everywhere transformation than they do on the hybrid. The buyer-side framing — what do we actually need where — is the framing that produces the lowest-cost, lowest-risk outcome.

SASE — why people are confusing this with SD-WAN

SASE (Secure Access Service Edge) gets pitched alongside SD-WAN constantly, and it's creating real confusion. Quick clarification:

SASE is the right direction for many mid-market companies, but it's also more expensive and more disruptive than just adopting SD-WAN. If your security architecture is already working — appliances at sites, working SSL inspection, working DNS filtering — there's no forced timeline to consolidate to SASE. Many of our clients do SD-WAN first, then layer in cloud security 12–24 months later. That sequencing controls risk.

The questions to ask any SD-WAN proposal

When you receive an SD-WAN proposal from any carrier (Spectrum, AT&T, Lumen, Verizon) or any pure-play vendor (Versa, VMware VeloCloud/Broadcom, Cisco Meraki, Fortinet, Aryaka, Cato Networks), ask these eight questions in writing:

  1. What's the total monthly cost at year 1 vs year 2 vs year 3 (some vendors discount year 1 then ramp)?
  2. Are SD-WAN appliances included or is hardware separate capex?
  3. What's the SLA for the overlay vs the underlay (these are often different)?
  4. Who do we call if a circuit goes down — the underlay carrier or the SD-WAN provider?
  5. Can we bring our own transport at any site (bring-your-own-circuit) or are we locked into your underlay options?
  6. What's the early termination clause and partial termination clause at individual sites?
  7. What's the architecture for cloud on-ramp (direct AWS/Azure peering) and does it cost extra?
  8. Is the SD-WAN platform on the vendor's roadmap for 5+ years, or has it been acquired/sunset?

The decision framework

At a five-question level, here's how we actually frame the choice:

Three or four "yes" answers → start the SD-WAN evaluation now. Two or fewer → renew the MPLS contract for a 24-month term (not 36, you want flexibility) and revisit.

Want us to apply this framework to your invoice or renewal?

Send us the document. We'll mark up the opportunity. Free invoice review — 24-business-hour turnaround, no obligation.

Related