Compare business internet & connectivity

Separate the access technology from the service

The access technology is the physical path: fibre, coaxial cable, fixed wireless, a copper pair, or cellular. The service is everything layered on top of it — how much bandwidth is dedicated to you rather than shared, whether the speed is the same in both directions, which service level commitments apply, how many usable addresses you receive, and what routing the provider will perform on your behalf. Two suppliers can quote you fibre at the same headline speed and be selling materially different products. When a proposal reaches you, split it into those two halves before comparing anything, because the access technology decides what is possible and the service decides what you actually get.

• Access: fibre, coax, fixed wireless, copper or cellular.
• Service: dedicated or shared capacity, symmetric or not.
• Commitments: what is promised, measured and credited.
• Addressing and routing: how many IPs, and who announces them.

Contention, committed rate and what an SLA really promises

Consumer and many small-business products are contended: the capacity you share with other subscribers in the same area is larger than the capacity you can use at once, and the product is priced on that assumption. A dedicated service instead states a committed rate you can rely on, which is why it costs more for the same headline figure. An SLA is the written part of that promise, and it is worth reading for three things: which measurements it covers, over what period, and what the remedy is. In practice the remedy is usually a service credit calculated against a small fraction of the monthly charge, which means an SLA is a remedy for inconvenience rather than insurance against loss. What matters most is the latency, jitter and packet loss your applications actually experience, so measure them yourself over a representative week.

• Contended capacity versus a stated committed rate.
• Which metrics the SLA covers, and over what window.
• What the credit is, and how it is calculated.
• Your own latency, jitter and loss measurements, taken over a full week.

When dedicated access is justified and when it is not

Dedicated internet access costs more than a comparable broadband service, and the justification is almost always one of four things: real-time traffic such as voice, cloud applications where latency degrades the work, payment or transaction systems where an outage stops revenue, and multi-site connectivity where one site's failure becomes everyone's problem. The arithmetic that settles the question is downtime tolerance rather than speed. Write down how many minutes of outage a normal trading day can absorb before the business notices, then check the availability figures you are being offered against that number, remembering that an advertised availability percentage is a commitment to a level of service and not a guarantee that no failure will occur.

• Real-time traffic, latency-sensitive applications, revenue systems, multi-site dependence.
• Minutes of tolerable outage per trading day, written down.
• The offered availability figure, expressed as downtime per year.
• What happens to every dependent service when the circuit drops.

Addressing and routing: the parts you must own

The service includes technical details that become painful to change later. Usable addresses matter if you run servers, publish services or maintain site-to-site connections, because addresses handed back at the end of a contract have to be replaced everywhere they appear. Reverse DNS entries for those addresses usually belong to whoever owns the block, not to you. If your business needs to advertise its own address space or manage failover between providers, that requires dynamic routing, which is a different service tier and a different conversation. Ask who owns the addresses, whether reverse DNS is delegated to you, and whether the provider will support routing between your own sites or between two providers.

• How many usable addresses, and who owns them contractually.
• Whether reverse DNS is delegated to you or held by the provider.
• Whether dynamic routing is available, and at what tier.
• What would have to change if you moved to another provider.

Diversity is a design, not a second circuit

Two circuits bought from two suppliers still fail together when they run through the same duct, enter the same building through the same wall, or terminate on the same street cabinet. Real diversity means following the paths rather than the invoices: different last-mile technologies, different physical entries where possible, and different upstream routes. The honest test is a tracing exercise rather than a sales conversation — compare the route each circuit takes to reach the same destination and see how far they share the journey. Where genuinely diverse paths are not available, that is worth knowing before the incident rather than during it, because it changes which of your services can be made resilient and which cannot.

• Different last-mile technology, not just a different supplier.
• Separate physical entry into the building where possible.
• Compare routes to the same destination, and measure where they overlap.
• What is genuinely protectable, and what is not — decided in advance.

What to gather before asking for pricing

Connectivity quotes depend on the address and the requirement, so both need to be specific. Bring the full service address with any unit or suite detail, the number of users and devices expected at the site, the applications that matter and what they need in terms of latency and upload as well as download, the current circuit's measured performance rather than its advertised speed, any addressing requirements, the out-of-contract date on the existing service and its notice period, and the date by which any change has to be complete. Add anything that constrains construction: listed buildings, unusual access, or a landlord who controls entry.

• Full address including unit, and any access constraint.
• Users, devices, and applications with latency and upload needs.
• Current measured performance, not the advertised figure.
• Existing contract end date and the notice period attached to it.

Buy by address: the location-by-location request

The unit of a multi-site purchase is the address, not the account. Each location is qualified, priced, engineered, ordered, installed, tested and billed as its own access circuit even when one master agreement covers them all, because the cost and the feasibility belong to the building rather than to the buyer. Start the request with a location schedule and require the answer to come back as a site-level file, one line per address, rather than a corporate summary. Useful columns are the provider and the underlying access carrier, whether the site is on-net, near-net or off-net, the access technology, the committed and burst rates, the recurring and one-off charges, any construction charge, the estimated installation interval, the service level attached, the demarcation point, who owns the last mile and the building entry, the known diversity limits, and the billing account. A single headline proposal hides exactly the sites that will not work, and those are the sites that consume the project. Send the schedule out with the requirements attached: what is needed at each site, the cutover constraint, and the date by which any change has to be complete.

• One access circuit per address, even under a single master agreement.
• Require a site-level response file, not a corporate proposal.
• Per line: underlying carrier, on-net status, technology, charges, install interval.
• Require the construction charge and the demarcation owner to be stated.

Serviceability is an address question, and the map is a first screen

A serviceability check returns one of four answers: on-net, where the provider already has facilities at the building or demarcation; near-net, where its facilities are close enough that a short extension may be economic; off-net, where access must come from another carrier or a substantial build; or not serviceable for the product, speed and service level requested. Fibre on the street is not the same as a dedicated circuit that can be ordered, because the building entrance, riser, conduit, splice point, poles and rights of way all have to support the service. The FCC's National Broadband Map gives an address-level first screen from its Broadband Serviceable Location Fabric, but it is built around mass-market availability: the map marks the buildings it expects to buy enterprise or commercial-grade service rather than mass-market broadband, so it triages a site, it does not quote one. Keep three quote stages distinct. Preliminary serviceability is database-based and subject to validation. A firm quote rests on network engineering and commercial review. A construction-validated quote rests on a site survey and, where necessary, outside-plant design. Ask what each stage assumes, what it excludes, and when it expires.

• Four answers: on-net, near-net, off-net, not serviceable for the product.
• Fibre on the street does not mean the building can take the circuit.
• The FCC map is a mass-market screen, not a carrier engineering quote.
• Separate preliminary, firm and construction-validated quotes, with expiry dates.

Diversity is a route question, not a second invoice

Two circuits from two providers still fail together when they share the segment that actually breaks. Diversity has to be tested where a single fault can reach both: the access provider, the building entrance, the conduit and riser, the pole or trench route, the splice point, the serving central office or aggregation node, the backhaul path, and the upstream interconnection. A circuit sold by a second provider but delivered over the first provider's local loop is commercially separate and physically shared, which is the most expensive kind of resilience to discover during an incident. The market structure is what makes this concrete. The FCC's Business Data Services framework governs certain dedicated, high-capacity services historically called special access, and these sit outside the mass-market broadband label that attaches to retail offerings rather than to individually negotiated enterprise circuits. The last mile itself is split between incumbent local exchange carriers, which own the legacy local infrastructure, and competitive carriers, which may own fibre, lease loops, resell, or aggregate several underlying networks into one invoice. Ask each bidder, per site, whether it is the incumbent, a facilities-based competitor, a reseller or an aggregator, and which entity owns the last mile and performs the repair. Require a route statement or diagram, and treat unknown as an answer: it identifies the circuit that cannot actually be protected.

• Test diversity at the layer that fails: entrance, conduit, node, backhaul.
• A resold circuit over the first provider's loop is not physically diverse.
• Business Data Services governs dedicated special-access services.
• Ask each bidder whether it owns the last mile or leases it, and who repairs it.

PUT THE OPTIONS SIDE BY SIDE

Your comparison checklist

Common questions