The SLA is a liability, not a promise
A service level agreement measures what a vendor will admit, not what your business loses.
Enterprise contracts still negotiate availability percentages as if they were insurance. They are not. An SLA is a capped refund policy dressed up as a guarantee, and the number that should sit in the board deck is the cost of a minute of downtime, not the rebate.
What the contract actually promises
Read an enterprise connectivity SLA carefully and three things become obvious. First, the percentage is almost always about availability, not recoverability. Second, the credit is a fraction of the monthly fee, not a fraction of the damage. Third, exclusions for 'scheduled maintenance', 'force majeure' and 'third party dependencies' leave a lot of daylight. We have read contracts where the exclusion list is longer than the guarantee clause itself, and where 'availability' is measured at the provider's edge, three hops before the traffic ever reaches your building.
The vendors are not being dishonest. They are pricing a product the way any sensible supplier prices risk: cap the downside, keep the premium modest, and let the customer believe the number on the cover page is the number that matters. It works because most procurement teams compare SLA percentages across bids the way they compare warranty years on a car, without asking what the warranty actually pays out if the engine fails on a motorway at rush hour.
An SLA is not a safety net. It is a pre-negotiated apology.
The commodity move is to stop treating the SLA as a risk transfer and start treating it as one input into a resilience design you control. Two diverse paths, automated failover, and a runbook that has actually been rehearsed do more than any contract clause. The SLA becomes a floor, not a strategy: something you keep in the drawer in case you need to claw back a few hundred euros, not something you build the business continuity plan around.
The number to own internally
- Cost per minute of downtime, by system, with revenue and operational parts separated.
- Recovery time objective and recovery point objective that the business actually signed off.
- Mean time to detect and mean time to repair, measured from real incidents.
- A failover test schedule that runs at least quarterly, documented and signed.
- A named owner for each critical path, someone who gets paged, not a distribution list.
Once those numbers exist, the SLA negotiation changes. You are no longer haggling over nines. You are haggling over the rebate for a failure mode you have already decided is unacceptable, and that usually means the right answer is to engineer around the vendor, not to squeeze them harder. We have sat in renewal meetings where the customer asked for 99.999% and the vendor happily quoted a 40% price premium; the customer would have been better off spending that premium on a second carrier.
Most finance teams have a cost of capital number memorised to two decimal places, but ask the same team what an hour of checkout downtime costs on a Friday afternoon in December and you get a shrug. That gap is where the real risk lives. Build the number once, refresh it twice a year, and every infrastructure decision after that gets faster because the trade-off is explicit instead of political.
Commodity discipline in practice
We have never seen a contract credit cover the real cost of an outage. We have seen redundant paths cover it many times. The money spent on diversity is insurance with a claims process that actually works: the traffic reroutes before anyone calls a lawyer. The trick is diversity that is actually diverse, not two circuits from the same wholesale provider running through the same duct under the same road, which we have found more times than we would like to admit during due diligence.
The quarterly rehearsal matters more than the design on paper. A failover path that has not been tested since it was commissioned is a hope, not a plan. We ask clients to schedule the test the same way they schedule payroll: on the calendar, with an owner, with a report at the end that says what broke and who fixed it. Half the time something breaks. That is the point of testing it before the real thing happens on a Tuesday morning with customers watching.
Who actually reads the exclusions
A surprising number of technical leads have never read the exclusions clause on their own connectivity contract, because procurement negotiated it and legal filed it. That is a mistake specific to enterprise ICT: the people who understand the failure modes are rarely in the room when the words that define compensation are chosen. Get an engineer into that meeting, even for thirty minutes, and the exclusions list gets shorter or the price gets clearer.
- Ask for the definition of 'outage': partial degradation rarely counts, only total loss does.
- Ask how availability is measured and from where: edge-of-network numbers flatter every vendor.
- Ask for the credit calculation worked through with last year's actual incident, not a hypothetical.
- Ask what happens on a multi-vendor outage: most contracts go silent the moment blame is shared.
What to change on Tuesday
- 01Replace 'what is our SLA?' with 'what does a minute cost us, and who owns that number?'
- 02Demand diversity, not nines. Two 99.9% paths beat one 99.99% path in the real world.
- 03Rehearse failure quarterly. A runbook that has never run is a fiction.
- 04Treat SLA credits as a small refund, not a risk transfer.
- 05Put an engineer in the room when the exclusions clause is negotiated, not just procurement.
More dispatches.
The catamaran is a computer at sea
Multihulls did not win because they are fast. They won because a stable platform lets sensors, models and crew think clearly.
The cat is now a connected device
Pet tech stopped being a gadget niche. It is a health data business with whiskers.
The incubator is a portfolio, not a factory
Ventures do not share a mould. They share a treasury, a network, and a discipline for killing things early.