We spend a lot of time arguing that you should outsource operational reliability — the pager, the triage, the tuning, the 3 AM diagnosis. That’s true, and we’ll keep saying it. But there’s an honest counterpoint that most managed-service marketing conveniently omits: some reliability work cannot be handed off, and pretending otherwise sets everyone up to fail.
This isn’t a hedge. It’s the thing that determines whether outsourcing reliability actually works. The partnerships that succeed are the ones where both sides are clear about the line — what the provider owns, and what stays yours no matter who you hire. Get the line wrong and you either keep work you shouldn’t or, worse, hand off decisions nobody outside your business can make.
The Line Is Between Operating and Deciding
The clean way to think about it: reliability has an operational layer and a business-judgment layer. The operational layer is highly outsourceable. The business-judgment layer is not — not because a provider isn’t capable, but because the inputs live inside your company.
The operational layer is executing the reliability function well: carrying the pager, keeping the signal clean, diagnosing failures fast, remediating, running the postmortem loop. This is craft and continuous discipline, and it’s exactly what a good managed partner should own — most teams execute it worse than a dedicated provider would, because for them it’s a distraction and for the provider it’s the whole job.
The business-judgment layer is deciding what reliability means for your business. What counts as “down.” Which failures are catastrophic and which are tolerable. What you’re willing to trade for speed. These aren’t operational questions with technically correct answers. They’re business questions whose answers only exist inside your company.
Outsource the first layer aggressively. Own the second layer deliberately. Confusing them is where reliability partnerships go wrong in both directions.
What Stays Yours No Matter Who You Hire
Four responsibilities don’t transfer, and being honest about them is what makes the transferable parts work.
Defining what “reliable” means for your business. A provider can hold any SLO you set. They cannot tell you it should be 99.5% or 99.99%, because that depends on what your customers tolerate, what your contracts promise, and what you’re willing to pay. Someone can execute an SLO for you. Only you can decide what it should be — that number encodes a business risk appetite, and risk appetite is owned, not delegated.
Knowing which failures actually hurt. Your provider can watch every endpoint. They can’t know, without you, that a five-minute outage of your billing webhook is a compliance incident while an hour of degraded search is a shrug. That prioritisation comes from understanding your business model, your contracts, and your customers — knowledge that lives with you. You supply the map of what matters; the provider navigates it.
The tradeoff between reliability and speed. How much velocity you’re willing to spend on reliability, and vice versa, is a strategic call tied to your stage and your competitive position. An early-stage company racing to product-market fit makes a different bet than a company selling to regulated enterprises. A provider executes within the tradeoff you set. Setting it is yours.
Owning the customer relationship when it breaks. When an outage hurts a customer, the conversation about trust, credits, and what it means for the relationship is yours. A provider resolves the incident and hands you a clear account of what happened. They don’t — and shouldn’t — own your customer’s trust. That’s the relationship you’re in business to protect.
What You Can and Should Hand Off
Being clear about the four above is what earns you the right to aggressively offload everything else — and there’s a lot of everything else.
Carrying the pager. Continuously tuning alerts so the signal stays clean. Diagnosing at 3 AM with cross-system pattern recognition your small team can’t accumulate. Remediating known failure modes from runbooks. Running the mechanical parts of the postmortem loop and feeding fixes back into monitoring. Keeping observability sharp and its cost honest. This is the bulk of the hours reliability consumes, and almost none of it benefits from being done in-house by people who’d rather be building your product.
The mistake small teams make isn’t outsourcing too much. It’s outsourcing too little because they’ve conflated the operational layer with the judgment layer — they keep the pager because they think keeping the pager means keeping control of what matters. It doesn’t. You keep control by owning the judgment layer, which frees you to hand off the operational layer entirely.
Why This Distinction Makes Outsourcing Work Better
Naming the non-outsourceable work isn’t an argument against managed reliability — it’s the thing that makes managed reliability succeed. The partnerships that fail are the ones with a fuzzy line: the provider is vaguely expected to “own reliability,” which quietly includes business judgments they can’t make, so when a failure lands in the judgment layer, nobody owned it and everyone’s surprised.
The partnerships that work have a crisp division. You own what “reliable” means, which failures matter, the speed tradeoff, and the customer relationship. The provider owns executing reliability flawlessly against those definitions. Each side owns what it’s actually positioned to own. That clarity is the responsibility boundary drawn in the right place — and drawing it right is the whole game.
So the honest pitch isn’t “outsource everything and stop thinking about reliability.” It’s “own the handful of business judgments only you can make, and hand off the operational mountain to people who do it as their whole job.” That’s not less control. It’s control located where it belongs.
Own the Judgment, Delegate the Execution
Reliability isn’t fully outsourceable, and any provider who tells you it is either misunderstands the work or is selling you something that will fail at the boundary. The right model is a clean split: you own the business judgment, a partner owns the execution, and the line between them is explicit from day one.
Vigil by IOanyT owns the operational reliability mountain — the pager, the tuning, the 3 AM diagnosis, the remediation, the improvement loop — against the definitions you own, with the responsibility boundary drawn explicitly so the judgment that has to stay yours, stays yours.
Your team owns what reliability means. We own making it real.
