In engineering, we naturally reach for engineering terms when we talk about reliability. Uptime percentages. Recovery times. Deployment frequency. All useful numbers, but on their own they miss the point of why any of it matters. Reliability only matters because the business depends on it. Every outage, every slow deployment, every unplanned incident has a cost attached, and that cost is paid by the whole business, not just by whoever is on call that night.
\nI learned this properly while leading Azure cloud and platform strategy for a portfolio of machine learning and data products inside a regulated insurance environment. The platform supported multiple product teams, all shipping into a space where downtime or data exposure carries real regulatory and financial consequences. Reliability wasn’t a nice to have. It was the product.
\nWe designed and owned an enterprise grade Azure platform covering API Management, ML Workspace, Container Apps, and a fairly complex networking layer, built specifically for zero downtime deployments and full disaster recovery compliance. None of that architecture existed because it was technically interesting. It existed because a failure in any of those layers would have stopped a regulated business process, and we would have had to account for that.
\nThat framing changed how I made decisions. Every architectural choice got tested against a simple question: what happens to the business if this breaks, and how fast can we recover. That question is more useful than any SLA target, because it forces you to think about actual consequences rather than abstract metrics.
\nThere is a tendency to treat cost optimisation and reliability as competing priorities, as if a cheaper platform must be a riskier one. In practice, the opposite was true for us. Through architectural rationalisation, rightsizing, and better governance, we reduced Azure operational costs by 40 to 45 percent, and the platform came out more reliable, not less. Removing waste forced us to understand our own architecture properly. You cannot rightsize what you do not fully understand, and understanding your platform is the first step toward making it dependable.
\nNone of the businesses we have supported ever asked for a well architected platform for its own sake. What we were really being asked for was confidence that the systems we all depend on will be there when needed, at a cost that makes sense. Reliability and cost efficiency turned out to be the same conversation, just told from two different angles.
\nDay to day, the teams building on our platform were not thinking about uptime numbers. What they were really relying on was trust, trust that the platform would be there so they could focus on their own work. That trust is the actual product we deliver as a platform team. We earn it through the boring, unglamorous decisions we make together: proper DR planning, sensible governance, and architecture that fails safely rather than catastrophically.
\nWhen platform engineering gets described purely as a technical discipline, I think that undersells it. The technical work is the mechanism. The outcome is confidence, and confidence is a business asset with a measurable value, whether or not anyone puts a number on it.
\nThat is the lens we try to bring to platform work. Not just “is this technically sound,” though it has to be, but “does this make the business more confident in what it depends on.” Everything else follows from that.
","htmlAst":{"type":"root","children":[{"type":"element","tagName":"p","properties":{},"children":[{"type":"text","value":"In engineering, we naturally reach for engineering terms when we talk about reliability. Uptime percentages. Recovery times. Deployment frequency. All useful numbers, but on their own they miss the point of why any of it matters. Reliability only matters because the business depends on it. Every outage, every slow deployment, every unplanned incident has a cost attached, and that cost is paid by the whole business, not just by whoever is on call that night."}]},{"type":"text","value":"\n"},{"type":"element","tagName":"p","properties":{},"children":[{"type":"text","value":"I learned this properly while leading Azure cloud and platform strategy for a portfolio of machine learning and data products inside a regulated insurance environment. The platform supported multiple product teams, all shipping into a space where downtime or data exposure carries real regulatory and financial consequences. Reliability wasn’t a nice to have. It was the product."}]},{"type":"text","value":"\n"},{"type":"element","tagName":"h2","properties":{},"children":[{"type":"text","value":"The architecture had to justify itself in business terms"}]},{"type":"text","value":"\n"},{"type":"element","tagName":"p","properties":{},"children":[{"type":"text","value":"We designed and owned an enterprise grade Azure platform covering API Management, ML Workspace, Container Apps, and a fairly complex networking layer, built specifically for zero downtime deployments and full disaster recovery compliance. None of that architecture existed because it was technically interesting. It existed because a failure in any of those layers would have stopped a regulated business process, and we would have had to account for that."}]},{"type":"text","value":"\n"},{"type":"element","tagName":"p","properties":{},"children":[{"type":"text","value":"That framing changed how I made decisions. Every architectural choice got tested against a simple question: what happens to the business if this breaks, and how fast can we recover. That question is more useful than any SLA target, because it forces you to think about actual consequences rather than abstract metrics."}]},{"type":"text","value":"\n"},{"type":"element","tagName":"h2","properties":{},"children":[{"type":"text","value":"Cost discipline is part of reliability, not separate from it"}]},{"type":"text","value":"\n"},{"type":"element","tagName":"p","properties":{},"children":[{"type":"text","value":"There is a tendency to treat cost optimisation and reliability as competing priorities, as if a cheaper platform must be a riskier one. In practice, the opposite was true for us. Through architectural rationalisation, rightsizing, and better governance, we reduced Azure operational costs by 40 to 45 percent, and the platform came out more reliable, not less. Removing waste forced us to understand our own architecture properly. You cannot rightsize what you do not fully understand, and understanding your platform is the first step toward making it dependable."}]},{"type":"text","value":"\n"},{"type":"element","tagName":"p","properties":{},"children":[{"type":"text","value":"None of the businesses we have supported ever asked for a well architected platform for its own sake. What we were really being asked for was confidence that the systems we all depend on will be there when needed, at a cost that makes sense. Reliability and cost efficiency turned out to be the same conversation, just told from two different angles."}]},{"type":"text","value":"\n"},{"type":"element","tagName":"h2","properties":{},"children":[{"type":"text","value":"Reliability is a trust product"}]},{"type":"text","value":"\n"},{"type":"element","tagName":"p","properties":{},"children":[{"type":"text","value":"Day to day, the teams building on our platform were not thinking about uptime numbers. What they were really relying on was trust, trust that the platform would be there so they could focus on their own work. That trust is the actual product we deliver as a platform team. We earn it through the boring, unglamorous decisions we make together: proper DR planning, sensible governance, and architecture that fails safely rather than catastrophically."}]},{"type":"text","value":"\n"},{"type":"element","tagName":"p","properties":{},"children":[{"type":"text","value":"When platform engineering gets described purely as a technical discipline, I think that undersells it. The technical work is the mechanism. The outcome is confidence, and confidence is a business asset with a measurable value, whether or not anyone puts a number on it."}]},{"type":"text","value":"\n"},{"type":"element","tagName":"p","properties":{},"children":[{"type":"text","value":"That is the lens we try to bring to platform work. Not just “is this technically sound,” though it has to be, but “does this make the business more confident in what it depends on.” Everything else follows from that."}]}],"data":{"quirksMode":false}},"excerpt":"In engineering, we naturally reach for engineering terms when we talk about reliability. Uptime percentages. Recovery times. Deployment…","timeToRead":2,"frontmatter":{"title":"Why Platform Reliability Is a Business Metric, Not Just an Engineering One","userDate":"13 July 2026","date":"2026-07-13T14:00:00.000Z","tags":["Platform","All"],"image":{"childImageSharp":{"fluid":{"base64":"data:image/png;base64,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","aspectRatio":1.9057377049180328,"src":"/static/0e2e57dc8dee511fcbc6c66cedef584b/0ef64/banner.png","srcSet":"/static/0e2e57dc8dee511fcbc6c66cedef584b/c5c6c/banner.png 930w,\n/static/0e2e57dc8dee511fcbc6c66cedef584b/fe828/banner.png 1860w,\n/static/0e2e57dc8dee511fcbc6c66cedef584b/0ef64/banner.png 2400w","sizes":"(max-width: 2400px) 100vw, 2400px"}}},"author":{"id":"Dushyant","bio":"Check about page for details.","avatar":{"children":[{"__typename":"ImageSharp","fixed":{"base64":"data:image/jpeg;base64,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","width":400,"height":388,"src":"/static/ca50b870313d79772bec7be9f9d9786f/64b17/dp-profile.jpg","srcSet":"/static/ca50b870313d79772bec7be9f9d9786f/64b17/dp-profile.jpg 1x,\n/static/ca50b870313d79772bec7be9f9d9786f/39f27/dp-profile.jpg 1.5x,\n/static/ca50b870313d79772bec7be9f9d9786f/a1eb1/dp-profile.jpg 2x"}}]}}}},"relatedPosts":{"totalCount":2,"edges":[{"node":{"id":"d7750fb5-fe7c-5de6-b0ea-b691249e4e7a","timeToRead":2,"excerpt":"For a long time, conversations about delivery performance in the teams I worked with were mostly anecdotal. Someone would say releases felt…","frontmatter":{"title":"DORA Metrics in Practice - Giving Leadership Real Visibility Into Delivery Performance"},"fields":{"slug":"/dora-in-practice/"}}},{"node":{"id":"dc071f4a-6e03-560c-99ce-75d8ffbd1e3f","timeToRead":2,"excerpt":"In engineering, we naturally reach for engineering terms when we talk about reliability. Uptime percentages. Recovery times. Deployment…","frontmatter":{"title":"Why Platform Reliability Is a Business Metric, Not Just an Engineering One"},"fields":{"slug":"/platform-reliability-as-a-business-metric/"}}}]}},"pageContext":{"slug":"/platform-reliability-as-a-business-metric/","prev":{"excerpt":"Introduction As software delivery becomes faster and more complex, teams often need to decide how their Azure DevOps pipelines should be…","timeToRead":7,"frontmatter":{"title":"Single Multi-Stage Pipelines vs Multiple Pipelines in Azure DevOps","tags":["Azure","DevOps","All"],"date":"2026-07-03T00:00:00.000Z","draft":false,"image":{"childImageSharp":{"fluid":{"aspectRatio":2.3785166240409206,"base64":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAABQAAAAICAIAAAB2/0i6AAAACXBIWXMAAAsTAAALEwEAmpwYAAAA4klEQVQY02P4jwr+/P335++//8QBBkyhf///7zj9dem+z++//AVx/xGtec3+FzVzn8R1vYzpfHHp3o//////JUYzxIbK6bd0U+64V72oXfD6249/EIcQZfPfv/9fvP/bueKNoNeJU9c+QILg/7//YAJEQhg4Nb/5+O/dh18rdz19//EXfg+jaIb47fD9r5uvf7n8+teBe18vv/i54uKn7Te/bLvxZcm5jwfuft19+8ut1z/hhqLY/O/f/ztvf+2+/WXXrS/HHn67+frng/e/tt34svXGl0P3vp57+h2k+Q1CMwBd08VzZ+r1pwAAAABJRU5ErkJggg==","sizes":"(max-width: 1000px) 100vw, 1000px","src":"/static/9aa7ca38858eb3ecea7574cbb0696d75/31987/banner.png","srcSet":"/static/9aa7ca38858eb3ecea7574cbb0696d75/c5c6c/banner.png 930w,\n/static/9aa7ca38858eb3ecea7574cbb0696d75/31987/banner.png 1000w"}}},"author":{"id":"Dushyant","bio":"Check about page for details.","avatar":{"children":[{"fixed":{"src":"/static/ca50b870313d79772bec7be9f9d9786f/64b17/dp-profile.jpg"}}]}}},"fields":{"layout":"post","slug":"/single-multi-stage-pipelines-vs-multiple-pipelines-in-azure-devops/"}},"next":{"excerpt":"For a long time, conversations about delivery performance in the teams I worked with were mostly anecdotal. Someone would say releases felt…","timeToRead":2,"frontmatter":{"title":"DORA Metrics in Practice - Giving Leadership Real Visibility Into Delivery Performance","tags":["Platform","All"],"date":"2026-07-17T14:00:00.000Z","draft":false,"image":{"childImageSharp":{"fluid":{"aspectRatio":1.7816091954022988,"base64":"data:image/png;base64,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","sizes":"(max-width: 976px) 100vw, 976px","src":"/static/95d9cf702ac83fdb9cb836fcfb7008a5/3bd0d/banner.png","srcSet":"/static/95d9cf702ac83fdb9cb836fcfb7008a5/c5c6c/banner.png 930w,\n/static/95d9cf702ac83fdb9cb836fcfb7008a5/3bd0d/banner.png 976w"}}},"author":{"id":"Dushyant","bio":"Check about page for details.","avatar":{"children":[{"fixed":{"src":"/static/ca50b870313d79772bec7be9f9d9786f/64b17/dp-profile.jpg"}}]}}},"fields":{"layout":"post","slug":"/dora-in-practice/"}},"primaryTag":"Platform"}},"staticQueryHashes":["3100086437","794779027","990608418"]}