{"id":1108,"date":"2026-11-15T19:40:00","date_gmt":"2026-11-15T19:40:00","guid":{"rendered":"https:\/\/planespart.com\/blog\/?p=1108"},"modified":"2026-11-17T18:40:00","modified_gmt":"2026-11-17T18:40:00","slug":"sre-parts-counter-team-story","status":"publish","type":"post","link":"https:\/\/planespart.com\/blog\/sre-parts-counter-team-story\/","title":{"rendered":"How a Small Parts Supplier Tamed Its Order Page With SRE"},"content":{"rendered":"<p>The following story is invented, a composite of the kind of week small aviation parts suppliers sometimes have. A supplier we will call Northfield Aero Parts ran a modest online catalogue and order desk, five people on the counter and two part-time developers. Every few weeks the order page slowed to a crawl or threw errors, and every time it did, the owner phoned a developer in the evening. This is how that team stopped living on adrenaline by borrowing the habits of SRE, short for site reliability engineering.<\/p>\n<h2>The week the order page fell over<\/h2>\n<p>It started on a Monday. A developer pushed a tidy change to how stock levels were displayed. By lunchtime customers were seeing &#8220;out of stock&#8221; on items that were sitting on the shelf, and the counter staff were fielding calls. The fix took twenty minutes. The cleanup, apologies and manual re-keying of orders took until Wednesday.<\/p>\n<p>The owner&#8217;s reaction was the common one: &#8220;Slow down, test everything twice, no more changes on Mondays.&#8221; That felt sensible, and it was also the wrong lever. According to <a href=\"https:\/\/www.xopsschool.com\/sre\/\">the SRE entry in the Ops Atlas<\/a>, a large share of outages, around seventy percent in the figure it cites, come from changes made to live systems. Stopping change entirely trades one problem for another, because nothing improves and the backlog of fixes grows. What the discipline offers instead is a way to decide how much change the system can absorb right now.<\/p>\n<h2>Step one: agree what &#8220;working&#8221; means<\/h2>\n<p>The team sat down with a whiteboard. The question was blunt: when is the website doing its job? They settled on two statements. First, a customer who submits an order should see a confirmation. Second, a search for a part number should return results quickly.<\/p>\n<p>They counted attempts rather than guessing. Over a quiet fortnight, the order form was used a few thousand times and failed on a small handful. They did not need fancy tools; the web host&#8217;s logs and a spreadsheet were enough. The Atlas entry describes the same recipe at industrial scale, dividing good events by valid ones, and cautions against treating raw server readings as a stand-in for customer happiness.<\/p>\n<h2>Step two: pick a target short of perfect<\/h2>\n<p>The owner wanted &#8220;never down.&#8221; The developers explained that every extra decimal place of reliability costs disproportionately more. Per the Atlas entry, each step up the ladder may be priced at about a hundred times the step beneath it. A parts shop with seven staff cannot buy that.<\/p>\n<p>They chose 99.5 percent of order attempts succeeding, measured over a rolling four weeks. The leftover half a percent became an allowance. On roughly three thousand attempts a month that allowed about fifteen failures. Seeing the number on paper changed the conversation immediately. Monday&#8217;s stock display mistake had burned most of a month&#8217;s allowance in a single morning.<\/p>\n<h2>Step three: a rule that everybody signed<\/h2>\n<p>A target without consequences is a poster on the wall, and the source warns against exactly that, listing objectives with no signed policy as a known anti-pattern. The team wrote five short lines and the owner initialled them:<\/p>\n<ol>\n<li>If the allowance is healthy, changes go out as usual on weekdays.<\/li>\n<li>If more than half the allowance is spent, every change needs a second person to read it first.<\/li>\n<li>If the allowance is gone, only urgent fixes ship until the four-week window recovers.<\/li>\n<li>Any single event that eats a large slice triggers a short written review.<\/li>\n<li>The review ends with named actions and dates.<\/li>\n<\/ol>\n<p>Compare that with the owner&#8217;s earlier &#8220;no changes on Mondays&#8221; edict. The new rules were less rigid and far more honest: they moved with the real state of the site.<\/p>\n<h2>Step four: small releases, like checking a part before fitting it<\/h2>\n<p>Mechanics do not install a batch of fasteners without checking each one. The developers adopted a similar attitude to releases. Instead of pushing a change to everybody, they switched it on for a small fraction of visitors first. The Atlas entry offers neat arithmetic for why this helps: if five percent of traffic sees a fault that fails twenty percent of the time, overall errors rise by only one percent while you notice and roll back. A bad change that would have hit every customer hit a few instead.<\/p>\n<h2>Step five: fewer midnight calls<\/h2>\n<p>The last piece was the phone. Northfield stopped alerting on every wobble and defined two levels. A fast, sharp rise in failures would call the on-duty developer. A slow leak would create a ticket for the next morning. The entry&#8217;s discussion of burn rate gives precise thresholds for large operations, but a small shop can start simpler: &#8220;page if more than a tenth of orders fail for an hour; ticket if one in fifty fail for a day.&#8221;<\/p>\n<p>They also wrote a one-page playbook for the commonest failure, restarting the stock sync and checking the supplier feed. The source claims that playbooks bring roughly a threefold improvement in recovery time, and while a two-person shop cannot promise that figure, the instinct is sound. Whoever is on call at midnight should not have to remember anything.<\/p>\n<h2>The objections they heard along the way<\/h2>\n<p>Not everyone was thrilled. One counter clerk asked whether all this measuring was just extra paperwork. The answer was that it replaced a different kind of paperwork: the long apologetic emails and the manual re-entry of orders after every crash. Another worry was that a published target would be used against the developers. The owner&#8217;s reply set the tone: a missed target is a prompt to investigate, not a reason to hunt for someone to blame.<\/p>\n<h2>What changed after three months<\/h2>\n<p>Nothing dramatic. The site was not flawless. But the owner stopped phoning developers in the evening, because the alerts only fired when they mattered. Releases happened weekly instead of monthly, because the team knew how much room they had. Monday&#8217;s mistake remained the biggest event on the log, and the review of it was the moment everyone realised the measurements were useful.<\/p>\n<h2>Lessons for your own parts business<\/h2>\n<ul>\n<li>Define success from the customer&#8217;s side, not the server&#8217;s.<\/li>\n<li>Choose a target below perfect and compute the allowance in plain numbers.<\/li>\n<li>Write the rule, sign it, and follow it when the allowance runs low.<\/li>\n<li>Release in small pieces and watch before widening.<\/li>\n<li>Alert on drain rate, and write a playbook for the repeat offenders.<\/li>\n<\/ul>\n<p>If you are tightening up other parts of the buying chain too, our guide to <a href=\"https:\/\/planespart.com\/blog\/control-system-parts-buying-mistakes-to-avoid\/\">avoiding common slips when ordering control system components<\/a> covers the physical side of the same discipline, and the primer on <a href=\"https:\/\/planespart.com\/blog\/complete-guide-to-aircraft-fasteners-and-hardware\/\">how fasteners and hardware are picked for airframes<\/a> is a good reminder that small, well-checked pieces keep large machines safe. Reliable software and reliable hardware share a common root: decide what good looks like, measure it honestly, and act on what you find.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>An invented story of a five-person aircraft parts counter that stopped firefighting its website by setting a realistic target and a signed change rule.<\/p>\n","protected":false},"author":6,"featured_media":1109,"comment_status":"","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[617],"tags":[],"class_list":["post-1108","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-aviation-maintenance"],"_links":{"self":[{"href":"https:\/\/planespart.com\/blog\/wp-json\/wp\/v2\/posts\/1108","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/planespart.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/planespart.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/planespart.com\/blog\/wp-json\/wp\/v2\/users\/6"}],"replies":[{"embeddable":true,"href":"https:\/\/planespart.com\/blog\/wp-json\/wp\/v2\/comments?post=1108"}],"version-history":[{"count":1,"href":"https:\/\/planespart.com\/blog\/wp-json\/wp\/v2\/posts\/1108\/revisions"}],"predecessor-version":[{"id":1211,"href":"https:\/\/planespart.com\/blog\/wp-json\/wp\/v2\/posts\/1108\/revisions\/1211"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/planespart.com\/blog\/wp-json\/wp\/v2\/media\/1109"}],"wp:attachment":[{"href":"https:\/\/planespart.com\/blog\/wp-json\/wp\/v2\/media?parent=1108"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/planespart.com\/blog\/wp-json\/wp\/v2\/categories?post=1108"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/planespart.com\/blog\/wp-json\/wp\/v2\/tags?post=1108"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}