<p><em>How the metrics we inherited from manual warehouses became the metrics we least trust, and why the new ones can be just as misleading.</em></p>
<p>The quarterly review opened the way it always did. The automation dashboard was green across the board. System uptime for the quarter: 98.7 percent, comfortably above target. Mean time between failures had improved for the third consecutive quarter. Someone on the project team, understandably, felt good about presenting those numbers.</p>
<p>Then the customer service report landed on the table. Order accuracy for the same period was the worst it had been in six months. Nobody in the room could immediately explain why, because nobody had been looking at both numbers together. The machines had been doing exactly what they were told to do, reliably. The warehouse had not been doing what it existed to do.</p>
<p>This is the trap that swallows most automation projects sooner or later. Not a technology failure. A measurement failure, and a subtle one, because every individual number on the dashboard is telling the truth.</p>
<h1>What Uptime Actually Tells You</h1>
<p>Uptime measures whether a machine is willing to work. It says nothing about whether the work it is doing is the right work, done in the right order, for the right outcome. A conveyor that is up 98.7 percent of the time can spend that time moving the wrong totes to the wrong stations in a sequence nobody asked for, and the uptime number will not move an inch.</p>
<p>This is not a flaw in the metric. Uptime and MTBF were never designed to answer the question &#8220;is the warehouse succeeding.&#8221; They were designed to answer a narrower, legitimate question: is the equipment we paid for behaving the way engineering promised it would. That is a maintenance question, not a business question. The mistake is not measuring it. The mistake is letting it quietly replace the business question in the room where decisions get made, simply because it is the number that updates in real time and looks the most technically impressive on a slide.</p>
<h1>The KPI That Didn&#8217;t Get Invited</h1>
<p>Walk through enough site visits where an automation business case has just been presented, and you hear the same promise every time: order accuracy will improve. On paper it usually does, at least at the aggregate level. What the presentation never mentions is that accuracy quietly stops being treated as the primary measure of success and starts being treated as an assumption. The project team optimizes what the new system exposes clearly: throughput, uptime, picks per station per hour. The old KPI, the one that actually described whether the customer got what they ordered, keeps running in the background on a report nobody reviews in the same meeting as the automation numbers.</p>
<p>That separation is the real cause of the scene above. Two functioning measurement systems, reporting to two different audiences, on two different cadences, neither one wrong, neither one sufficient on its own.</p>
<h1>A Station That Is Fast and Wrong at the Same Time</h1>
<p>Consider a goods to person station running at its designed picks per station per hour, hitting target every shift. Upstream, the wave release logic has been tuned to keep that station fed and busy, because keeping the station busy is what the WES is scored on. What the wave logic is not scored on is whether the totes it releases match the actual priority of outbound orders. So the station processes exactly what it is fed, quickly and accurately, while the highest priority orders for that afternoon&#8217;s truck sit further back in the queue.</p>
<p>The KPI board shows a station performing at or above design capacity. The dock shows a truck leaving with the wrong orders on it. Both are correct readings of what they measure. Only one of them was ever the point.</p>
<h1>Anchoring the New Metrics to the Old Question</h1>
<p>The fix is not to abandon uptime, MTBF, or throughput. Automated systems are genuinely large capital investments and they need to be watched for their own sake. The fix is refusing to let any machine centric metric stand alone in a review without its paired outcome metric next to it. Uptime paired with order accuracy for the same shift. Throughput paired with on time departure. Picks per station paired with priority fulfillment, not just volume fulfilled.</p>
<p>When the pair moves together, the system is doing what it was built for. When the pair diverges, that divergence is the actual signal worth investigating, far more valuable than either number alone. A warehouse that treats a gap between machine performance and business performance as an anomaly to be quietly explained away will keep having quarters like the one above. A warehouse that treats that gap as the primary thing to monitor will catch the problem before the customer does.</p>
<p>The KPI did not become obsolete when the forklifts were replaced by conveyors. It just stopped being the number everyone was looking at in the room where it mattered.</p>

Uptime Is Not the KPI You Think It Is
