The bottom line
87% of analytics projects never reach production operational impact. The failure is almost never the model or the dashboard - it is the missing bridge between an insight on a screen and an action in an operational workflow. Closing that last mile requires three things: embedding analytics into the tools and processes where decisions are actually made, creating a feedback loop so the system learns from outcomes, and ensuring the data foundation is reliable enough that people trust the outputs enough to act on them. Without those three, dashboards remain reporting artefacts - not decision infrastructure.
In This Article
Why 87% of Analytics Projects Miss the Mark
Gartner's estimate that 87% of AI and analytics projects never reach production operational impact is not a commentary on data quality or model sophistication. (Source: Gartner, 2023) It is a commentary on what happens after the model is built. The model works. The dashboard is built. The project is declared complete. And the production floor continues unchanged, because the insight is on a screen and the action is somewhere else entirely.
The analytics project ends at the dashboard. There is no mechanism for the insight to trigger an operational response. The analyst closes the laptop, the operations manager looks at the chart in the weekly review, and the decisions that the chart was supposed to inform continue to be made the same way they were made before the analytics programme started.
This is the analytics last mile problem. It is not a technology problem. The technology to close the gap - workflow integration, alert routing, work order creation, feedback loops - exists and is mature. The problem is that last mile work is unglamorous, hard to scope, and falls in the gap between the analytics team's mandate and the operations team's system ownership.
87% of analytics projects fail not because the insight was wrong - but because the insight never reached the decision that needed changing.
What the Last Mile Actually Is
The analytics last mile is the layer between the insight on a dashboard and the action in an operational workflow. It includes alert routing (who gets notified, through which channel, when a threshold is breached), workflow triggers (what happens in the operational system when an exception is detected), work order creation (the maintenance request, the purchase requisition, the quality hold that the analytics insight should generate), escalation paths (what happens if the first responder does not act), and feedback loops (how the system learns whether the action resolved the issue).
A predictive maintenance model that identifies a bearing approaching failure has completed the analytics work. The last mile is: creating the maintenance work order in the CMMS, pre-positioning the replacement bearing, routing the alert to the maintenance supervisor and not the quality team, and recording whether the intervention prevented the failure or occurred too late. Without that last mile, the model produces an insight that gets ignored.
The last mile is different for every operational domain. In supply chain, it is the replenishment trigger. In quality, it is the production hold and the root cause investigation workflow. In logistics, it is the customer notification and the rerouting decision. Each domain requires a different integration, different routing logic, and different feedback measurement. That specificity is why last-mile work is scoped separately from the analytics build - and why it is so often left out.
Closing the Insight-to-Action Gap
The practical approach is to define the operational response before building the analytics. Start with the decision: what decision should change as a result of this insight? Who makes it? What system do they use to act on it? What information do they need to make it? The analytics build should be designed to deliver exactly that information, in exactly that system, at the moment of the decision.
For maintenance teams, that means the alert lands in the CMMS, not in an email. For supply chain teams, it means the replenishment recommendation appears in the planning tool, not in a BI dashboard. For quality teams, it means the production hold is triggered automatically, with the supporting data attached, not communicated via a report that someone may or may not read before the next batch runs.
Organisations that close the last mile consistently - that define the operational response before the analytics build, and integrate the insight into the workflow where the decision happens - are the ones in the 13% that Gartner describes as reaching production impact. The 87% are not building worse models. They are building better dashboards that nobody acts on.
The gap between insight and action isn't a technology problem - it's a process design problem. Who sees this insight? When? What are they expected to do with it? The organisations that close the analytics last mile define the answers to those questions before they build the dashboard, not after.