Predictive Maintenance

Every maintenance job, in real time

The dashboard brings maintenance status into one place, showing High Priority, Open and Booked job counts updated in real time, along with a "The Equipment is not Broken" status that separates equipment still working normally from the genuine work queue. A pie chart breaks jobs down by level of urgency, so team leads can assign engineers and resources in order of need rather than in the order the tickets came in.

• High Priority / Open / Booked job counts, updated continuously
• Pie chart splitting jobs by urgency (Priority Levels)
• Filters by period, branch, product, team and job type

AI works out which equipment should be serviced first

AI/ML models analyse Time to maintenance per device and per product type, so preventive service intervals fit what is actually happening rather than a fixed calendar cycle. The output is a table of equipment due for servicing, ranked by the model's confidence score.

What lets the team trust the results is that the dashboard also opens up what sits behind them: model quality metrics such as Accuracy, Recall and F1 Score; a Feature Importance chart showing which variables weigh most on the prediction; a correlation heatmap of the numeric variables; and the distribution of confidence scores, together with a data quality panel reporting the proportion of missing and duplicate values. Before trusting a prediction, the team can check whether the source data is clean enough.

• Time to maintenance analysed per device and per product type
• Table of equipment due for servicing, ranked by model confidence
• Model quality reported with Accuracy, Recall and F1 Score
• Feature Importance, correlation heatmap and a data quality panel

Reading recurring problems out of the text on maintenance tickets

The notes engineers and site staff type into problem reports usually just sit there. The system pulls that text out and analyses it as a Word Cloud, so the problems that keep coming back are immediately visible: "POS", "WiFi", "won't print" or "touch not responding", for example. That is the starting point for fixing the root cause rather than working ticket by ticket. Alongside it is a maintenance history page summarising the total number of devices, how many are in Critical status, the date of the last service and the next maintenance cycle.

• Word Cloud built from the text of real problem reports
• Maintenance history: total devices and how many are in Critical status
• Last service date and next maintenance cycle, per device

An executive summary page, and an assistant you can ask in Thai

The dashboard is built on Microsoft Power BI and includes a separate Executive Overview page for management, gathering the top-level metrics: total tickets, tickets opened and closed, the number of devices with maintenance jobs, average time to resolution, total Downtime hours and SLA Compliance, plus a Downtime trend chart and a map of service locations.

Embedded in the same page is a Data Assistant window. You can ask it directly in Thai (to summarise the chart in front of you, for instance), and the answer comes back as short bullet points.

• Executive Overview page covering tickets, average resolution time, Downtime and SLA
• Downtime trend chart and a map of service locations
• Data Assistant for asking about the dashboard's data in Thai