Definition
KPI Dashboard (KPI Dashboard) is a visual board on the ICMP platform that displays key business metrics such as card activation rate, usage, revenue, and cost. Together with Global Dashboard (Global Dashboard) and Usage Analytics (Usage Analytics), it supports operational decision-making, performance evaluation, and anomaly detection.
Typical KPIs
Typical KPIs on the KPI Dashboard include: Card-related — total cards, activation rate, suspension rate, termination rate, new/lost cards; Usage-related — total traffic, per capita usage, Overage proportion of over-usage, pool utilization rate; Revenue and Cost — MRR, ARPU, TCO, gross margin; Operations-related — number of alerts, ticket response, customer satisfaction. It supports drill-down by time, customer, region, and project, as well as threshold alerts and trend comparison.
Relationship with Global Dashboard
The KPI Dashboard focuses on key performance indicators and is usually a subset or thematic view of Global Dashboard. The Global Dashboard provides an overall view, while the KPI Dashboard can be customized for different roles such as management, operations, and finance, highlighting the metrics they care about. Both rely on underlying data sources such as Usage Analytics, CDR, and billing data.
Implementation in ICMP
The SimtoLink ICMP KPI Dashboard supports multiple roles, multiple dimensions, and is configurable; it is deeply integrated with Usage Analytics and Cross-Border Billing; it supports export and API integration; combined with Event Subscription, it enables KPI anomaly alert push notifications.
Define the Measure before Drawing the Chart
A KPI dashboard shows a small set of performance measures that have already been defined, on a cycle, so a specific role can judge whether operations or the business have drifted from the target. It is not a container for every operational number. Each measure must first state its numerator, its denominator, its time zone, and whether it counts only closed records. If the denominator of an active share includes stock SIMs not yet delivered, the inventory plan pulls the figure down. If usage adds a shared pool again once per SIM, the total sits above real consumption. Charge measures must also say whether they use the charging currency or the settlement currency. Otherwise a comparison across countries has no meaning.
Before use, the measure owner, the refresh cycle, and the thresholds must already be agreed. Management looks at outcomes, such as revenue mix and the number of customers over a tenant quota. Operations looks at process, such as alerts still open and provisioning requests that failed. The two kinds of number can come from the same platform, but they should not be drawn on one chart with no note. A threshold alert must point at an action someone can take, not only turn a color red. Quota utilization fits as a process measure. Formal revenue still follows the invoice. When a measure’s definition changes, either recompute the history or mark a break on the chart, so readers do not see a definition change as a jump in the business.
Treating Process Counts as Business Results
A common error is to call alert counts, API call counts, or the number of people online a performance result. More alerts may mean the rule became more sensitive, or that devices are genuinely abnormal. Without the rule version, the number cannot judge operational quality. Another error is to compare across cycles by force. Usage counted by calendar month and usage counted by billing month differ, in the last days of the month, by a whole stretch of late roaming CDRs. Putting test SIMs, employee SIMs, and commercial SIMs into the same activation rate also makes a pilot look like large-scale churn.
If a drill-down drops its filters, the detail behind “activation rate in one region” becomes a worldwide SIM list, and the reader thinks the measure was calculated wrong. An exported measure must also be traceable to a definition document or an audit trail, showing who changed the formula and when. When reading a KPI dashboard, first see which role the chart belongs to, which cycle it uses, and what the denominator includes, then decide whether it can enter a business conclusion.
When a measure is retired or renamed, the old chart should not silently point at the new formula. Keep the stop date of the old name, and note on the new chart from which day the new denominator applies. When comparing two regions, confirm that both sides exclude test SIMs and stock not yet delivered. Excluding them on only one side turns a regional difference into a definition difference. Write that note beside the chart, so the reader does not have to reconstruct the counting rule from memory.
A snapshot and a queue are not the same kind of number. The SIM count online on a given day is the SIMs still in service at the end of that day. The count of SIMs activated at any time this month includes SIMs already terminated during the month. Dividing those two produces a ratio that cannot be explained. Each measure should declare whether it is a point-in-time snapshot, a total over an interval, or a count of distinct events. Alert counts should also declare whether they count events or SIMs. When one SIM fires several times in a day, the two methods differ by a long way.
Counting only SIMs still in service hides failures that have already been terminated. If the denominator of an activation rate excludes SIMs terminated this month, churn looks like an improvement. The opposite, a denominator that keeps adding every historical SIM, lets early scrap depress today’s ratio for a long time. The definition should say how long terminated SIMs stay in, from which day test SIMs are excluded, and whether stock SIMs enter the denominator. On the day the definition changes, break the chart. Do not join it to the old series as one line.
A threshold is a prompt to look at the detail. It is not an automatic conclusion. Several days above the threshold may mean the rule is too strict, or that one project’s usage mix has changed. The handling record should say which detail was checked, and whether the end was a threshold change, a business change, or a confirmed anomaly. A red point with no handling record cannot, at the next review, explain why nobody acted. Keep charts for different roles separate, so finance does not use an operational process measure to judge revenue.