Home · Solutions · Other solutions
Solution · Other solutionsOne page of service desk numbers finance trusts, rebuilt every night from the raw data
Cost per ticket and automation share on one page, daily
Robots pull tickets, robot runs and finance's cost sheet every night, reconcile the counts and publish one Power BI page with resolution, automation share, backlog age and cost per ticket.
Executive summary
Three ticket counts, one analyst merging exports, and a pack that arrives ten days after the meeting that needed it.
We build a small, governed data product rather than a dashboard on top of the same exports.
One number per metric: every source is joined under one written definition and a robot checks the count before anyone sees it.
Power BI semantic model and report in Microsoft Fabric; Power Automate alert flow; SharePoint library with the extracts, the definitions page and the lineage record
Business problem
IT reporting
The service desk is measured by people who do not share a definition of a ticket. ServiceNow reports what agents closed. UiPath Orchestrator knows which tickets robots resolved without an agent touching them. Finance divides headcount cost by whatever count it was handed. Each source is right on its own terms, and the monthly pack is a hand-made merge of the three, so it is late, it is a snapshot, and every number in it has been touched.
Because the join is manual, the meeting argues about the denominator instead of the trend. Automation share, the one number that should show what the robot investment returned, is not in the pack at all, because nobody owns the link between a ticket and the robot run that closed it. Queue managers keep private tabs to defend their queues, and finance keeps its own sheet to defend its budget line.
It persists because reporting is nobody's product: an analyst's monthly chore, run from a recipe of exports and pivot tables held in one head, and the people who read the pack have long since learned to discount it.
How it works today
- WaitingThe month closes, but the pack waits until the last queue has been tidied; the analyst's calendar blocks the first three working days
- PersonThe analyst exports closed and open tickets from ServiceNow, then robot job logs from Orchestrator, and asks finance for the headcount cost figure
- PersonEverything is merged in Excel; category names that changed during the month are corrected by hand so that the trend still joins
- PersonCharts go into PowerPoint; queue managers query the numbers and the analyst rebuilds the tabs
- WaitingThe pack leaves on the tenth working day, describing a month nobody can change any more
- SystemFinance builds its own cost per ticket from a separate ServiceNow export with different filters and a different idea of a closed ticket
- Risk of errorThe two numbers meet in the operations review, the denominator is argued about, and nobody acts; next month repeats
Why the current process costs more than it appears
The budget shows headcount, not what it is spent on.
- Analyst days are the visible part. Decisions made late or not at all are the invisible one: which category to automate next, whether a queue is ageing, whether a reopen spike points at a knowledge gap.
- A pack that arrives ten working days late describes a month that cannot be changed; by the time a queue manager sees week two's backlog, it is week six.
- Numbers finance does not trust get used against IT at budget time, and a cost per ticket built on a smaller denominator always looks worse than IT's.
- Automation investment without a measured automation share cannot be defended or extended: the robots resolve tickets every day, and nobody can say how many.
- One analyst knows how the three exports fit together, and the organisation's view of its own service desk leaves with her.
Cost of inaction
Volume follows headcount and every system the centre takes on, so the manual pack takes longer and lands later each quarter. Each new automation adds a source that is not in the pack, so automation share stays unmeasured at exactly the moment the investment is questioned. Finance keeps its own number, and at budget time that number wins.
The quiet part is dependency. Queue ageing is discovered at month end, after the centre's internal customers have noticed. One analyst holds the recipe for three exports, and the organisation's picture of its own service desk leaves on her last day.
A plausible organisation with realistic proportions. The figures are there to be recalculated on your data; they are not a client result.
The shared-services centre of an industrial group, serving 5,100 employees with IT, HR and finance queues in ServiceNow; UiPath automations already resolve password resets and several request types without an agent; a Microsoft Fabric capacity used by finance and a Power BI licence for every manager.
About 3,300 tickets a month, roughly 39,600 a year, across three queues; three data sources; one monthly pack for the operations review.
A reporting analyst spends about 28 hours a month assembling the pack from ServiceNow exports, Orchestrator job logs and a finance cost sheet; six queue managers spend meeting time reconciling it; it arrives ten working days after month end.
No written definition of a ticket, a reopen or an automated resolution, so every join is manual and every number is arguable, and automation share is not reported at all.
A nightly UiPath robot extracts tickets, categories, timestamps and reassignment history through the ServiceNow API; Orchestrator supplies robot-resolved tickets and robot hours; finance keeps the cost base in Excel Online on SharePoint; a Power BI semantic model in Microsoft Fabric joins them under one definitions page, a reconciliation robot checks the counts before publication, and alerts reach queue managers in Microsoft Teams.
The pack becomes a page current every morning, the analyst's 28 hours turn into analysis, six managers stop reconciling, and about a third of reopens become avoidable once their causes are visible. The figures are modelled, not measured.
Proposed solution
We build a small, governed data product rather than a dashboard on top of the same exports. Every night a UiPath Robot extracts ticket records through the UiPath Integration Service ServiceNow connector: identifiers, queue, category, timestamps, reassignment history and resolver, never the description text. UiPath Orchestrator queue and job data say which tickets the existing robots closed and how long they ran, while UiPath Insights stays the automation team's own run-time view. Finance keeps the cost base in an Excel Online sheet on SharePoint that only finance can edit.
A Power BI semantic model in Microsoft Fabric joins the three sources under one definitions page: what a ticket is, when it counts as resolved, what a reopen and an automated resolution are, and how cost per ticket is calculated, in finance's own formula, printed next to the number. A second robot reconciles the model's counts by queue against ServiceNow's totals each night and holds publication on a variance, telling the analyst the difference in Teams. The report is one page of six metrics with drill-down by queue and category.
Delivery is where the managers already are. Power BI subscriptions send the page every morning; a data alert on the backlog-age KPI tile starts a Power Automate flow that posts to the queue manager in Microsoft Teams; the report sits as a tab in the operations channel. Quarterly, UiPath Process Mining with the ServiceNow template runs on the same extracts to show where tickets loop between teams. No AI component: a management pack must give the same answer twice.
Power BI semantic model in Microsoft Fabric with scheduled refresh, subscriptions, data alerts and row-level security; UiPath Orchestrator queues, time triggers and audit; UiPath Insights dashboards; UiPath Process Mining ServiceNow template; Microsoft Teams channel tab for the report; Power Automate flow started by a Power BI data alert
The definitions page and the data model, the extraction and reconciliation robots, the cost-per-ticket logic agreed with finance, the one-page report with drill-downs, the alert thresholds, the quarterly mining configuration and the lineage documentation
ServiceNow extraction through the UiPath Integration Service ServiceNow connector with a read-only integration user; the finance cost sheet through the Microsoft OneDrive & SharePoint connector (Excel Online)
How the automated process works
- AutomationEvery night the extraction robot pulls yesterday's ticket changes from ServiceNow: identifiers, queue, category, timestamps, reassignments and resolver, never the description
- SystemOrchestrator queue transactions and job history are read for every ticket a robot closed, with run time; the cost sheet is read from SharePoint
- AutomationThe semantic model in Microsoft Fabric refreshes and applies the definitions page: ticket, resolution, reopen, automated resolution, cost per ticket
- AutomationThe reconciliation robot compares the model's counts by queue with ServiceNow's totals and holds publication if they differ beyond the agreed tolerance
- PersonOn a variance the analyst gets a Teams message with the difference and the queue, and decides: fix the source, change the definition, or release with a note
- AutomationThe page publishes; subscriptions reach the six queue managers, finance and the CIO before the working day starts
- PersonWhen backlog age in a queue crosses its threshold, the data alert posts to the queue manager in Teams, who acts that day, not at month end
- SystemOnce a quarter Process Mining loads the same extracts into the ServiceNow template to show reassignment loops and the next automation candidate
Human-in-the-loop model
Automation handles
- Nightly extraction from ServiceNow, Orchestrator and the SharePoint cost sheet, with a record of every run
- Applying the agreed definitions and refreshing the model
- Reconciling counts against the source and holding publication on a variance
- Subscriptions, threshold alerts in Teams and the quarterly mining dataset
People decide
- Finance signs off the cost base and the cost-per-ticket formula, and is the only editor of the sheet
- The service desk lead owns the definitions page; a change is versioned with the model
- Queue managers act on backlog and reopen signals
- The analyst investigates variances and reopen causes instead of building exports
Before and after
Systems and integrations
Everything below runs on licences and systems you already hold, or would need anyway.
Inputs
- ServiceNow tickets, categories, timestamps and reassignment history
- UiPath Orchestrator queue and job data
- the finance cost sheet in Excel Online on SharePoint
Automation layer
- UiPath Orchestrator
- UiPath Robots
- UiPath Integration Service
- UiPath Insights
- UiPath Process Mining
Target systems
- Power BI semantic model and report in Microsoft Fabric
- Power Automate alert flow
- SharePoint library with the extracts, the definitions page and the lineage record
Human touchpoints: the Power BI page as a Microsoft Teams tab and by subscription; backlog alerts to queue managers in Teams; variance notices to the analyst in Teams; the definitions page owned by the service desk lead
Technologies used
the governed model with the six metrics, the one-page report, scheduled refresh, subscriptions and row-level security
Aa threshold breach on the backlog-age tile starts a flow that posts to the queue manager in Teams
Anightly extraction and reconciliation on time triggers; queue and job data for robot-resolved tickets; run history and audit
AAPI extraction of tickets, reading the finance cost sheet, variance messages to the analyst
Arun-time, queue and ROI dashboards for the automation team, checkable against the page
Aquarterly reassignment-loop and conformance analysis on the same extracts
Awhere the page is pinned, where alerts and variance notices land, and where the review reads one number
Asource of tickets, categories, timestamps and reassignment history; read-only integration user
AIllustrative economic model
A model, not a promise.
Three pools, each on its own rate, none of them measured at a client: the analyst's 28 hours a month at €44, of which 85% becomes robot work once extraction and reconciliation are automated; six queue managers reconciling 3 hours a month each at €79; and rework on reopened tickets, 6% of 39,600 a year, 18 minutes each at €42, a third of it assumed avoidable once reopen causes are visible. The reopen share is the soft number, and the first the page replaces with your own.
Business benefits
- One number per metric: every source is joined under one written definition and a robot checks the count before anyone sees it
- The pack arrives every morning instead of ten working days after month end, so managers act on the month they are in
- Automation share becomes a reported figure, which makes the next automation decision defensible to finance
- Cost per ticket is trusted by finance because finance owns the formula and the sheet feeding it, and the formula is printed next to the number
- The analyst's 28 hours a month move from assembling exports to explaining variances and reopen causes
- Adding a queue or a new automation adds a source to the model, not a tab to a workbook
The management view
- Status without a meeting: backlog age by queue, reopen hot spots and automation share on one page, refreshed daily
- A queue ageing in week two is an alert, not a month-end surprise
- The definitions page ends the denominator argument and gives internal audit a documented reporting control
- The quarterly mining view shows where tickets loop between teams, which is where the next quick win usually sits
Board-level KPIs
Security and governance
Control is not an add-on.
- The extraction robot uses a read-only ServiceNow integration user; ticket descriptions and attachments never leave the ITSM tool, and the model carries categories, timestamps and identifiers only
- Passwords and API keys are resolved at run time from Azure Key Vault through the Orchestrator credential store, and every run is logged
- Row-level security in Power BI shows queue managers their own queues; finance edits the cost sheet in a SharePoint library with versioning and a Microsoft Purview sensitivity label
- The definitions page is versioned with the model, so any historical number traces to the definition in force when published
- Data stays inside your Microsoft 365 tenant and, for the robots, in the EU region of UiPath Automation Cloud
Why now
Finance functions are asking for cost per ticket as automation budgets rise; a number IT cannot defend will be replaced by one it did not produce, and the modelled €39,500 a year is the smallest reason to move
Power BI on Microsoft Fabric makes a governed semantic model with nightly refresh, subscriptions and alerts a configuration exercise on a capacity many groups already pay for
UiPath Orchestrator and Insights expose robot-resolved work as data, so automation share can finally be joined to tickets instead of being estimated
Relevant executive roles
One trusted page for the operations review and an automation share that justifies the roadmap
Backlog ageing and reopen hot spots seen in week two, not at month end
The same page for IT, HR and finance queues, with a comparable cost per ticket
A cost per ticket built on finance's own cost base and formula, printed next to the number
Common questions and objections
It counts what is in ServiceNow. It does not know robot hours, finance's cost base or what a reopen means to your CFO; the join under one written definition, checked by a robot before publication, is what we build.
Which is why finance owns the formula and the sheet, and the report shows the formula next to the number. A dispute becomes a versioned proposal to change a definition, not a second spreadsheet.
It is a data model, two robots and one page, built so the next three automation decisions are made on numbers. It pays for itself the first time it stops an argument, and it surfaces the candidates for larger projects.
When this is not the right solution
- Ticket data without categories or reliable timestamps; fix the ITSM configuration first, or the page will publish a precise version of noise
- A BI team already delivering a governed service desk model; then we contribute only the automation data and the reconciliation robot
- Fewer than about five hundred tickets a month, where a weekly export answers the same questions
A question for the next management meeting
Given that a cost per ticket finance does not trust will be used against IT at the next budget round, is the number this company reports today one that IT produced, or one it was given?
Implementation approach
The first week looks the same at every client: we look at the data.
We deliver
- A definitions workshop with the service desk lead and finance: the six metrics, what a ticket, a reopen and an automated resolution are, and the cost-per-ticket formula
- The extraction robot for ServiceNow and Orchestrator data, and the reconciliation robot that gates publication
- The Power BI semantic model in Microsoft Fabric, the one-page report with drill-down by queue and category, and row-level security
- Subscriptions, the backlog-age alert to Teams, and the report tab in the operations channel
- A rebuild of last quarter's pack from raw data, showing where it diverges from the published one
- The quarterly Process Mining set-up on the ServiceNow template
- Lineage and definitions documentation, hand-over to the analyst, light support afterwards
We need from you
- API access to ServiceNow with a read-only integration user, and read access to Orchestrator and Insights
- A Microsoft Fabric capacity or Power BI workspace, and a finance owner for the cost sheet
- The service desk lead as owner of the definitions page, and last quarter's pack with its source exports
Stages
Definitions
The six metrics agreed with the service desk lead and finance; the ServiceNow fields identified
Rebuild
Last quarter's pack rebuilt from raw data; divergences from the published version settled
Build
Extraction and reconciliation robots, the semantic model, the page, subscriptions and alerts
Parallel run
One month beside the manual pack, with every variance explained
Scale
HR and finance queues added, the mining view configured, hand-over to the analyst
Quick win. Effort depends on the number of ServiceNow queues and categories, how far the three sources' definitions differ today, and whether finance's cost base exists as a sheet or as a conversation.
Same desk, same month, two ticket counts. The meeting believed one of them.
Send us one month of ticket extracts and your current monthly pack. We rebuild its five headline numbers from the raw data and return a one-page read-out of where they diverge and why.
Recount one month of your ticketsThe neighbouring process usually has the same problem
The board pack should not depend on which analyst merged which spreadsheet on which day.
View solution IT & servicesTask mining: what your shared-service desks really doYour automation backlog is a wish list until somebody measures what the desks actually do all day.
View solution Other solutionsTickets triaged and routed before an analyst reads themPriority follows the requester's capital letters, and a P1 waits in the queue until a dispatcher has time to read it.
View solutionIndustries we deliver this in most oftenManufacturing & industryTransport & logisticsServices & ITShared services