Business Data & Reporting
Business intelligence dashboard: numbers people trust
A business intelligence dashboard people trust starts with metric definitions. Why QuickBooks, Jobber and ServiceTitan reports disagree, and how to fix it.
- Published
- Author
- Cass & York Team
- Series
- Cass & York
- Reading time
- 8 min read
It is Monday morning and the owner wants October revenue. ServiceTitan says $164,000. QuickBooks, run on cash basis, says $140,000. The sales manager's spreadsheet says $182,000. Somebody spends the next two hours exporting CSVs to work out which one is wrong.
All three are correct for the question each one answers. Each report counts a different event on a different date. A business intelligence dashboard is a single screen of the few numbers a team runs the business on, pulled from the source systems on a schedule, with each number defined tightly enough that two people asking the same question get the same answer. Most of the effort goes into those definitions: which event counts, which date field, which system, how often it refreshes and who owns it. The charts come last.
This guide is for owners and ops or finance leads at field service, home service and contracting companies whose QuickBooks, Jobber or ServiceTitan reports and spreadsheets disagree. We work through four reasonable definitions of one month's revenue on example inputs, the bridge that ties them together, the small operating dashboard we would start with, and when business intelligence dashboard software beats building your own view.
Why reports disagree: the causes we check first
When two revenue numbers disagree, we check these six causes in order. In our experience the cause is usually a definition, and the data underneath is fine.
Cause | What it looks like | Documented example |
|---|---|---|
Different event | One report counts work sold, another counts work completed, another counts invoices, another counts cash | ServiceTitan's dashboard filters most revenue metrics on Job Completion Date, revenue from invoices outside a job on Invoice Date, and Total Sales on the estimate sold on date (ServiceTitan dashboard business metrics) |
Different date on the same event | Two invoice reports for "October" with different totals | Jobber's Invoices report filters on Created, Issued, Due or Paid date, chosen from a dropdown (Jobber Invoices report) |
Accounting basis | Two people run "the P&L" and get different income | QuickBooks accrual reports show income whether or not customers have paid; cash reports show income once cash arrives (QuickBooks cash and accrual). Any single report can switch method without touching the company setting (QuickBooks accounting method) |
Scope and filters | The dashboard covers all branches; the report covers one | ServiceTitan lists mismatched date ranges, a narrower Business Unit filter, and a dashboard that drops cents as causes of report and dashboard mismatches (ServiceTitan mismatch article) |
Report template | Same data, same dates, different totals | A ServiceTitan administrator reported $10,000 in sales on one template and $15,000 on another for the same date range (ServiceTitan Community) |
Money in transit | Payments in Jobber exceed deposits in the bank | Jobber Payments groups deposits into lump payouts, net of fees, which arrive in 2 to 4 business days after the first payout (Jobber Payments reports). Deposits taken on quotes can post to QuickBooks with no invoice to match (Jobber Community) |
Rolling windows add a seventh, quieter cause. The Issued figure on Jobber's invoice list covers the last 30 days and leaves out today (Jobber key metrics). A tile that reads "last 30 days" and a calendar month report will disagree on 30 days out of 31.
We hear about this setup in discovery calls with field service operators: Jobber for jobs and invoices, QuickBooks for the books, and spreadsheets in between. One of them is an exterior cleaning business growing by acquisition.
One month, four revenue numbers
Here is a home service company's October on example inputs. Every figure below is invented for illustration; the mechanics are what matter.
Example inputs for October | Amount |
|---|---|
Work sold in October (estimate sold date) | $182,000 |
Of that, scheduled for November | $56,000 |
Work sold in September, completed in October | $38,000 |
Completed in October, invoiced November 2 | $12,000 |
Completed in late September, invoiced October 1 | $9,000 |
Maintenance plan invoices with no job attached | $6,000 |
October invoices still unpaid on October 31 | $65,000 |
Cash collected in October on September invoices | $31,000 |
Deposits collected on quotes for November jobs | $7,000 |
The chart shows the same October four ways; on the site, toggle a definition to see which events it counts.
Each number answers a different question. Booked revenue tells sales how much work they won. Completed revenue tells operations how much work crews delivered. Invoiced revenue tells finance what the accrual books will show. Collected revenue tells the owner how much cash came in. Calling all four "revenue" is what starts the Monday argument.
The bridge below ties them together line by line. When every line has a name and a list of records behind it, the disagreement turns into a reconciliation.
Step | Amount | Running total |
|---|---|---|
Booked in October | $182,000 | |
Less: sold in October, scheduled for November | −$56,000 | $126,000 |
Plus: sold in September, completed in October | +$38,000 | $164,000 completed |
Less: completed in October, invoiced in November | −$12,000 | $152,000 |
Plus: completed in September, invoiced in October | +$9,000 | $161,000 |
Plus: maintenance plan invoices with no job | +$6,000 | $167,000 invoiced |
Less: October invoices unpaid at month end | −$65,000 | $102,000 |
Plus: cash collected on September invoices | +$31,000 | $133,000 |
Plus: deposits collected on November quotes | +$7,000 | $140,000 collected |
The bank balance adds one more step. Card payments taken on October 30 and 31 may land in November, and payouts arrive net of processing fees, so deposits in the bank can trail collected revenue for a few days.
How we resolve the disagreement
When a client brings us two numbers that disagree, we follow the same seven steps.
- Put the numbers side by side with their sources. Name the system, the report, the template and every filter used for each one.
- Name the event and the date field. Write down "estimate sold on date," "job completion date," "invoice issued date" or "payment date" for each number. Most disagreements end here.
- Build the bridge. Pull the records that sit in one number and are missing from the other, and group them into named lines like the table above. A line that stays unexplained is a data problem, such as a duplicate job, a voided invoice that synced, or a deposit with no invoice.
- Assign a question to each definition. Sales keeps booked. Operations keeps completed. Finance keeps invoiced and collected. Every one of them stays on the dashboard under its own name.
- Rename the tiles. Replace the tile labeled "Revenue" with three tiles labeled "Completed work," "Invoiced" and "Collected."
- Test on a closed month. The invoiced tile for September should tie to the QuickBooks accrual P&L for September, and the collected tile to the cash basis P&L. Every gap gets a bridge line or a fix at the source.
- Lock closed months. Once finance closes a month, the dashboard stops recalculating it. Late edits in Jobber or ServiceTitan show up as adjustments in the current month, each with a reason.
What belongs on a small operating dashboard
Start with six tiles that fit on one screen. Microsoft's own dashboard guidance says it is best to have all the tiles on one screen, to remove everything but essential information, and to put the highest level numbers top left (Microsoft Learn). A wall of 30 charts gets opened once.
Tile | Question it answers | Typical source |
|---|---|---|
Completed work this month, with invoiced and collected beside it | How much did we deliver, bill and collect? | Jobber or ServiceTitan jobs and invoices, QuickBooks payments |
Booked backlog | How much sold work is waiting to be scheduled or finished? | Sold estimates and open jobs |
Completed, not invoiced | What work is done and still unbilled? | Jobs completed with no issued invoice |
Receivables over 30 days | Who owes us money, and how much? | QuickBooks or Jobber open invoices by age |
Gross margin by job type | Which kinds of work make money? | Invoices plus labor and materials costed to each job |
Close rate | What share of estimates turn into sold work? | Estimates sent versus sold |
Each tile shows one headline number, last month for comparison and a trend line. Clicking a tile opens the rows behind it.
Write a definition card for every number
The card is what makes a number trustworthy. Each tile links to one, and nobody changes the card without the owner's sign off.
Field | Example: Completed work |
|---|---|
Question it answers | How much work did crews finish this month? |
Event | Job marked complete |
Date field | Job completion date, calendar month, company time zone |
Source | Jobber jobs and invoice line items (or ServiceTitan Completed Revenue) |
Includes | Line items on invoices for completed jobs |
Excludes | Maintenance plan billing with no job, deposits, sales tax, voided invoices |
Refresh | Every morning at 6:00, plus on demand |
Shared filters | Branch, job type, technician |
Drill down | Job number, client, completion date, invoice number, amount |
Business owner | Operations lead |
Data owner | Whoever maintains the pipeline |
Change log | Date, change, approver |
Update frequency: match the decision
Refresh speed should follow how fast someone acts on the number. Dispatch and unbilled work matter daily. Margin by job type is a weekly conversation. Closed books are monthly. Power BI limits semantic models on shared capacity to eight scheduled refreshes a day and allows up to 48 on Premium or Fabric capacity (Microsoft Learn). For an operating dashboard, a morning refresh plus a button usually covers it. A faster refresh also stays limited by how often the pipeline collects from Jobber or ServiceTitan, which we covered in how fast customer data needs to update.
Filtering and drill down
Filters have to apply to every tile at once. If the branch filter changes completed work and leaves receivables alone, the screen tells two stories. Keep the date basis out of the filter bar: each tile's card fixes its date field, so a user can change the month and leave the definition alone. Drill down should end at rows with source record IDs, such as the Jobber invoice number or the QuickBooks transaction, so anyone can check a total against the system it came from.
Ownership
Every number needs a business owner who decides what it means and a data owner who keeps the pipeline running. When we built a quoting and pricing portal for a commercial roofing contractor, admins managed pricing, labor and margins inside the portal, so every quote used the same rules. Dashboards work the same way: one owner per definition, one place it is written down. Our guide to a single source of truth across business systems covers assigning authority when two teams own overlapping facts.
A definition card template
Each card fits on one screen and answers the same eight questions. Name: what the number is called on the dashboard. Question: the decision it supports, in one sentence. Event: what has to happen for a record to count, such as invoice created or job completed. Date: which date places it in a period. Source: the system and report or table it comes from. Filters: what is excluded, such as voided invoices, test jobs or internal accounts. Owner: the person who decides what the number means. Refresh: how often it updates and when it was last loaded.
Fill one card for each number before anyone builds a chart. When two reports disagree, compare their cards first; the difference is almost always in the event, the date or the filters.
Business intelligence dashboard software or a custom interface
Most companies already have a dashboard inside each tool. ServiceTitan, Jobber and QuickBooks each report well on their own data and stop at their own edge, which is why the owner ends up with a spreadsheet. The next step is a BI tool, or a view built on your own data layer.
Option | Published price | Good fit | Watch for |
|---|---|---|---|
Reports built into ServiceTitan, Jobber, QuickBooks | Included | Questions answered inside one system | Each sees its own data |
Microsoft Power BI | Pro $14 per user per month, Premium Per User $24, both billed yearly (Microsoft) | Teams already on Microsoft 365 | Sharing needs Pro or Premium; refresh limits above |
Tableau Cloud | Standard edition from $15 per user per month for view-only seats, annual contract (Tableau) | Teams with an analyst who will build in it | Building dashboards needs a Creator seat, priced well above the $15 viewer seat; annual contract required |
Metabase | Open source free; Starter $100 a month with 5 users; Pro $575 a month with 10 users (Metabase) | Data already in a database you control | Needs a database underneath |
Google Looker Studio | Free for creators and viewers; Pro $9 per user per project per month (Google Cloud) | Google Workspace shops, light use | Connector coverage for field service tools varies |
Custom view on your data layer | Build and upkeep cost | People act on the number in the same screen: approve, assign, chase an invoice | Someone has to own it after launch |
Our take: buy the BI tool and spend the budget on the data model. Five Power BI Pro seats cost $840 a year at Microsoft's list price of $14 per user per month, paid yearly. The larger cost is the pipeline that pulls Jobber or ServiceTitan and QuickBooks into one model, deduplicates clients and jobs, applies the definition cards and keeps running when a vendor changes an API. That layer is what makes any of these tools agree with your books.
A custom view earns its cost when the dashboard has to drive an action, for example a queue of completed jobs with no invoice where the office manager clicks to send the invoice. At that point it is an integration or an AI agent writing back to Jobber, with a screen on top. Our article on when to build instead of adding another SaaS tool lays out that decision, and integration platforms versus custom code covers the pipeline side. If you are combining several companies after an acquisition, see connecting reporting after an acquisition.
Start with the two numbers that disagree
A dashboard people trust gives each number a name, an event, a date field, a source, a refresh time and an owner, and it shows the bridge when two of them differ. Get that right for six tiles and the Monday reconciliation goes away.
If your Jobber, ServiceTitan and QuickBooks reports tell different stories, send us the two reports that disagree. We will name the definitions behind each, build the bridge on your last closed month and show you what the six tile version would look like on your data. Our reporting and data platform services cover the pipeline and the dashboard.
Related reading: creating a single source of truth and data silos in a service business. See our data integration services, or bring us the two numbers that disagree.
Sources
- Dashboard business metrics, ServiceTitan Help Center, updated November 2, 2026
- What is causing the mismatches between my reports and modular dashboard metrics?, ServiceTitan Help Center, updated October 16, 2026
- Reporting is Inaccurate and Causing Issues for Our Company, ServiceTitan Community, February 3, 2023
- Differentiate Cash and Accrual basis, QuickBooks Support, updated August 4, 2026
- Choose between cash and accrual accounting methods in QuickBooks Online, QuickBooks Support, September 22, 2026
- Invoices Report, Jobber Help Center, updated October 29, 2026
- Invoices List Page and Key Metrics, Jobber Help Center, updated June 25, 2026
- Jobber Payments Reports, Jobber Help Center, updated May 4, 2026
- QuickBooks deposits not matching invoices, Jobber Community, undated
- Data refresh in Power BI, Microsoft Learn, updated September 18, 2026
- Tips for designing a great Power BI dashboard, Microsoft Learn, updated October 20, 2025
- Power BI pricing, Microsoft, accessed November 3, 2026
- Tableau pricing, Tableau, accessed November 3, 2026
- Metabase pricing, Metabase, accessed November 3, 2026
- Data Studio (Looker Studio), Google Cloud, accessed November 3, 2026
Share this article