Nobody is lying. Everybody is counting differently.
Almost never a reporting-tool problem. Four systems answer the same question honestly and produce four numbers, and the meeting spends an hour deciding which to believe.
Thirty minutes · no predetermined package
Which of these is actually your problem?
Six sentences. The one that sounds like your month decides which page to read next.
Revenue attribution
Milestones sent server-side from your own records, so the platform learns from revenue.
Read more →Executive dashboards
One definition per metric, read from the operational records rather than an export.
Read more →Closed-loop reporting
Spend at the top and revenue at the bottom, in one view, defined once.
Read more →Warehouse and modelling
A modelling layer so a definition changes in one place, not nine.
Read more →Metric definitions
The document that outlives the dashboard. Usually the cheapest fix on this page.
Read more →Power BI, Tableau, Looker
The tool is rarely the constraint. The pipeline underneath it usually is.
Read more →What untrustworthy reporting looks like from inside.
Four people, four answers, all defensible.
Cost per lead is reported weekly. Cost per customer is not reported at all.
The monthly pack takes days to assemble and is out of date when it is presented.
A channel gets cut because it looked expensive per lead, and revenue drops a quarter later.
Two dashboards exist for the same question because neither team trusted the other's.
Someone exports to a spreadsheet to do the calculation the dashboard was supposed to do.
Six questions before building anything.
The first one is the one nobody can answer, and it is why the rest exist.
Five stages, and the writing comes before the charting.
Write the definitions
one pageEach metric, its formula, its source and its owner. Signed off by the people who will argue about it later.
Reconcile the sources
the rulesWhere two systems disagree, which wins and why. This is the actual project.
Instrument the milestones
eventsQualified, application, funded — classified from record history rather than assembled by hand.
Send and report
closed loopMilestones pushed server-side to the platforms, and cost per customer reported beside cost per lead.
Hand over the definitions
maintainableThe document, not just the dashboard. A metric nobody can re-derive quietly drifts.
The longer the sale, the wider the gap.
Lending and financial services
A funded deal arrives weeks after the click, and the platform never learns which enquiries were worth having.
Home services
High spend on Local Services Ads with booked revenue sitting in a dispatch system the ad account cannot see.
Legal intake
Very high cost per click and a signed matter worth a hundred times an unqualified enquiry.
Professional services
Long consideration, referral-weighted pipelines, and a first touch nobody was tracking.
What people ask about data and reporting.
What is business intelligence, practically?
Turning records a business already holds into numbers people act on. Most of the work is agreeing what each number means, not building the charts.
Why do our systems report different numbers?
They count different things at different times. The ad platform counts a form fill on the day of the click; finance counts revenue when it is recognised. Both are right about different questions.
What is closed-loop reporting?
Connecting spend at the top to revenue at the bottom in one view, so cost per customer can be compared across channels rather than only cost per lead.
Do we need a data warehouse?
Only when data genuinely lives in several systems and reconciling them is ongoing work. For one or two sources, reporting on the operational records is cheaper and fresher.
What is server-side conversion tracking?
Sending conversion events from your own systems rather than the visitor's browser. It survives ad blockers and consent refusal, and it can report a funded deal instead of a form submission.
Should we use Power BI, Tableau or Looker?
Whichever your team already knows. The tool is rarely the constraint — the definitions and the pipeline underneath it are, and those problems follow you anywhere.
How do we measure marketing ROI properly?
By cost per customer rather than cost per lead, which requires the sale to be connected back to the click. Most stacks lose that connection somewhere in the middle.
What is a single source of truth?
One system designated authoritative for each fact, with every report reading from it rather than a copy. It is a governance decision before it is a technical one.
How often should reporting refresh?
As often as a decision could change because of it. Daily suits most executive reporting; real time is usually a cost with no matching decision.
Can you build this on our existing CRM?
Usually, and it is the cheapest option when there is one main source. Reporting on the operational records is fresher than anything built on a nightly export.
How long does a reporting project take?
The definitions take longer than the build. Expect the first week to be conversations, and expect them to change what everyone thought they wanted.
Who should own reporting once it is built?
Whoever owns the decision it informs. A dashboard owned by whoever built it stops being maintained the moment they move on.