Two kinds of question that need two kinds of answer — and what happens when one system is asked for both.
Operational reporting answers questions about current state from live transactional data: what is on back order, which invoices are overdue today. Analytical reporting answers questions about pattern and trend across time and dimensions: how margin by customer segment has moved over three years. The data shapes and the performance characteristics are different.
Running heavy analytical queries against a live ERP is the most common cause of ERP performance complaints that are not configuration problems. A query scanning three years of line items to produce a trend competes with the transactions people are trying to post. Separating the two is usually cheaper than tuning a system being asked to do both.
Run from the ERP, against live data, scoped tightly — today's despatches, open orders, overdue invoices.
Built over extracted or replicated data, structured for aggregation by dimension and period.
Nightly is adequate for most trend analysis. Near-real-time costs more and is rarely needed for questions about patterns.
Revenue should mean one thing. Measures defined independently in each report is how two numbers for the same thing arise.
A curated set of reports and measures, rather than hundreds of user-built variants nobody can reconcile.
A sales analysis report takes forty minutes to run and times out during business hours. The request is for a faster server.
The report is scanning three years of transaction lines directly from operational tables, joining to customers, products and territories, and aggregating at run time. It is an analytical question being asked of a structure designed for transactional writing. No amount of hardware fixes that comfortably, and running it during business hours competes with the system's actual job.
The right answer is to separate the workloads: operational reports read live transactional data and return small, current result sets, while analytical reports read a structure built for aggregation and refreshed on a schedule. The same question then answers in seconds against pre-aggregated data that is a few hours old — and for a three-year sales trend, a few hours old is indistinguishable from live. Trying to serve both needs from one structure is why many ERP reporting complaints never resolve.
| Term | What it means |
|---|---|
| Operational report | A list used to do work today — orders to pick, invoices to chase, stock to count. |
| Analytical report | Aggregated data used to understand a pattern or make a decision. |
| Data latency | How current the data is. Operational needs now; analytical usually tolerates overnight. |
| Normalised vs star schema | Transactional structures optimised for writing; analytical structures optimised for reading. |
| Semantic layer | Business definitions sitting over the data, so everyone's revenue means the same thing. |
It degrades the transactional experience, and the cause is rarely diagnosed as reporting.
Two reports with different revenue figures destroys trust in both.
Ask what decision changes with fresher data. Usually none, and the cost difference is large.
If every report is re-cut in Excel, the reporting layer has failed and the numbers have forked.
| Metric | What it tells you | What distorts it |
|---|---|---|
| Report execution time distribution | Which reports are slow, rather than an average that hides them. | |
| Reports run during business hours | Analytical load competing with transactional work. | |
| Data freshness against requirement | Whether the latency actually matters for each report, which is often assumed rather than asked. | |
| Duplicate report definitions | The same question answered differently in several places, which is how numbers start disagreeing. |
We build and support the systems this page describes. Start with a free working proof of concept — no cost, no commitment.