6.5Field feedback
The essentials in 30 seconds
- Understand: joins and grain explain 90% of wrong totals — push the work to the source.
- Prove: a wrong report is worse than none — cross-check, parameterise, test under real conditions.
- Doubt: preview OK does not mean report OK — check the PDF and the real volume, no myths.
Farès thread — what the quarter taught him
After one quarter, Farès sums up his practice in Algiers in five sentences: understand his data model, push the work to the source, never distribute without cross-checking, parameterise everything that varies, test under real conditions. And four myths shelved: the preview is enough, more colours is better, performance is tuned at the end, do everything in Crystal.
Keep one principle and dismiss one myth: that is how experience settles in.
What practitioners keep repeating
Beyond features, a few principles keep coming back among those who produce statements every day.
✓ Do
- “Understand your data model”: joins and grain explain 90% of wrong totals.
- “Push the work to the source”: filtering and aggregating server-side changes everything on large volumes.
- “A wrong report is worse than none”: always cross-check before distributing.
- “Parameterise everything that varies”: one generic model beats ten copied reports.
- “Test under real conditions”: what runs fast on 100 rows can collapse on one million.
✗ Avoid
- “If the preview is fine, the report is fine” — no: check the PDF and a real volume.
- “The more colours and charts, the better” — clarity comes first.
- “Performance is tuned at the end” — it is designed from the connection and joins on.
- “Do everything in Crystal” — sometimes an upstream SQL view/command is far more robust.
Ready-to-say script — talking like a practitioner (30 s)
“My data model: [tables and grain understood]. My filter: [pushed to the server, SQL checked]. My statement: [generic through parameters, cross-checked, tested on a real volume with PDF checked].”
Copy-paste template — practitioner routine (5 lines)
practitioner_routine = {
"data_model": "joins + grain understood",
"source_work": "filter + aggregation server-side",
"generic": "parameters {?Mois} / {?Region}",
"proof": "cross-check + real-volume test + PDF",
"dismissed_myth": "preview / colours / perf-at-the-end / all-Crystal: ..."
}
Small business / tight budget
One principle per quarter is enough: data model + cross-check first, the rest next.
Algeria SME
Adapt to local closes: cross-check against accounting and real-volume test every month.
Remote / hybrid
Prioritise written traces at a distance (verified SQL, cross-checked figure, checked PDF, dated).
Cross-cultural
Prioritise local transposition as soon as codes differ: question peers, compare practices.
48h challenge — apply tomorrow
Apply 1 practitioner principle (data model, source work, cross-check, parameters, real test) to your current statement and dismiss 1 myth.
- I know the 5 practitioner principles and the 4 myths to dismiss.
- My practitioner routine (model, server, proof) is started.
- I have done the 48h challenge or scheduled its date.
Farès wants to progress like the practitioners. What does the field say?
Experience cannot be bought: it is built by cross-checking and testing, statement after statement.