Data Systems Engineer

In Progress

EconRadar

Economic intelligence across borders — many sources, many calendars, one comparable read.

Multi-SourceNormalizationMacro DataAnalysts & Researchers
  • Multi-country macro data
  • In active development
  • Case study to follow

The Premise

EconRadar pulls macroeconomic data from several countries and several institutions, normalizes it into one schema, and makes the result comparable — the same indicator, the same units, the same calendar, side by side.

That sentence sounds like plumbing, and it is. It is also the entire job. Anyone who has tried to put two countries' inflation prints on the same axis knows that the analysis is the easy half; getting to the point where the comparison is honest is the work.

The Operational Friction

Macro data is published, not distributed. Every statistics agency, central bank, and multilateral institution ships its own numbers on its own schedule, in its own units, against its own base year, with its own definition of the indicator and its own quiet habit of revising last quarter without telling anyone.

So the analyst does the reconciliation by hand. Download, rebase, convert, align the calendars, decide which source wins where they disagree, and note the revision that moved the series after the chart was already made. It is hours of careful, unrewarding work per question — and it has to be redone every time the question changes, because the reconciliation was never written down anywhere a machine could re-run it.

The cost isn't only the hours. It's that the reconciliation lives in one person's spreadsheet and one person's memory, which means nobody can check it, and the next comparison starts from zero.

The Reconciliation Problem
RECONCILEFIVE PUBLISHERS · ONE INDICATORONE COMPARABLE SERIES

Same indicator, five sources, five different numbers. The analysis isn't the hard part — the reconciliation is.

Illustrative — the shape of the problem, not a data extract

Where the Build Stands

The trace runs amber through the work that is finished and thins where it hasn’t reached yet. It is the whole status report.

  1. INGESTBuilt

    Source connectors

    Pulling the same indicator from several national and institutional publishers, each on its own schedule and its own format.

  2. NORMALIZEBuilt

    One schema, one unit, one calendar

    Rebasing, unit conversion, and period alignment, so two series from two publishers can sit on the same axis without quietly lying.

  3. RECONCILEBuilt

    Where the sources disagree

    Explicit precedence rather than a silent last-write-wins — the disagreement is recorded, not resolved out of sight.

  4. SURFACEIn build

    The comparable read

    Turning the reconciled series into something a person can actually interrogate across countries and across time.

  5. HARDENPlanned

    Revisions, provenance, and the write-up

    Handling silent upstream revisions, keeping every number traceable to the publisher it came from, and the case study that documents all of it.

What Isn’t Claimed Yet

What is deliberately not claimed yet: any number. There is no measured outcome on this page because the system has not finished earning one, and a portfolio that reports results before they exist is worth less than one that waits.

The case study follows the same seven-section template as the others — including the postmortem — and it lands when the build does. Until then this page is what it says it is: a system in progress, described at the resolution it has actually reached.

The full case study — architecture, decisions, what failed, and the numbers — publishes when the build is done.

In the meantime

Read a system that is finished.

Back to Selected Work

Want the write-up when it publishes? me@tesfa.pro