Skip to main content
ELSTAT

Services

From framing the question to putting the number on a decision-maker's screen.

01

Survey & sampling design

Sampling frames, allocation, and instruments that survive being argued about.

Most survey error is baked in before anyone reaches the field. We settle it at design time: define the target population explicitly, choose a sampling scheme that matches the population structure, size the allocation against the precision actually required, then test the instrument until the questions stop being ambiguous.

  • Sampling design document with assumptions and allocation formulas
  • Sampling frame and a reproducible selection procedure
  • Calibrated instrument with enumerator guidance
  • Survey weights and variance estimation procedure
02

Statistical analysis & modelling

From raw data to estimates and their uncertainty, with the computation left open.

Analysis is written as code, not as a sequence of manual spreadsheet steps. That means results can be re-run, inspected by someone else, and refreshed when new data lands without redoing the work. Every estimate ships with its confidence interval.

  • Documented, re-runnable analysis scripts
  • Point estimates with confidence intervals
  • Data cleaning and nonresponse handling notes
  • Sensitivity checks against the load-bearing assumptions
03

Composite index construction

Collapsing many indicators into one number without hiding the choices behind it.

A composite index is a powerful communication tool and an easy one to abuse. Normalisation, weighting, and aggregation choices drive the result, and are often never written down. We write them down: what each dimension measures, why the weights are what they are, and how far the ranking moves when those weights shift.

  • Conceptual framework and per-dimension definitions
  • Documented normalisation, weighting, and aggregation scheme
  • Sensitivity and rank-robustness analysis
  • A computation pipeline that can be re-run each cycle
04

Dashboards & data visualisation

Interfaces that make numbers usable by the person deciding, not just by the analyst.

We build dashboards as products, not as exported charts. Each view answers one clear question, carries the indicator definition where the number appears, and shows when the data was last refreshed. Anything that answers nobody's question does not get built.

  • Responsive web dashboard with access control
  • Indicator definitions surfaced alongside the numbers
  • Data and chart export for reporting needs
  • Technical documentation and operational handover
05

Regional performance & budget monitoring

Programme progress, budget realisation, and agency performance in one view that leadership reads — rather than one assembled the night before a meeting.

Local government rarely lacks data. The problem is that it sits across many agencies under different definitions, so answering one question from leadership means sending requests to six places and waiting a week. We start by reconciling indicator definitions across agencies, connect to the sources that already exist rather than standing up a new database, and assemble the result into one summary that can be traversed down to sector, district, and work unit.

  • An indicator dictionary agreed across agencies
  • Integration with sources already in use, including regional statistics and the performance accountability system
  • Executive summary with traversal by sector, district, and work unit
  • Budget and transfer-fund realisation tracked in one frame
  • Programme status flagging, so delays surface while they can still be corrected
  • Documentation and training so the dashboard stays alive after handover

How we engage

Three shapes that cover most work. Scope and deliverables are written down before we start.

Research project

One question, one full cycle: design, collection, analysis, report. Fits recurring surveys, baselines, and programme evaluation.

Methodology advisory

Your team executes, we hold the method: sampling design review, instrument validation, index computation audit.

Build the data system

When the work repeats every year, a report is the wrong deliverable. We build the platform that runs the methodology.

How we work

  1. 01

    Sharpen the question

    Establish which decision is waiting on the result, and which number is actually needed to make it. This stage routinely cuts the initial scope in half.

  2. 02

    Design before collecting

    Sampling design, indicator definitions, and the analysis plan are written and agreed first. Once the data is in, the method can no longer be chosen neutrally.

  3. 03

    Compute and validate

    Computation runs as reproducible code, cross-checked against an independent source, and stress-tested against the assumptions doing the most work.

  4. 04

    Hand over so it gets used

    Results are handed over in a form your team can run: methodology documentation, the code, and where warranted the platform that runs it.

Start a conversation