Observatoire des tout-petits
The Observatoire des tout-petits gathers and publishes early childhood data for Québec. Every year, hundreds of indicators arrive from about ten providers and more than 100 sources, almost all in Excel and never in the same layout. Witify built a data management platform where artificial intelligence reads those files and imports them in minutes.
- Client
- Observatoire des tout-petits
- Year
- Business type
- Non-profit
The starting point: hundreds of Excel files updated by hand
The Observatoire's website publishes more than 200 indicators on young children and their families. Most are broken down by region and year, often by sex or age group too. They come from the Institut de la statistique du Québec, government departments and specialized surveys.
Before the project, every update ran into the same obstacles:
- Unpredictable layouts: Excel files make up more than 80% of the sources, each with its own merged headers, tabs and footnotes.
- Manual data entry: values were copied over by hand, with the risk of error that comes with it.
- Scattered data: files, documents and web pages lived in several places, with no change history.
- Charts to redraw: every update often called for a developer.
What the platform changes every year
- saved every year
- 2,500 h
- The data entry, checking and chart rework the team no longer does.
- indicators updated
- 200+
- Hundreds of provincial and regional indicators imported every year.
- data points a year
- 100M+
- Hundreds of millions of values read, structured and validated.
AI reads the Excel files, the team validates
The team drops in the provider's file as is, without touching it. Artificial intelligence takes care of the rest:
- Reading: it finds the useful tables, even spread across several tabs or surrounded by notes.
- Dimensions: it recognizes regions, years, sex or age group, wherever they sit.
- Mapping: it matches each series to the variables already defined for the indicator.
- Grid: it fills the raw data grid, shown next to the source file.
The team checks, fixes what needs fixing and saves. A provider can change its layout from one year to the next: there is no import template to maintain.
One database for all the knowledge
What used to live in scattered files is now structured, linked and reusable.
Indicators
More than 200 indicators sorted into 13 themes, with their dimensions and sub-dimensions.
Variables
185 reusable variables (regions, age groups, family types) shared across indicators.
Providers and sources
10 providers and 105 sources, cited automatically under every chart.
So what
50 analyses that turn the data into takeaways.
Charts configured, not redrawn
Once the data is saved, the charts update themselves. The team configures each visualization in the platform, with no outside tool:
- Chart type: lines, grouped bars or stacked bars.
- Regional comparisons: one panel per year, all 17 regions on the same scale.
- Visitor filters: picking one or several regions, showing the values.
- Context: title, chart notes and source under every chart.
- Publishing: on the public site, every chart can be downloaded and opened full screen.
A dashboard for the whole life cycle
Every dataset update opens a cycle that follows the same stages:
- Request to the provider and data reception
- Analysis and import
- Writing the highlights and the "So what"
- Internal, then external validation
- Publication
The dashboard shows how many datasets sit at each stage, which ones have been stuck for more than 100 days and the average cycle time. The team knows where to focus without keeping a separate list.
Rigor and traceability at every step
Every figure the Observatoire publishes has to be explainable. The platform keeps track of everything.
Every publication creates a version you can look back at, with its notes.
Every change keeps its date and its author.
Comparability breaks between years are flagged in the data.
Data can be marked provisional, revised, confidential or paid. Lists filter on these statuses.
An embargo end date holds back a dataset until the right moment.
Methodological notes and highlights are written right next to the data.
The files received from providers stay attached to their dataset.
The team discusses each indicator and each dataset right where it lives.
Every published value can be documented, checked and explained.
Discover our other projects
View all projects
