Sector databases: the missing link in public policy

We celebrate data collection, the field survey, the digital questionnaire, the major statistical operation. We often forget what makes them lastingly useful: the sector database, that structured information system which receives, normalises and preserves the data of an entire sector, campaign after campaign, year after year. Without it, every survey starts from scratch, every programme reinvents its own indicators, and institutional memory dissolves at the pace of projects. Public debate readily focuses on national statistical capacity or on open data; the truly missing link lies below that, at the operational level: the health, agriculture, education and energy information systems and registers that keep an administration running, or fail to. That link is what this article puts at the centre.

The isolated-survey syndrome

Let us begin with the most ordinary and most costly observation. A large share of the information produced in West Africa takes the form of one-off, project-financed surveys, designed in their own formats, then archived in files that become unreadable as soon as the team disperses. The scale of this scarcity is staggering: according to a pan-African compilation published in Scientific Data in 2025, an African household appears in a well-being survey less than once every thousand years, roughly a hundred times less often than a household in the United States. The data exist, but they are rare, scattered and non-cumulative. With no shared reference framework, each operation reinvents its definitions, units and nomenclatures, and the result is a structural inability to compare over time, to pool across programmes and to learn over the long run.

A sector database breaks precisely this cycle. It fixes, once and for all, the definition of indicators, the data-entry formats, the quality rules and the access rights, then it receives successive data within a coherent framework that outlives teams and funding. The difference is not cosmetic: it separates information that evaporates from an asset that compounds.

Where the statistical system really cracks: the sources

To understand why the sector level is the weak link, we must look at exactly where the deficit sits. The World Bank's Statistical Performance Indicators (SPI) break a national system into five pillars. The pillar that buckles most is not the one for products or dissemination, but Pillar 4, data sources: censuses, surveys, administrative and geospatial data. That is exactly where sector databases live, in the administrative registers and routine information systems of line ministries. Sub-Saharan Africa posts the world's lowest average SPI score, around 54 out of 100, and the World Bank identifies the sources pillar as the most lagging of all. In other words, the weak link in West African statistical systems is not the final dashboard, it is the data factory that should feed it.

When a sector builds a real system: the proof from health

Health offers the best proof that a structured sector information system changes the game, because it has a widely deployed standard: the DHIS2 platform. The figures speak for themselves. DHIS2 is now used by the health ministries of 80 low- and middle-income countries, and more than 100 countries once NGO-led programmes are included. The countries that have deployed it are home to about 3.2 billion people, close to 40 % of the world's population, and more than 40 African nations rely on this platform to manage their routine health data. This is no longer an experiment: it is the backbone of much of public health in the Global South.

Above all, what such a system delivers can be measured. The Senegalese case is precisely documented for malaria surveillance between 2014 and 2017, just after DHIS2 was rolled out. Over that short window, the share of health facilities submitting their monthly report rose from 85.4 % to 97.5 %, and the proportion of expected malaria cases actually reported into the system jumped from 76.5 % to 94.7 %. In three years, the system went from patchy coverage to near-complete reporting. This improvement is in no way automatic: it is the product of a structured database, normalised indicators and continuous monitoring of completeness. It illustrates the central point of this article: it is not the act of collecting that creates value, it is the system that organises and preserves it.

Three sectors, three levels of maturity

What health has largely achieved, other sectors have not yet begun, and the maturity gap is revealing. Comparing the information systems of health, education and agriculture shows exactly where West Africa's sector lag plays out.

• Health: a mature standard. With DHIS2, most countries have monthly, facility-level reporting of normalised indicators. The data are recent, geolocated and cumulative. It is the most advanced sector, though not free of quality problems.
• Education: an annual census, but aggregated. Education management information systems (EMIS) still rely largely on the annual school census, delivered as aggregate data with no individual pupil record, often late, and fragmented across several ministries and institutions. We know how many pupils there are, rarely who drops out, where and why.
• Agriculture: the most broken link. Lacking recent censuses, many countries fly blind. Ghana, for example, has not run an agricultural census in over three decades. Administrative statistics overestimate yields by about 32 % on average compared with field measurements, distorting even the targeting of agricultural policy.

This gradation is no technical fate. It reflects choices about investment and consistency. Where a sector has built a structured, normalised and maintained information system, the data become reliable and useful; where it has gone without, it keeps deciding on ageing estimates. The nuance is worth stating for education, because the lag there is less an absence of data than an absence of granularity. An aggregate school census shows enrolment numbers, but not a pupil's trajectory: as long as the system does not reach the individual level, linking enrolment, attendance, results and dropout stays out of reach, and retention policies advance without knowing precisely who is dropping out, or when. Energy, a sector in full structuring, illustrates the opposite challenge: its databases are to be built almost from scratch, which offers the rare chance to design them interoperable and disaggregated from the outset, instead of repeating the silos inherited from older sectors.

Data without a structured base are a perishable asset. Structuring them sector by sector turns them into a patrimony that outlives teams and projects.

The real problem is not collection, it is interoperability

Having a good system per sector is not enough if those systems ignore one another. The most insidious ailment of African sector data is not only scarcity, it is fragmentation: registers scattered across ministries and agencies, built on different identifiers, incompatible nomenclatures and misaligned calendars. The 2025 pan-African compilation puts it bluntly for agriculture: the data come from incompatible sources, national ministries, regional governments, household surveys, plot measurements, with systematic biases between data types. The same problem runs through every sector. A child can appear in the school register, the immunisation register and the civil register under three different identities, with no system knowing it is the same person.

This is where the value of a well-designed sector database lies: not in accumulation, but in the ability to talk to others. Interoperability means shared identifiers, common indicator dictionaries, standardised exchange formats. Without that discipline, data entry is duplicated, effort is wasted, and the overall view a public policy needs becomes impossible. The data exist, but they remain trapped in silos that do not speak to one another.

The cost of inaction is quantifiable, and heavy

The absence of reliable, interconnected sector systems is no technical inconvenience: it costs billions. The health case is the best documented. According to Africa CDC, up to 40 % of the continent's health spending is lost every year to inefficiencies directly tied to failing information systems: fragmented planning, duplicative delivery systems, poor payroll management, ghost workers and weak procurement. In the Democratic Republic of Congo, Africa CDC estimates losses of about 800 million dollars a year from fraudulent payroll entries alone, with some regions recording up to 40 % ghost staff. These losses are not inevitable: they are the exact mirror of the absence of reliable, linked registers. A clean health-workforce register, cross-checked against payroll, mechanically eliminates the ghost worker.

The logic holds well beyond health. Without complete civil registration, we do not know who is born or who dies: in Sub-Saharan Africa, completeness of death registration remains below 3 %, depriving health policy of its most fundamental data. Without an up-to-date land or agricultural register, we subsidise blind. Without an individual education system, we cannot link enrolment, attendance and dropout. Each time, the cost of inaction is paid twice: in wasted money and in decisions taken on false figures.

What the data shop windows hide

A well-kept open-data portal, an elegant annual statistical report, a colourful ministerial dashboard: these shop windows reassure, but they can mask an empty foundation. An aggregate national indicator says nothing of the completeness of the reporting that produced it, of how old the source is, or of how many facilities never reported. One can publish a flawless enrolment rate built on an incomplete school census, or a national agricultural yield overestimated by close to a third. The quality of a policy is not read in the elegance of its presentation, but in the soundness of the base that feeds it.

This is precisely the conviction that guides CRAD's approach. A sector database is not just software: it is a design effort (data model, indicator dictionary, validation rules, access rights) coupled with an ownership effort by the staff who will feed it and keep it alive. CRAD treats both dimensions as one, because a database no one maintains dies as fast as an abandoned spreadsheet. The goal is never to deliver a tool, but to leave behind an administration able to produce, correct and use its own figures, in full sovereignty. It is the exact opposite of the isolated-survey syndrome: moving an institution from a project logic to a logic of permanent steering.

Designed to last: what separates a living base from a data graveyard

Durability is not improvised, it is designed. Three conditions, too often neglected, separate a base that lives from one that dies when the funding ends. First, access governance must be thought through from the start: who enters, who validates, who consults, under what rules, failing which the data degrade in silence. Second, quality must be instrumented continuously, through automated completeness and consistency checks, as the Senegalese example shows, where it was precisely the monitoring of completeness that pulled reporting upward. Third, local ownership trumps sophistication: a simple, mastered base fed by the staff is worth infinitely more than an advanced system no one can run. The ultimate test of success is not functional richness, it is the survival of the base three years after the provider leaves.

Ultimately, the missing link in West African public policy is neither the will to collect nor the ambition to open data. It is the intermediate layer, that of structured, interoperable and durable sector databases, which turns a succession of surveys into institutional memory and that memory into decision-making capacity. The sectors that build it, such as health with DHIS2, gain reliability and sovereignty over their own figures; those that do without keep steering by guesswork. Building this link means deciding that a sector's data ceases to be a one-off expense and becomes an asset that compounds.

← All insights