While preparing a ranking of the suppliers that sell most to the state, the LuxIA team ran into an impossible figure: a process from the municipality of Onzaga — a Santander town of about five thousand people — appears on SECOP II with an awarded value of COP 3,355,609,172,320,196,000. That is more than two thousand times the annual gross domestic product of the whole of Colombia, in a single record. A few places further down, a foundation appears with COP 938 trillion across just four processes, far above the entire national budget for a full year.
There is no corruption scandal here. There is a keyboard. Public information systems are fed by human beings, and typing errors are a documented phenomenon in any administrative database: extra zeros, values entered in cents into a field that expects pesos, the total value of a framework agreement loaded into every line item, fields swapped while filling in the form. SECOP II validates formats, not plausibility: if the official types eighteen digits, the system stores them.
The problem is not that the error exists, but what happens when someone aggregates figures without looking at them. A "total contracted by municipality" sum ends up dominated by a single absurd record. A supplier ranking crowns an unknown foundation above the country's largest construction firms. And a citizen complaint built on the raw figure — "Onzaga contracted for quintillions!" — falls apart in seconds, taking with it the credibility of whoever made it. In citizen oversight, rigour is not decoration: it is the asset.
Four habits protect against these mirages. First, check the order of magnitude against the entity's budget: a sixth-category municipality spends a few thousand million pesos a year; any contract approaching that annual total already deserves arithmetical suspicion, and one that exceeds it a thousand times over is an error, not a finding. Second, always open the process on SECOP II and read the documents: the platform is the primary source and the signed contract states the real value. Third, compare against similar processes at other entities: the price of a kilometre of concrete track or of a school meal has known ranges. Fourth, distrust on principle any single figure larger than the nation's annual budget: it does not exist.
At LuxIA we live with this problem daily, and our policy is not to "correct" the source quietly — the data are shown as the state published them — but to bound our aggregations so that one absurd record cannot contaminate them: defined time windows, exclusion of placeholder values, comparison against the annual procurement plan and against the entity's own history, and a direct link from every figure to the original process on SECOP II so that you can verify it. When a number on our platform strikes you as impossible, the click through to the official file is one step away: that is by design.
The moral goes beyond SECOP: open data without method produces noise, and noise serves opacity — every false alarm weakens the real one. Effective transparency is not only about publishing millions of records, but about reading them with judgement. That is exactly the difference between a data point and a finding.