The Hidden Cost of Poor Data Quality in Agriculture

Many farms believe their biggest data problem is that they do not collect enough.

The real problem is usually the opposite: they collect a lot of data, in many places, in slightly different units and definitions — and the numbers do not agree with each other.

Confident decisions based on inconsistent data are more damaging than gut-feel decisions, because they carry the false credibility of being 'data-driven'.

When the feed mill log, the barn sheet and the office spreadsheet disagree, every KPI built on them is suspect.

Suspect KPIs erode trust in the entire dashboard. Once managers stop trusting the dashboard, the weekly review cadence collapses — and with it, the decision discipline that drives performance.

Bad data quality also distorts capital allocation. Operations that look better than they are get scaled; operations that look worse than they are get starved. Both are expensive errors.

Data quality has six recognised dimensions: accuracy (does it reflect reality?), completeness (are all records present?), consistency (do the systems agree?), timeliness (is it current?), uniqueness (are records duplicated?), and validity (does it match the expected format and range?).

A simple weekly audit — sample five records at random, trace each to its source document, reconcile any discrepancy — surfaces around 80% of quality issues before they reach the dashboard.

Define a single source of truth for every metric: feed, weights, mortality, sales. When two sources disagree, the rule is that one of them is authoritative and the other must reconcile to it.

Validate at entry, not at reporting. A field that requires a date should not accept text; a weight that must be in kilograms should reject pounds.

A finishing unit reported FCR of 2.4 — well below industry benchmark. The owner celebrated, scaled the operation, and committed capital to a second site on the strength of the number.

A subsequent audit revealed the source of the magic: feed deliveries were being recorded net of moisture (after a moisture-correction adjustment), but consumption was being recorded as-fed.

True FCR, on a like-for-like basis, was 2.7 — entirely in line with industry benchmark, but materially below the assumed operating margin used in the expansion business case.

The cost of bad data was not the wasted few hours of recording. It was the $90,000 annual margin gap on the existing site, plus the millions in misallocated expansion capital on the assumption of a number that was never real.

• Define one source of truth for every metric — feed, weights, mortality, sales — and document which system owns each.
• Standardise units and timing of capture across barns, sites, and reporting periods.
• Run a 15-minute data audit every week before the KPI review. Sample five records, reconcile to source, log discrepancies.
• Validate at entry, not at reporting. Reject incomplete or out-of-range records at the point of capture.
• Tie data quality to a named owner per data domain. Quality without ownership decays in weeks.
• Treat any KPI that beats the industry benchmark by more than 10% as suspect until independently audited.

The most expensive number on a farm is the one you trust but should not. Data quality is not an IT problem — it is a P&L line, and an under-managed one in most operations.