DATA YOU CAN AUDIT
Most panel providers publish results. We publish the reasoning behind them. Every statistical choice, every correction factor, every quality gate that stands between a raw purchase record and a client-facing metric. Because the fastest way to earn trust in this industry is to make verification easy.
Who is in the panel matters more than how large the panel is. If your panelists over-represent Riyadh, every national metric carries a Riyadh tilt. PanelX calibrates against four demographic dimensions (region, socioeconomic class, nationality, household size) using iterative proportional fitting against national census estimates.
Panelists under-report purchases they perceive as unhealthy and over-report purchases they consider responsible. PanelX applies category-specific correction factors derived from cross-validation against retail audit benchmarks.
The longer the gap between a purchase and its recording, the less accurate the record. PanelX targets same-day capture through a mobile-first design that makes recording a purchase faster than forgetting it.
Panelists who stay 12+ months tend to be systematically different from those who leave. If your panel loses 50% annually, the surviving half is a self-selected group. PanelX's behavioral retention architecture maintains panel composition stability.
The bias no legacy panel attempts to correct. When 25-35% of consumption happens in channels your panel does not cover, every metric carries a structural gap. PanelX captures across all channels, eliminating coverage bias at source.
Four dimensions. One system.
A purchase panel is a sample. The value of that sample depends entirely on how accurately it represents the population it claims to describe. PanelX uses iterative proportional fitting (IPF) across four demographic dimensions: region (10 zones), socioeconomic class (4 tiers), nationality (Saudi, Expat Arab, Expat Asian), and household size (3 brackets).
Each dimension is calibrated against national population estimates from GASTAT census data and Ministry of Human Resources labor reports. The algorithm adjusts household weights until the weighted panel distribution matches the national distribution on all four dimensions simultaneously. Convergence is verified mathematically: if the system cannot converge within 10 iterations, the weighting run halts rather than producing unreliable estimates.
Weight capping prevents any single household from exerting disproportionate influence. No household weight exceeds 3x the median (5x absolute ceiling). Effective sample size is monitored: if the weighting process compresses the effective panel below 60% of the raw panel, the underlying sample needs correction through recruitment, not heavier statistical adjustment.
Region
10 zones
SEC
4 tiers
Nationality
3 groups
HH Size
3 brackets
Every purchase record checked for mandatory fields. Incomplete records flagged for follow-up or imputation.
Hash-based and temporal proximity rules identify duplicate submissions. Duplicates suppressed, not deleted.
Missing or implausible prices corrected via IQR cascade: brand-level median, category-level, then historical modal price. Every imputation logged.
Extreme values (1st/99th percentile) winsorized, not excluded. Panel data reflects real behavior, including bulk purchases.
Velocity checks, GPS anomalies, image hash matching, timestamp clustering. Severity-graded: hold, warn, or suspend.
Every published weighting run includes convergence diagnostics, effective sample size, and design effect. Failed runs are not published.
Key categories cross-validated against retail audit data. Penetration deviations greater than 5pp trigger investigation before release.
Legacy Panels
Approximately 65-75% of Saudi FMCG consumption.
PanelX
100% of the consumption perimeter.
25-35% of FMCG spend in Saudi Arabia happens in channels that no legacy panel measures. This is not a marginal improvement. It is an entirely different measurement perimeter.
A 12-page document covering sampling design, weighting algorithms, correction factors, and quality gates. No marketing. No hand-waving. Just the statistical reasoning behind every number.
Download Methodology PaperRequires a business email address.