Ensuring Store Data Quality in O2O Products: Processes and Rules

1. Introduction

Store data is a core resource for O2O products and directly impacts user experience; accurate store information is essential for users to locate and consume services.

O2O products aim for both coverage and high data quality, focusing on ensuring the reliability of store attributes such as name, address, phone, and coordinates.

2. Typical Online Issues

Common problems include incorrect store coordinates, non‑standard phone numbers, unreasonable recommended pricing, and other data anomalies that hinder user consumption.

3. Process Overview

The data quality assurance work proceeds from three angles:

Clarify downstream data requirements and define data specifications for each attribute.

For existing (stock) data, identify and fix issues such as missing or erroneous attributes.

For new (incremental) data, enforce strict validation rules to block low‑quality entries at the source.

4. Problem Discovery from Data Application Perspective

Beyond manual field visits, the article proposes leveraging existing online data to detect issues.

5. Single‑Attribute Rules

Validate each attribute individually (e.g., address must not be empty, store name cannot contain special characters, coordinates must be within legal ranges, phone numbers must be valid, and cover images must be present).

6. Multi‑Attribute Cross‑Validation

Check logical relationships between attributes, such as matching coordinates with address, phone numbers with city, and brand with recommended dishes, to uncover deeper inconsistencies.

7. Auxiliary Information Checks

Use supplementary data sources like sales visit logs and merchant verification records to validate store coordinates and detect closed stores.

For coordinate validation, multiple noisy location points are denoised (e.g., using minimum‑circle filtering) and compared against the stored coordinate; a large distance indicates a potential error.

Closed‑store detection relies on the absence of recent consumption verification records and spikes in refund/complaint metrics, prompting sales teams to confirm and update store status.

8. Conclusion

By applying the described single‑attribute, cross‑validation, and auxiliary checks, O2O products can systematically discover and fix problematic store data, improve data reliability, and enhance overall user experience.

Future work includes leveraging user‑generated content and behavior analytics to further automate anomaly detection.