This paper uses unpublished retailer-level microdata underlying UK consumer price indices to investigate price rigidity. Based on the conventional method, little rigidity is found in frequency of price change, since the implied price duration is only 5.5 months. However, it significantly underestimates the true duration (9.3 months) as suggested by cross-sectional method. Results also exhibit conspicuous heterogeneities in rigidity across sectors and shop types but weak difference across regions and time. The overall distribution of duration can be decomposed by sector into a decreasing component and a cyclical component with 4-month cycles. Both time and state dependent features exist in pricing. These findings support New Keynesian theories and enable a better calibration to improve the performances of macroeconomic models.