Inflation gives rise to inefficient price dispersion in New Keynesian models. Yet empirical analyses suggest that such costs are small (see e.g., Nakamura et al. (2018)). We study price dispersion using a new model that enriches the canonical sticky-price model with an information-gathering activity that describes the firm’s effort to identify the best price to be chosen upon adjustment (akin to e.g., Caballero (1989); Reis (2006)). We refer to this activity as “price-research”, involving for instance a review of the firm’s budget sheets to measure production costs. Firms optimally manage prices allocating resources to price-changing and price-research activities. When inflation is low firms invest heavily in price-research and the reset price is accurately chosen. Instead, high inflation shrinks the firm’s markup rapidly and the firm engages in price changes with less precise information on its own idiosyncratic costs. High inflation leads firms to shift resources from pricing-research towards price-adjustment, generating a less efficient distribution of relative prices. We use a granular dataset from during 2019-2025, to study the size and frequency of price changes. We focus on a period of “moderate” two-digit inflation which subsequently records a four-fold increase. We use the micro pricing patterns to parametrize our structural model and its fundamental frictions. The calibrated model fits several empirical moments better than a simple model where the only friction is due to “sticky prices”. The model suggests that increasing inflation from 20% to 80% per year, as observed in Turkey, increases the costs of inflation by about 150 basis points of GDP, which is about four times larger than the costs predicted by a canonical sticky-price model with only price-adjustment frictions.