Purpose: Quantitative uncertainties are a direct consequence of averaging, a common procedure when building life cycle inventories (LCIs). This averaging can be amongst locations, times, products, scales or production technologies. To date, however, quantified uncertainties at the unit process level have largely been generated using a Numerical Unit Spread Assessment Pedigree (NUSAP) approach and often disregard inherent uncertainties (inaccurate measurements) and spread (variability around means). Methods: A decision tree for primary and secondary data at the unit process level was initially created. Around this decision tree, a protocol was developed with the recognition that dispersions can be either results of inherent uncertainty, spread amongst data points or products of unrepresentative data. In order to estimate the characteristics of uncertainties for secondary data, a method for weighting means amongst studies is proposed. As for unrepresentativeness, the origin and adaptation of NUSAP to the field of life cycle assessment are discussed, and recommendations are given. Results and discussion: By using the proposed protocol, cross-referencing of outdated data is avoided, and user influence on results is reduced. In the meantime, more accurate estimates can be made for horizontally averaged data with accompanying spread and inherent uncertainties, as these deviations often contribute substantially towards the overall dispersion. Conclusions: In this article, we highlight the importance of including inherent uncertainties and spread alongside the NUSAP pedigree. As uncertainty data often are missing in LCI literature, we here describe a method for evaluating these by taking several reported values into account. While this protocol presents a practical way towards estimating overall dispersion, better reporting in literature is promoted in order to determine real uncertainty parameters. © 2013 Springer-Verlag Berlin Heidelberg.