Each national data file is examined upon submission. Error checks are performed on all submitted data to find and correct any data recording errors and harmonized the format of each variable from country to country. Each variable is examined for out-of-range codes or unusually high rates of missing or refused responses. The frequency distribution for all key indicators is compared to that for other countries and to previous years, to see if there are any possible anomalies. All potential skip logic errors (questions asked that should be skipped, and questions skipped that should be asked) are examined and all extraneous data deleted from the data file. Each team is sent an initial data quality review, which informs them of any errors in their data, allowing them to respond to or fix the problem. Sometimes, if there is excessive missing data, a team may be asked to either re-contact the respondents which should have been asked the question or to resample enough respondents to make-up for the missing data.

Vendor-supplied weights are examined to ensure that they provide a sample that matches the age and gender distribution for the country, have no high leverage values, and properly represents the age and gender distribution within each sample strata (if applicable). If the weighted distribution does not match the national population, the weights are adjusted by age and gender population data, either provided by the team or, if not, derived from the US Census International Population estimates. If the resulting weights are still not representative, GEM disregards the vendor-provided weights and calculates new ones. Likewise, if the weighted distribution does not match the national population divided by strata, GEM calculates new weights from the age/gender/strata population data provided by the team.

The data files are processed and made available to National Teams two times before the results are finalized. The teams are required to review their data during these initial data releases to check for any potential errors made during the data recording or harmonization process.