Correcting the record on retail sales, improving quality across economic statistics

Under new senior leadership the ONS is urgently focusing its resources on core economic statistics as part of a wider strategic recovery plan. In this post, incoming Director General for economic social and environmental statistics James Benford says the recent postponement and correction of the retail sales figures underlines the need for the Office to prioritise and put more resources behind work to improve economic statistics and to bring more transparency on how these plans are progressing.
Today we have published retail sales figures for July 2025, alongside updates to previous months’ data to correct for an error in how we were seasonally adjusting them, which required further time to address and quality assure.
I apologise both for the delay to this release and the error in how we have been seasonally adjusting these data. The focus of this blog is what we have changed, why we have made those changes and the broader learnings that follow from them.
During July we started our annual seasonal adjustment review of the retail sales data. In light of feedback from users, we undertook a more comprehensive and in-depth assessment this year.
For retail sales, the approach to seasonal adjustment is unusually complex. Many retailers report their sales on a ‘retail calendar’ basis which split the year into planning and reporting periods that have the same number of weekends, which tend to have a larger percentage of sales, and consistently align holidays and shopping events, such as Easter and Black Friday, each year. Reflecting that, our retail sales figures are collected for blocks of four weeks, four weeks and then five weeks (a ‘4-4-5′ approach) and can both sit within a month or span multiple months. This adds complexity to our seasonal adjustment process, given the need to align the collected data to calendar months. The process is complicated further by the pattern of the blocks of four and five weeks sometimes changing.
Our review revealed that we had not seasonally adjusted data correctly because we did not properly account for the adjustment in retail reporting calendars. We have now corrected for this mistake.
With the corrections, we have revised down the quarterly growth rate of retail sales from 1.3% to 0.7% in Q1 and up from 0.2% to 0.3% in Q2. While the overall pattern of quarterly growth is similar to our previous estimates, the month-on-month changes are now much less volatile.
Retail sales is used to measure the output of the retail sector and part of household expenditure. It accounts for 4.8% of GDP. This means that the contribution of retail sales to GDP growth has been revised down from 0.06pp to 0.03pp in the first quarter and is unrevised to two decimal places in the second quarter. These small corrections will be incorporated into GDP with the Quarterly National Accounts release at the end of September.
We will be conducting a full lessons-learned exercise and will improve our processes and procedures to make sure the error will not be repeated. Retail sales are the only data where we need to align the 4-4-5 approach to data collection to calendar months, so there are no other outputs affected by the error. We will also be working with retailers and plan to move to calendar months reporting periods to remove some of these complexities by the end of next year.
Our broader work as part of our statistics and surveys improvement plans will help reduce the likelihood of these types of errors. We have stopped the integrated data programme and dialled back on broader analytical work that is not related to producing statistics to put more resources into statistical production and survey collection. In addition, we are working to prioritise down to a narrower set of core statistical outputs, to free up more resource. These steps will relieve pressure on stretched teams and allow for more improvement work generally. We will update on our prioritisation work in due course as we engage with our stakeholders and make further decisions.
We will also bring more transparency on the statistical outputs where we have a quality review, or other methodological improvements underway, so users can understand what is driving that and the period over which any concerns will be resolved.
I look forward to working with all users to further improve quality and restore confidence in the ONS’s economic statistics that underpin so many critical decisions. If you have any comments or suggestions to improve our approach we would like to hear from you. You can contact us on econstatsplan@ons.gov.uk.
You can read the ONS improvement plans for ONS economic statistics and surveys here. The findings of the independent Devereux review are here.

James Benford – Office for National Statistics Director General for Economic Statistics