Priya.VEditor
A missing statistic is not a zero and it is not permission to substitute an unrelated number. The Office for National Statistics (ONS) information page for its Digital Economy Survey says collection was paused and data in respect of 2023 were not collected. That matters because the survey was intended to provide evidence about UK businesses' use of information and communication technology and the value of electronic sales. A reader studying business ecommerce cannot treat that missing year as if the original measurement programme continued. Nor can they fill the gap with consumer card payments or a retail internet sales share without explaining that these sources cover different populations.
The practical problem is a break in directly comparable business-oriented evidence. UK firms may sell through marketplaces, trade electronically with other businesses, receive digital orders or provide services online. Some sales are paid for by invoice, not a consumer payment card. The retail internet sales series describes a narrower retail context, while the ONS's Visa analysis provides insight into UK cardholders' payment channels. These sources remain valuable but answer different questions. The correct research response is to make the survey pause visible, distinguish what is still measured and avoid implying that a polished chart makes incompatible numbers comparable.
.jpeg)
What the survey was intended to measure
The ONS Digital Economy Survey provided evidence on information and communication technology use by businesses and electronic sales under the survey's statistical definitions. Those subjects extend beyond the checkout activity of conventional online retailers. A manufacturer might receive structured electronic orders from distributors. A software business might sell subscriptions to other enterprises. A service firm might use digital systems for clients while processing payments through normal business invoicing. A business survey can ask questions about these activities that consumer card data cannot observe. However, what counts as an electronic sale depends on the survey questionnaire, reporting population and reference period. Researchers must retain those definitions when comparing historical values.
Why the reference year 2023 matters
According to the original ONS survey notice, no data were collected in respect of 2023 following the pause. That is a specific statement about a statistical collection, not a statement about whether the UK's digital economy functioned. Businesses continued to trade, but the missing survey observation cannot reveal the exact value of their electronic sales. A time-series chart should show a break rather than draw a continuous line through the absent year. If an analyst constructs an estimate, it should be clearly labelled as a separate model with an explained method and uncertainty, not published as though it were an ONS survey response.
Do not confuse the website update date with the study period
A statistical notice can be updated after the observation year to which it refers. The earlier version of this article recorded 19 March 2024 as a webpage update date; that should not be misread as a 2024 business survey measurement. The publication date of a notice, the planned collection year and the reporting year of the underlying business data are different pieces of information. A research citation must distinguish them. It is also unsafe to infer from an earlier pause notice that all later collection was permanently discontinued. For any new report, investigate whether the ONS published a successor programme or documented a methodological restart.
What a missing point in a chart should look like
When a survey has no observation for a year, a chart should identify that year as missing and explain why. It should not represent zero because zero would claim the measured activity did not occur. It should not silently interpolate between adjacent values because the result would be a calculated estimate, not direct evidence. It should not re-label a percentage from another source as the old survey's output simply to avoid a visual gap. A visible break is analytically useful: it reminds decision makers that their confidence about the trend is lower than the smooth graphic might otherwise suggest.
Why retail internet sales cannot repair the survey
The ONS Retail Sales Index includes information about internet sales within the defined retail business population. This can be very useful when the question concerns the share or value of retail goods sold through online channels. But business ecommerce includes more than retailers. Manufacturers, wholesale businesses and professional service suppliers may operate digital order systems outside the retail definition. Even among retailers, a channel-share percentage differs from a whole-economy value of electronic sales. Converting one into the other would require additional evidence about the population, units, coverage and business categories rather than merely copying a number into the missing year.
Why a consumer payment series is another different population
The ONS receives aggregated, anonymised Visa data that distinguish spending with domestic and international merchants and identify card-not-present transactions. These records capture certain household purchasing patterns, but many enterprise sales are paid through bank transfers, credit accounts, invoicing platforms and corporate procurement systems. A Visa-based online share does not measure all business electronic sales. It may cover consumer purchases from merchants who also sell to businesses, but that overlap does not make its denominator equivalent to a survey of enterprises. An accurate analyst treats it as a separate line of evidence about payments, not a replacement observation.
The category of digital trade adds another distinction
The ONS's 2026 digital trade research includes developing estimates of household digitally ordered imports using payment-data proxies. That concept concerns cross-border purchases arranged over qualifying digital networks; it does not describe the value of all electronic sales by UK businesses. Software might be ordered online and delivered electronically, while clothing can be ordered online and delivered physically. A transaction may also take place through a platform intermediary. The OECD, IMF, UNCTAD and WTO Handbook on Measuring Digital Trade distinguishes these concepts so researchers do not treat online payment, ordering and delivery as synonymous.
Unit consistency is non-negotiable
An enterprise survey can report sales values, percentages of businesses adopting technology or other measures. Retail statistics can report pounds, indices or internet share of sales. Consumer card datasets can report spending values or proportions of card activity. These are not directly convertible without additional information. A table containing one percentage beside a pound figure can be informative when the categories are explicitly explained, but it cannot support a numerical growth calculation across the rows. The correct procedure is to preserve units, reference dates and denominators and explain what comparisons are descriptive rather than statistical equivalences.
Survey design defines which businesses count
Business surveys usually specify which industries, company sizes, geographic units and reporting periods are included. A business with several branches or legal entities may report in a different unit from a smaller independent operation. Changes in coverage could alter a survey total even if underlying ecommerce activity did not change. Researchers comparing an older Digital Economy Survey output with a later replacement should investigate the sampling frame, questionnaire and calculation method. A similarly named measure in a new report does not guarantee direct comparability with the earlier collection.
Distinguish survey non-collection from suppression
A pause in collecting data for a period is different from suppressing a value to protect confidentiality. In the first case there may be no survey responses from which to compile the intended number. In the second, relevant observations may exist but are not published in detail. Delayed publication and poor-quality responses are also different explanations for missing evidence. Researchers should use the precise reason supplied by the ONS. Making unsupported allegations that data were hidden or that ecommerce collapsed would go beyond the stated notice and would not improve the analysis.
What the absence means for trend forecasts
When an annual business series has a missing year, it becomes harder to distinguish a smooth underlying trend from unusual changes in specific periods. The safe response is not to extend a fixed growth rate across the gap and call the result an observation. A forecast may still be built, but it should state that some inputs are modelled, identify its assumptions and describe the range of plausible outcomes. Businesses planning new ecommerce services should use more direct evidence, such as prospective customer requirements, existing transaction data and sector-specific market studies, rather than invent a precise national result.
What an analyst can still do well
Start by defining one research question. If it concerns UK retail internet share, consult the official retail internet-sales dataset and its revisions. If it concerns cardholders' remote payment behaviour, use the ONS Visa analysis with clear coverage caveats. If it concerns household cross-border digital ordering, use the developing digital trade methodology and note its modelling limitations. If it specifically concerns UK enterprise electronic sales across sectors during the uncollected year, explain that the original survey does not supply the requested observation. The appropriate measure follows the question, not the convenience of an available headline.
Make a clear source matrix
A useful research brief records each source's responsible organisation, publication date, statistical reference period, covered population, unit, method and known limitations. It should state whether the measure is direct survey evidence, observed payment activity or a modelled estimate. This simple matrix helps prevent a team from mistakenly treating a consumer card percentage as a business-sales total. It also makes future revisions manageable. When a new data release appears, an analyst can decide whether it replaces an older observation or adds an entirely different kind of evidence.
Why internal business records help but do not fill the national gap
A software company can know exactly how many orders its own platform processed and how long its customers waited for fulfilment. This can support improvements to checkout performance, integrations and customer service. It cannot estimate national electronic sales reliably without a representative sampling or estimation method. Conversely, national statistics will rarely identify the cause of a particular business's failed checkout or invoice discrepancy. Internal and national evidence serve complementary roles. A responsible report should state the limits of both rather than present a convenient company example as a proxy for the entire UK economy.
What journalists and marketers should avoid
An article claiming that the ONS has stopped measuring ecommerce entirely would overstate the Digital Economy Survey notice. Other statistical programmes continue to publish information about retail internet sales, card spending and digital trade. But it would be equally inaccurate to say the missing business survey observation has been solved simply because those other data exist. An accurate headline identifies the paused programme and the missing reference year. The accompanying text explains how alternative evidence helps with narrower questions without restoring the uncollected measurement.
How to evaluate a future replacement survey
If the ONS publishes a new business digital economy collection, check whether it uses the same definitions, business sizes, sectors and sampling methods. Compare the reporting time periods and look for a documented bridge to earlier results. A new official figure may be useful for future planning without proving a continuous historical trend. Changes to categories, reporting technology or how digital sales are classified can affect measured levels. Only publish an uninterrupted long-run chart when the statistical authority provides appropriate comparability evidence or the analysis explicitly models the break.
Implications for digital-service planning
A UK software developer deciding whether to build a customer portal, business ordering system or integration should not need an invented national figure. Begin with the operational problem: which process is slow, which systems do not communicate and which errors damage outcomes? Ask for examples and measure the baseline. Review existing software products, information-security obligations and the costs of maintaining a custom platform. National statistical context may help understand general industry conditions, but specific customer evidence is required to justify product scope, service reliability and commercial investment.
The responsible conclusion
The ONS Digital Economy Survey pause creates a genuine gap in its intended 2023 business ecommerce evidence. It does not establish zero sales, prove national digital decline or eliminate the value of other ONS sources. The correct response is to preserve the missing observation, avoid unsupported interpolation and use alternative statistical programmes only for questions they genuinely answer. This is a useful discipline for technology reporting more broadly: when evidence is incomplete, a clear account of the limits is more trustworthy than a seamless story built from incompatible figures.
Primary sources: ONS Digital Economy Survey information page; ONS retail internet-sales dataset; ONS Visa consumer card research (6 February 2026); ONS Digital trade development (2026); OECD/IMF/UNCTAD/WTO digital-trade definitions. The exact survey pause notice was not accessible through this session's public web retriever; the pause statement remains drawn from the previous official source record and should be rechecked before formal external citation.