Why Online Shopping Didn't Take Over
Online shopping jumped in 2020, then gave back part of the jump. Six years on, it is about a sixth of US retail and just over a quarter of British.
Key takeaways
- The online share of US retail jumped from 11.9 per cent to 16.3 per cent in spring 2020, fell to about 14 per cent by 2022, and was 17.1 per cent in mid-2026.
- In Britain internet sales peaked at 37.8 per cent of retail in January 2021 and were 27.3 per cent in August 2026.
- Across seven large economies, UNCTAD estimated online retail at about 19 per cent of retail sales in 2020.
- The retail forecast we reviewed got the direction right and the speed wrong.
- Three questions for any forecast: what does the official series show, is it a level or a spike, and does it have a date?
In the spring of 2020, with shops shut across much of the world, online shopping looked as if it had won, and the high street as if it might not recover. Six years later, the official figures tell a different story. Online shopping grew, as everyone expected. It did not take over, and in Britain it gave back a large part of what it gained.
About seventeen dollars in every hundred of US retail
The United States Census Bureau publishes the share of retail sales made online every quarter, adjusted for seasonal swings. The series is the closest thing to a scoreboard for the argument.
| Quarter | Online share of US retail sales |
|---|---|
| Last quarter of 2019 | 11.2 per cent |
| First quarter of 2020 | 11.9 per cent |
| Second quarter of 2020 | 16.3 per cent |
| Third quarter of 2020 | 14.9 per cent |
| Second quarter of 2022 | 14.2 per cent |
| Last quarter of 2024 | 16.2 per cent |
| Second quarter of 2026 | 17.1 per cent |
The spring 2020 jump was real: from just under 12 per cent to over 16 in a single quarter. Then about half of it reversed. By the middle of 2022 the share had fallen back to about 14 per cent, and it did not pass its 2020 peak again until 2025. In the most recent figures, about 17 dollars in every hundred of American retail spending goes online, about one dollar in six.
That is a large business: the Census Bureau put online sales at about 340 billion dollars in the second quarter of 2026. It is also a minority of retail, after three decades of internet shopping and a pandemic that forced people online.
The figures come with a caveat that matters for anyone comparing numbers. The Census Bureau revises the series, and since 2025 it has covered only businesses with paid employees, so older published figures do not always match the current ones. The pattern, though, is clear in every version: a spike, a retreat, and a return to steady growth.
The arithmetic is worth doing once. The online share rose by more than four percentage points in a single quarter in 2020. Over the next two years it gave back about two of them. It then took until 2025 to climb back above the spring 2020 level. Measured that way, the pandemic lent online shopping a few years of growth and took about half of that back within two years; ordinary growth has since carried the share past the lockdown peak.
The pandemic lent online shopping a few years of growth, then took about half of it back.
A lockdown spike in Britain that mostly reverted
The British figures show the reversal more sharply. The Office for National Statistics publishes internet sales as a share of all retail sales each month. In February 2020, before the first lockdown, the figure was 19.1 per cent. In January 2021, during the third lockdown, it reached 37.8 per cent, the highest in the series.
Then it fell. The annual figure was 30.7 per cent in 2021, 27.1 per cent in 2024 and 27.5 per cent in 2025. In August 2026 it was 27.3 per cent.
Measured against February 2020, the British share is about eight percentage points higher now, against a peak gain of nearly nineteen. British shoppers buy far more online than they did before the pandemic, but the peak, when well over a third of spending went online, was a product of closed shops rather than changed habits, and it did not last.
A wider view from the United Nations trade agency, UNCTAD, gives the same order of magnitude internationally. Across seven large economies that together made up most of the world's business-to-consumer online sales, online retail rose from about 16 per cent of retail sales in 2019 to about 19 per cent in 2020. Even at the height of the pandemic, more than four fifths of retail spending in those economies was still not online.
UNCTAD's own country estimates differ from the national series, putting Britain at 15.8 per cent rising to 23.3 and the United States at 11.0 rising to 14.0, because each source defines retail and online sales slightly differently. That is a reason to compare like with like, not a reason to doubt the direction.
Why the forecasts kept running ahead of the shoppers
The forecasters were not wrong about the direction. Jörg Snoeck and Pauline Neerman's The Future of Shopping, a trade forecast for retailers in its second edition in 2022, is a good example of the genre. It cites the UNCTAD point that most retail sales are still made in physical shops, and its broad expectations about online growth have been borne out.
Our own review of the book compared its forecasts with what happened, and reached a verdict that would fit plenty of retail forecasting from the period: direction right, magnitude wrong, too fast and too far. Online share did rise. The pandemic did leave a lasting increase. What it did not do was move the share at the speed the most excited forecasts implied, and twelve of the book's fifty-two forecasts carried no date at all, which makes them impossible to check.
There are reasons rooted in shopping itself, though these are explanations rather than measurements. Buying online is good at some things: repeat purchases, items whose quality is known in advance, and comparison on price. It is weaker at others: things people want to touch, try or take home immediately, and purchases that are part of a day out. For many categories the shop is also where the product is seen first, even when it is later bought online.
And habit cuts both ways. Lockdowns forced people to try online shopping for things they had never bought that way, and some of those habits stuck. But they also showed people what they missed, and when shops reopened, a good part of the spending went back to them.
How to read the next prediction about shopping
The practical lesson is about forecasts rather than shops. A claim that something has changed forever deserves three questions.
- What does the official series show? Statistics agencies publish the figures. A forecast that cannot be set against them is an opinion.
- Is it a level or a spike? A jump during an unusual period, such as a lockdown, says little about the trend until it survives the return to normal.
- Does the claim have a date? A forecast with no horizon cannot be wrong, which is a sign it is rhetoric rather than prediction.
The same questions apply to forecasts in the other direction. Claims that the high street is recovering, or that shoppers have tired of buying online, deserve the same check against the series, which on the current figures shows neither.
For shoppers, the figures are a reminder that the choice between online shopping and the high street is not either-or. Most people now shop both ways, and the proportions vary by country and category. In Britain more than a quarter of retail spending is online. In the United States it is about a sixth.
For retailers, the figures suggest that online shopping is a growing channel rather than a replacement. The figures say nothing about any individual company, and some retailers have done very well online. But a business planning at the January 2021 peak on the assumption that the lockdown share was the new normal would have built for a market more than a third larger, in online terms, than the one that arrived in Britain.
There is a broader point about how forecasts are received. A prediction that the world has changed forever is more exciting than one that a trend will continue, and it is more likely to be repeated. That gives dramatic forecasts an advantage in attention that they have not earned in accuracy, and it is why checking them against the official series, slowly and a few years later, is worth the effort.
Online shopping has not taken over, and on the official evidence it is not about to. It is growing again, steadily, from a higher starting point, which is a less dramatic story than the one told in 2020 and a more useful one.
Common questions
What share of retail sales is online?
In the United States, about 17.1 per cent in the second quarter of 2026, according to the Census Bureau. In Great Britain, the Office for National Statistics put internet sales at 27.3 per cent of retail sales in August 2026.
Did online shopping fall after the pandemic?
In Britain, sharply: from a peak of 37.8 per cent of retail sales in January 2021 to about 27 per cent in 2024 and 2025. In the United States the share fell from 16.3 per cent in spring 2020 to about 14 per cent in 2022, then passed its 2020 peak again in 2025.
Will online shopping replace physical shops?
The official figures do not suggest so soon. Online share keeps growing, but even in Britain most retail spending is still in physical shops.
Sources
US Census Bureau. Quarterly Retail E-Commerce Sales, seasonally adjusted, release CB26-133, 18 August 2026, and time series tsadjustedsales.xlsx (revised 18 August 2026). https://www.census.gov/retail/ecommerce.html. Second quarter 2026 preliminary, subject to revision.
Office for National Statistics. Internet sales as a percentage of total retail sales, series J4MC, dataset DRSI, released 18 September 2026. https://www.ons.gov.uk/businessindustryandtrade/retailindustry/timeseries/j4mc/drsi
UNCTAD (2021). Estimates of Global E-commerce 2019 and Preliminary Assessment of COVID-19 Impact on Online Retail 2020. Technical Notes on ICT for Development No. 18. https://unctad.org/node/32767
Snoeck, J., & Neerman, P. (2022). The Future of Shopping: Where Everyone Is a Retailer, second edition. Lannoo. Read in full by this publication, with a reliability audit and a calibration against the Census series.
A note on scope: series are revised; the US series has covered only employer businesses since the 2025 benchmark. The reasons for the pattern are offered as a reading.