Singapore's online shopping stopped growing after 2022. Then it jumped.

The boring bit

Where the data comes from

Four tables from SingStat Table Builder, all downloaded 9 Oct 2026. Online Retail Sales Proportion, Monthly (opens in a new tab) (table M602191, also on data.gov.sg) gives, for each month, the share of sales that were online for all retail, for retail excluding motor vehicles, and for three industries: 104 months from January 2018 to August 2026 (the three industries start in January 2019, so they have 92 months). Retail Sales Value, Monthly (opens in a new tab) (M602171) gives the sales value of retailers in Singapore for the same 104 months, and for retail excluding motor vehicles from January 2019 (the table's title mentions the 2025 = 100 basis, but its values are S$ million: July 2026 is 4,413 in the table and $4.4 billion in SingStat's July 2026 release). Online Food & Beverage Sales Proportion, Monthly (opens in a new tab) (M602271) gives the share of food and beverage services sales that were online, 92 months from January 2019. Individuals Who Made Online Purchase By Age Group, Annual (opens in a new tab) (M920331, from IMDA) gives the share of Singapore residents who made an online purchase, each year from 2017 to 2024.

SingStat's own notes say the retail figures cover retailers in Singapore, whether they sell through shops, online or both, and come from a monthly survey of about 460 retail firms. See its July 2026 release (opens in a new tab), which also gives the total retail sales value for July 2026 as $4.4 billion and the online share as 15.4 in every 100 (the table has since been revised to 15.2 and S$4,413m). Online means the order was received remotely and the price and terms were agreed online, through a company's website, an online marketplace or an app. Orders by phone, fax, email or in-store kiosk or QR code do not count. For food and beverage services the same definition applies, regardless of how or where the food is delivered or collected.

What is not claimed

• This is the share of sales recorded by retailers in Singapore, not everything people in Singapore buy. Purchases from overseas websites are not counted, so the real online share could be higher.
• The data cannot tell us why the share moved. We name the circuit breaker only as the period SingStat identifies in 2020: it ran from 7 April to 1 June 2020 and most physical stores were closed in May 2020 (SingStat newsletter, issue 1 of 2021 (opens in a new tab)). We do not claim that it caused the later changes, and we do not say what caused the 2026 rise.
• The dollars are our estimates. SingStat publishes the total sales value and the online share, not the online dollars. We multiply them month by month: online dollars = sales value x online share, and in-store dollars = sales value minus online dollars (so anything not ordered online counts as in-store). They are dollars as spent, not adjusted for rising prices. Leaving out motor vehicles, parts and accessories gives online dollars 1% to 3% lower in every year, close enough to cross-check the method.
• Each year's figure covers the same eight months, January to August, so 2026 (which has only eight months) can be compared fairly with other years and the November shopping peak does not distort it. For all retail and for retail excluding motor vehicles, a year's share is the online dollars divided by all dollars of those eight months. SingStat publishes no sales values for the three industries or for food and beverage services, so for them it is the plain average of the eight monthly shares. For all retail the plain average gives different figures in some years (for 2020, 13.2 instead of 11.8) because the online share peaked in months when total sales were low.
• Only three retail industries are published (supermarkets and hypermarkets, computer and telecommunications equipment, furniture and household equipment). Clothes, books and everything else are inside the all-retail figure but cannot be looked at on their own. The names above are SingStat's. "Computers and phones" is our short form of computer and telecommunications equipment. "Food and drink" is food and beverage services (restaurants, cafes and the like), which SingStat counts separately from retail.
• The month chart is the average of the same month in 2021 to 2025, and leaves out 2020 and 2026. For all shops it is online dollars over all dollars across those five years; for the other kinds it is the plain average of the five published shares.
• 2020 and 2021 were COVID years, with the circuit breaker and later restrictions. Their figures describe those years and should not be read as a normal pattern.
• SingStat changed the basis of its retail and food and beverage series (from 2017 to 2025 as the reference year, with coverage and weights revised) from the January 2026 release (SingStat note on the change (opens in a new tab)). The online-share history was not revised, so 2026 is on a different basis from the years before it, and is marked with an asterisk. The check for retail is in the questions above and the code below: even if the whole extra 0.95 points in the average year-on-year gap (its January to August 2026 average against its August to December 2025 average) came from the new basis, 2026 would still be about 14.2 in every 100, above the 11.8 to 12.7 of 2021 to 2025, and 6 in every 10 of the extra dollars would still be online. That is a bound under that assumption, not proof of the cause: from January 2026 the online share of food and beverage services dropped to an average about 5 points below 2025's, so a change of basis may move these shares, and we cannot see by how much for retail. We do not compare food and drink in 2026 with earlier years.
• SingStat revises recent months after first release. For example, its April 2026 release (opens in a new tab) gave 15.4 in every 100 for April 2026, and the table now gives 14.8. We use the tables as they stand on 9 Oct 2026.
• The survey of individuals (IMDA) asks people whether they bought anything online. It is a different measure from the shop figures, and the table does not say whether purchases from overseas websites are counted, so the two are shown side by side but not added or compared as one number.
• Nothing here says anything about how well any shop or website is run.

Coverage

Every month in the tables up to August 2026 was used: 104 of 104 for all retail, retail excluding motor vehicles' share and sales value, 92 of 92 for each of the three industries and 92 of 92 for food and beverage services. All 8 yearly rows of the survey table (2017 to 2024) were used. Nothing was left out or estimated, apart from the online dollars described above.

Report an error. Corrections are logged publicly with a date.

import math
import pandas as pd


def half_up(x, places=1):
    # the house rule: round half up (Python and pandas round half to even)
    k = 10 ** places
    return math.floor(x * k + 0.5) / k


# One row per month, from SingStat's tables (see prep/prep_online_share.py)
share = pd.read_csv("online_share.csv", index_col="month", parse_dates=True)    # % of sales that were online
value = pd.read_csv("retail_value.csv", index_col="month", parse_dates=True)    # retail sales value, S$ million


def jan_aug(s):
    # the same eight months in every year, so 2026 (January to August) is comparable
    s = s[s.index.month <= 8]
    return s.groupby(s.index.year).sum()


# Dollars: SingStat gives each month's sales value and the share of it that was online
online = value["total"] * share["retail"] / 100          # online dollars each month, S$ million
total_y, online_y = jan_aug(value["total"]), jan_aug(online)
instore_y = total_y - online_y
retail = 100 * online_y / total_y                         # the share of January to August dollars that were online
print(retail.map(half_up))                                # 2019: 5.2, 2020: 11.8 ... 2026: 15.2

# The 2021 to 2025 band, and 2026 against it
band = retail.loc[2021:2025]
band_low, band_high = band.min(), band.max()              # 11.8 and 12.7
print(half_up(band_low), half_up(band_high), half_up(retail[2026]))

# Of the extra dollars each year, how many were online?
extra_online, extra_total = online_y.diff(), total_y.diff()
online_part = 100 * extra_online / extra_total            # 2024: 0, 2025: 44, 2026: 91
print(online_part.loc[2021:].map(lambda x: half_up(x, 0)))

# Could a change of basis in January 2026 explain the rise? Take the whole extra gap away as a worst case
gap = share["retail"] - share["retail"].shift(12)         # this month minus the same month a year earlier
extra = gap["2026-01":"2026-08"].mean() - gap["2025-08":"2025-12"].mean()
worst_case_2026 = retail[2026] - extra               # about 14.2, still above the band
worst_online_2026 = (value["total"] * (share["retail"] - extra) / 100)["2026-01":"2026-08"].sum()
worst_online_part = 100 * (worst_online_2026 - online_y[2025]) / extra_total[2026]    # about 61 of every 100 extra dollars
print(half_up(worst_case_2026), half_up(worst_online_part, 0))

# Which month? The same calendar month in 2021 to 2025 put together (2020 and 2026 are left out)
five = online.index.year.isin(range(2021, 2026))
month_online = online[five].groupby(online[five].index.month).sum()
month_total = value["total"][five].groupby(value["total"][five].index.month).sum()
month_share = 100 * month_online / month_total          # January 11.4 ... November 15.8 ... December 13.6
print(month_share.map(half_up))

Source: Online Retail Sales Proportion, Monthly (opens in a new tab), Retail Sales Value, Monthly (opens in a new tab), Online Food & Beverage Sales Proportion, Monthly (opens in a new tab) and Individuals Who Made Online Purchase By Age Group (opens in a new tab), SingStat Table Builder (the first is also on data.gov.sg (opens in a new tab)), downloaded 9 Oct 2026. Contains information from data.gov.sg made available under the Singapore Open Data Licence.