Using peekd data to predict the US
retail eCommerce market
October 2019
Felix van Litsenburg
October 2019
Description of the data sources
US Census Bureau data
This paper uses US Census Bureau unadjusted data from August 20191. This is an estimate of total
retail eCommerce sales across the United States, based on responses to the US Census Bureau’s
Monthly Retail Trade Survey. These estimates are released on a quarterly basis.
peekd data
peekd analysed sales data from over 400,000 online shops. We captured $52 BN of eCommerce
sales in Q2 2019, roughly 38% of the total market as estimated by the US Census bureau. We
capture the data daily, at source, and at a transaction level. Our data does not rely on sampling or
surveys.
1 https://www.census.gov/retail/index.html#ecommerce
Summary
This white paper describes how peekd data can be used to predict the US Census Bureau’s
estimate of the United States’ retail e-commerce market. This information in turn lends itself for
further analysis of the United States economy and consumer sentiment.
For Q2 2019, the Census Bureau reported $140 BN of eCommerce sales1.
We predict this will be $144 BN for Q3 2019 and $165 BN for Q4 2019.
peekd’s data can be used as a macro input for more frequent updates on US retail eCommerce
performance, or to predict its future performance. Moreover, it can be broken down into
different product categories, from fashion to furniture, down to a very granular level (e.g.
Fashion Sneakers). This can guide for example investment decisions for specific product
categories.
October 2019
Historical correlation of the data sources
The correlation between the eCommerce sales data tracked by peekd and the estimates of US
Census Bureau is 0.9, and gives an R2 of 0.8. This means peekd’s data mimics the rates of change
present in the US data very closely.
We see an increase in the share of the US eCommerce market that our data captures, because more
shops were added to our sample over time. In Q1 2016, peekd data represented 24% of the US
eCommerce market. In Q2 2019, this number was 38%.
Using peekd data to forecast the US retail eCommerce market
peekd data is available at a daily frequency, and correlates very highly with US Census Bureau data.
This suggests two use cases:
1. Provide more frequent updates on the state of US retail eCommerce
2. Forecast the US retail eCommerce market
For case 1, we exploit the historical 3:1 ratio of US Census Bureau data to peekd data to estimate the
US retail eCommerce market size. This would suggest a Q3 2019 market size of $144 BN.
For case 2, we forecast peekd’s data and then apply the ratio above to obtain a total market
forecast. We used a decomposable time series model2 to forecast our time series, using an additive
model where non-linear trends are fit with yearly, weekly, and daily seasonality, plus holiday effects.
2 Taylor SJ, Letham B. 2017. Forecasting at scale. PeerJ Preprints 5:e3190v2
0
50
100
150
2016
Q1
2016
Q2
2016
Q3
2016
Q4
2017
Q1
2017
Q2
2017
Q3
2017
Q4
2018
Q1
2018
Q2
2018
Q3
2018
Q4
2019
Q1
2019
Q2
US retail eCommerce$bn
Total US retail eCommerce (US Census Bureau)
Tracked by peekd
October 2019
15
16
17
18
19
20Ja
n/20
19
Feb/
2019
Mar
/201
9
Apr/
2019
May
/201
9
Jun/
2019
Jul/2
019
Aug/
2019
Sep/
2019
peekd tracked data$bn
$54BN tracked by peekd for Q3 2019. We estimate the US Census Bureau’s Q3 2019 figure will be 2.6x higher at $144 BN.
October 2019
This predicts the following US retail eCommerce numbers for 2019 (in BN):
Month Point
Estimate
Upper
Bound
Lower
Bound
July $17.3 Actual
August $17.3 Actual
September $19.5 $19.7 $18.9
October $19.1 $19.7 $18.5
November $23.9 $24.4 $23.3
December $19.3 $19.6 $18.7
Using the observed scaling ratio, we can then forecast the US Census Bureau’s numbers. This
forecasts, in BN:
Quarter Point
Estimate
Upper
Bound
Lower
Bound
Q3 2019 $144 $145 $142
Q4 2019 $165 $170 $161
Interpreting the data
peekd data allows for more up-to-date insights into the United States’ retail eCommerce
environment than the US Census Bureau’s data. This makes it a more agile tool to understand
consumer sentiment and the broader economic climate.
Due to its frequency, it can produce very accurate forecasts. Our forecast indicates that US retail
eCommerce will continue to expand throughout 2019, reaching a seasonal peak in the last quarter.
peekd’s data can be broken down further into more granular categories. This will allow industry
experts to make better informed decisions about their industry within retail eCommerce.
peekd c/o Cross Platform Solutions
125 Reichenberg Strasse 10999 Berlin, Germany
Managing Director: Moritz Thoma
About peekd
peekd is a Berlin-based data science and eCommerce start-up. We have developed a
proprietary online point-of-sales database on which we employ cutting-edge Big
Data and machine learning technologies to derive insights. We capture these in our
Online Retail Intelligence tool.
peekd
c/o Cross Platform Solutions GmbH
125 Reichenberger Strasse
Berlin, Germany
Managing Director: Moritz Thoma [email protected]