Culture

Business websites

.com

Brand protection

Online stores

Hosting

Data privacy

Policy-making

Registrars

Investing & reporting

Internet Basics Series

How to

People

Cybersecurity

Web trends

Tech news

Security

COVID-19

Technology detection

domains

Software

WordPress

TLDs

GDPR

Veronika Vilgis

 In fact, in the first four months of the pandemic (Feb 2020 - May 2020), we see a 151% increase globally in the number of newly discovered e-commerce websites compared with the same period in 2019 and in line with reports from individual countries. 

Two years on, we wonder how many of those e-commerce sites are still active and what distinguishes the successful ones from those that didn’t survive?

To do so, we analyze around 111,000 US e-commerce websites that were created between February and May of 2020. We then follow their monthly trajectory to see how many are still online two years later, in July 2022. 

First, we have a look at the geographical distribution of those newly created e-commerce sites (based on a subsample that has identifiable state information available). Not surprisingly, California is by far the state with the largest number, followed by New York, Florida and Texas. These states are among the most populous, so to get a better idea which state has advanced their online market, we adjust numbers by population (US Census Bureau). We then see that the most online shops per 100,000 inhabitants were created in Wyoming, Delaware and Nebraska. Wyoming seems to be an outlier here as the rate is twice as high as in Delaware: 56 vs 28 e-commerce sites per 100,00 inhabitants. Half of the states fall within the range of 5.3 to 8.5 e-commerce sites for every 100,000 inhabitants.

Next, we are interested in attrition over time. Figure 2 shows that as time progressed, more and more sites went offline. Specifically towards the end of the first year, many of the newly created sites disappeared. Those that made it through the first twelve months are more likely to stay online as the rate of sites going offline slowed down and then went up again towards the end of the second year.

But can we see if there are other factors that may play a role in whether an e-commerce site stays active or disappears?

To do so we look at two of our proprietary metrics: Changes and Economic Footprint. Changes refers to the number of month-on-month changes to any of the data fields we collect. Examples are changes to the word count on the website, keywords, email addresses and many more. The Economic Footprint, on the other hand, is a metric that gives an estimate of the economic impact of the website and takes into consideration information about the number of visitors, the turnover and the size of a particular business.

When we compare those e-commerce sites that are no longer active with those that are still active two years after their creation, we see that in the first few months, specifically around September 2020, those that disappeared had on average a higher Change score suggesting that these owners were more active. However, after this initial surge, the overall activity dropped off relatively quickly. In contrast, the ‘winners’ started off with less activity but maintained a relatively high activity level that would peak around December, perhaps suggesting that these businesses took advantage of the Christmas boom. 

When looking at the Economic Footprint, we observe that both ‘winners’ and ‘losers’ started nearly at the same level of economic activity. However, those that are no longer active after two years lost momentum more quickly and their Footprint dropped more quickly.

What are some of the possible reasons for such a high attrition? First of all, we don’t have a benchmark (yet). It’s perfectly possible that this is the normal course of many online businesses. Data from the U.S. Bureau of Labor Statistics shows that approximately 20% of new businesses across sectors fail during the first two years of being open, 45% during the first five years and 65% during the first ten years. The rate we see here is considerably higher: 39% of e-commerce sites are no longer active after the first year while 54% have dropped off after the second year. 

So, could this be due to the pandemic or is it just a phenomenon of online stores? When we take a deep dive into individual data points and look at the type of e-commerce that no longer exists, we see quite a lot of stores that were selling masks. The demand for masks has diminished rapidly after the initial shortages at the beginning of the pandemic, so it makes sense that these sites no longer exist after two years. As it’s relatively easy to set up an online business quickly, it may be uniquely suited to leverage short-lived trends. 

Another reason for the high attrition likely reflects the move back to in-store businesses in many sectors. As store owners no longer need to follow mandatory closures or social distancing measures, small business owners may decide to cease any online activities they have set up during the initial stages of the pandemic. 

Taken together, this analysis has shown that of the 110,000 US e-commerce sites created between February and May 2020 only around 28% are still actively operating two years later. Whether this is a unique phenomenon of the pandemic rush to go online or reflects the typical survival rate of online stores remains an open question. Regardless, overall rates of survival seem lower compared to traditional businesses perhaps due to the ease of creating an e-commerce site. 

Public Sector

Pandemic lockdowns forced many small businesses to move their operations online. For others, it offered new opportunities to create an income through e-commerce.

E-commerce: Pandemic winners and losers

Christian Branbergen

Prof. Dr. Michael D. König, associate professor at the Vrije Universiteit Amsterdam (VU Amsterdam) and affiliated with the ETH Zürich Switzerland, did research in collaboration with Dataprovider.com in which he used structured data from company websites to analyze the effects of the pandemic on businesses across time and different regions. 

"My research looks at how companies are affected by different types of shocks. In the past, we have mostly studied how companies are affected by technological change. More recently, we have looked into how companies are affected by the COVID-19 pandemic and we are looking into new ways how to measure this effectiveness by firms. One very promising way seems to be the analysis of company websites. In this respect, we were very happy to collaborate with Dataprovider.com to get information about company websites and how they write about the recent COVID-19 pandemic. So my research tries to use company websites as a new data source to measure how companies are affected by the COVID-19 pandemic and the different policy measures that have been implemented, for example, across different regions, sectors and also across time. For the purpose, we try to look at websites of companies and try to understand how companies write about the pandemic and whether we can get some information out of this. 

And we have done some preliminary analysis using the Dataprovider.com search interface which made it very convenient for us using a keyword-based approach. And we could, for example, investigate and understand how firms in Switzerland were affected during different times of the pandemic. In Switzerland in 2020 and 2021, there were two waves and we could see that in some sectors, firms were exposed more broadly to production shocks during the first wave of the pandemic while this was less of a problem in the second wave of the pandemic.

So about a year ago, we started the collaboration with Dataprovider.com using the website content to understand how or which companies mentioned the keywords corona or one of its synonyms. And then we tried to see which sectors and which countries across the world were mentioning corona the most. And we provided simple, descriptive statistics and rankings over sectors and countries and how they changed over the first month in the year 2020. 

Based on that, we were then interested to see in which context companies mentioned the corona pandemic on their websites. So we developed a set of keywords that allowed us to identify whether companies mentioned corona in the context of a production problem, a supply problem or a demand problem. And we were then able to better understand the specific problems companies were facing: which sectors are most affected by that and how that has changed across time and also across regions? 

We did that for a specific country, Switzerland, where we also had alternative data sources to kind of compare the outcomes from this web based approach. And we did find that the website information could be used to identify production problems on company websites. So my future research in collaboration with Dataprovider.com will try to look into how the effectiveness of companies measured through mentioning the COVID-19 pandemic on their websites differs across countries. 

We started with an analysis of a specific country, Switzerland. In the next step, we would like to expand this analysis to other countries because this would allow us to evaluate the policies that have been implemented by different countries and how the pandemic has spread across different countries over time. 

Another direction that we would like to explore is to see the frequency with which companies update their websites. This can be measured by an index that Dataprovider.com has developed, the Heartbeat indicator. So is it the case that the COVID-19 pandemic triggered certain firms to update and expand their websites? Or is it the case that companies that already have highly or frequently updated website were less affected by the pandemic and were, in this sense, more resilient to this shock? 

I believe that structured web data can be a very important data source for academic research because it has a very high coverage. When we look at company websites and almost every company has a company website nowadays, this would allow us to get an almost complete picture of the entire economy of a country or even worldwide. And it also allows us to have highly, frequently updated information about how companies are doing if they use their website as a communication tool to the outside world. So we have a very broad coverage that you cannot find in other data sources when it comes to including our companies and always in terms of highly, frequently updated information that we can find in this way." 

Prof. Dr. Michael D. König explains how he used web data from Dataprovider.com to research how companies were affected by the COVID-19 pandemic.

How were companies affected during the COVID-19 pandemic?

Andriana Boyrikova

Often these websites refer to each other as sources, thus forming networks. Combating misinformation has become a crucial part of dealing with the pandemic, and this is where technology can step in to uncover online networks that spread false and potentially harmful information.

Identifying websites that spread misinformation

We tend to refer to misinformation that is being spread online (and offline) as fake news. Yet, this encompasses not only fabrications and legitimate news that have been deliberately twisted, but also unresearched information and stories that contain some truth but aren't completely accurate. Inevitably, this impedes the identification of fake news and makes it ever more important to be able to read and analyze news critically.

Naturally, the most significant rule is to use common sense to evaluate if a message is plausible and collect information from multiple sources before drawing any conclusions. People tend to believe information that confirms their beliefs and discount information that contradicts what they support. However, keeping emotional responses in check and searching for what well-known and trusted sources have published on that subject is a crucial first step.

Next to that, checking the rest of the information on the website in question doesn’t take a lot of time and effort but plays a critical role in indicating how trustworthy it is: what other stories have been posted, is there any information on the mission, staff members or physical location of that organization or company?

When reading a piece of online content, it's also important to pay attention to the sources that have been used. For example, quotes, an interview, a report, survey data or official statistics and to check if these sources actually exist (as sometimes references to non-existent studies are made).

Uncovering networks of fake COVID-19 news

A significant number of these fake news websites are spin-offs of notorious websites publishing false health-related content and conspiracy theories. What’s more, they tend to link to each other as sources. So we can argue that if we find one website spreading misinformation about COVID-19, we can find more by, for example, following links on these websites. 

We will check if that holds true by doing an online investigation using our database of incoming links (links pointing to a particular website from other websites).

Our starting point is the infamous US website www.vaxxter.com. The homepage of vaxxter.com claims that the website publishes “scientific articles exposing vaccine myths and pharma foibles,” and states that it “promotes alternative health news.” However, it has published stories with misleading information and unsubstantiated claims about COVID-19. For example, an article from April 2020 headlined “Is 5G a Deadly Trigger for the Coronavirus?” suggests that 5G cell phone technology is related to the outbreaks of the new strain of coronavirus in Wuhan, Milan, and Iran. So far, there is no research that establishes a link between the COVID-19 pandemic and 5G. Additionally, the website doesn’t provide any ‘About’ information, physical location or contact details. 

Starting our search from Vaxxter.com, we collected 10 000 websites that link to it directly or through other websites. A cluster of 865 websites that are most likely related to fake news catches our attention as worth investigating. By applying a genetic clustering algorithm, we can determine that these 865 websites form a cluster, as they have a relatively large number of links among them but fewer links to other websites.

We pick one out of the 865 websites at random, Disinfo.news, and take a look at it. The website claims to publish “independent news on natural cures, food lab tests, cannabis medicine, science, robotics, drones, and more.” At the same time, the story in its header is headlined “Why Trump will win — compelling analysis no one else will report.” What’s more, an article from January 2020 headlined “Is coronavirus a manufactured bioweapon that Chinese spies stole from Canada?” claims that “these Chinese agents [...] may have infiltrated North America for the sole purpose of hijacking this deadly virus to unleash it at a later date.” So far, there is no evidence to support this claim. To top it all, the website doesn’t provide any information on its mission, staff members, contact details or physical location.

We can go one step further and dig deeper into that cluster of 865 websites by looking into keywords. After running an analysis, we came up with a keyword frequency plot based on the 33 most frequent keywords found on the websites. Among some of them are “vaccine,” “coronavirus,” “5g,” “children,” “truth,” “COVID-19,” “medic,”“science,” “billionaires,” and “freedom.” We can deduce that legitimate news or scientific articles can be writing about 5G, coronavirus or billionaires, but when they are used all together, this can be a red flag signaling fake news.

It’s interesting to note that some of these keywords - for example, “corona” and “5g” - also appear on websites that are outside the cluster but are somewhat connected to it; yet the further away we get from the cluster, the less frequently they show up.

The graph below identifies the high-risk websites in the cluster: the brighter colors signal a higher risk of COVID-19 misinformation based on the keyword frequency plot.

We select one of the websites from the highest risk group, www.unmaskohio.org, and take a look at it. (The website has been taken offline since)The homepage welcomes its visitors with the following statement “Unmask the truth: end compelled masks in Ohio,” and some of its featured headlines read “Masks ineffective” or “Most already immune.” When clicking on the section titled “The Science,” we find stories that claim that “Covid tests are not fit for purpose” and “Former Pfizer VP: 'No need for vaccines,’ ‘the pandemic is effectively over.’” What’s more, we can’t find any contact or ‘About’ information anywhere on the website.

If we compare the cluster that we are investigating to all 10,000 websites connected to Vaxxter.com, we can clearly see that the highest risk of fake news is still most prominent in the cluster where the websites link to Vaxxter.com directly or through other websites. The further away we get from the cluster, the lower the risk gets.

In an attempt to investigate the spread of online misinformation and fake news about the COVID-19 pandemic, we conducted research centered around Vaxxter.com: a US website notorious for failing to comply with basic standards of transparency and credibility. 

By using our database of incoming links, we could find a cluster of websites tightly connected to Vaxxter.com that publish inaccurate information or unsubstantiated claims about the COVID-19 pandemic. Our analysis indicates that websites that spread inaccurate information or conspiracy theories tend to link to each other, and technology can play an important role in discovering them.

As the COVID-19 pandemic has taken over the world, so has misinformation. The pandemic has led to an overabundance of inaccurate information and conspiracy theories: from sources claiming they offer treatments and miracle cures to outlets deliberately spreading false information to sway public opinion.

Follow us

Questions?

COVID-19

E-commerce: Pandemic winners and losers

How were companies affected during the COVID-19 pandemic?

Uncovering networks of websites spreading misinformation about COVID-19

Start today