E-commerce live streaming has become a massive retail channel, blending entertainment, social interaction, and instant shopping into a single digital experience. On platforms like TikTok, viewers watch hosts demonstrate products, ask questions in real time, tap the screen to leave “likes,” and click to buy without ever leaving the broadcast. For businesses, encouraging viewers to interact is a standard marketing goal. The basic assumption is that more engagement leads to more revenue.
A team of researchers set out to investigate exactly how this process unfolds over time. Instead of looking at a broadcast’s total interactions and total sales at the end of the day, Zhang Bolun and his colleagues at Harbin University of Commerce and Harbin Institute of Technology tracked the minute-by-minute flow of live streams. They wanted to see if the relationship between engagement and sales is a two-way street, where interactions drive sales and sales simultaneously drive further interaction. Their findings were published in Marketing Intelligence & Planning.
Tracking the Data
Prior research on live-stream shopping has generally treated the link between engagement and sales as a static, one-way event. The researchers argue that this perspective misses the fluid nature of a live broadcast. A stream is a constantly changing environment characterized by alternating periods of high and low activity. To understand how these fluctuations impact a retailer’s bottom line, the team needed granular data.
The researchers utilized a web crawler to gather information from TikTok, focusing on the platform’s top 100 live-streaming channels between August and September 2023. They captured activity during the peak evening shopping hours of 7:00 PM to 10:00 PM. Extracting some of this data required manual processing. Because total online viewer counts were not available as raw text, the team downloaded the videos, extracted screen images at one-minute intervals, and used text-recognition software to log the visible viewer numbers.
After cleaning and organizing the data, the final set included 2,876 individual live streams, representing nearly seven million minute-by-minute data points. For each minute, the researchers tracked four specific variables: the number of likes, the number of comments, the total online viewers, and the gross merchandise value, which represents total sales revenue.
In a live digital environment, multiple actions happen simultaneously, making it difficult to separate cause and effect. To address this, the researchers applied a statistical model designed to track how multiple variables influence each other over time. This approach allowed them to isolate how a spike in one specific activity predicted changes in the other variables during the minutes that followed.
A Two-Way Street
The analysis revealed that consumer engagement and sales revenue feed off one another. A surge in engagement metrics is linked to an increase in sales. At the same time, a surge in sales predicts a subsequent rise in user engagement.
The researchers attribute this mutual influence to the public nature of the live-streaming environment. When viewers buy a product, notifications or host announcements often make those purchases visible to the rest of the audience. The researchers interpret this as a signal to other viewers, prompting them to either join the conversation or make their own purchases, which keeps the momentum going.
The study also broke down engagement by the level of effort required from the user. Tapping a screen to send a “like” is a low-effort action that indicates basic approval or interest. Typing a “comment” requires more time and thought, often involving questions, opinions, or reactions to the host.
These two types of behavior impact sales differently. The data showed that a spike in likes provides a fast but short-lived boost to sales. A spike in comments, on the other hand, produces a slower but significantly larger and longer-lasting increase in sales. The analysis also showed that the different types of engagement influence each other. High-effort interactions tend to stimulate low-effort interactions. Specifically, a wave of comments predicts a strong and persistent increase in the number of likes, while likes have a much smaller effect on generating new comments.
The Role of Viewers and Hosts
Interestingly, the sheer volume of people watching a stream does not directly lead to higher sales. The statistical model indicated that a rise in online viewers has no immediate, direct impact on revenue or on the number of likes. Instead, higher viewer counts increase the pool of people who might leave a comment. Those comments then go on to drive both sales and additional likes.
The researchers also tested how the specific characteristics of a live stream affect the financial return on user engagement. They categorized the broadcasts based on the type of host, the size of the channel’s fan base, and the overall popularity of the store. Streams led by influencer hosts, rather than standard corporate representatives, saw a stronger link between engagement and sales. Similarly, having a larger pre-existing fan base and higher store popularity amplified the translation of likes and comments into revenue.
The Five-Minute Window
For retailers and e-commerce managers, the study outlines a specific timeline for how interactions translate into revenue. The cumulative impact of an engagement spike on sales is concentrated almost entirely within the first five minutes. After ten minutes, the effect becomes negligible.
The researchers point out that merchants can use this predictable timeline to optimize their sales strategies in real time. Instead of following a rigid, pre-planned script, hosts can watch their live metrics and adjust their pitch on the fly. For instance, the data suggests that introducing a new product or making a primary sales push within two to four minutes of a peak in user comments is the optimal window to capture the resulting wave of purchases.
The study does note a few limitations to the findings. The data was collected from an aggregate perspective, meaning the researchers tracked the total volume of activity per minute but could not track individual users to see if the specific people commenting were the exact same people making the purchases. Additionally, the data is drawn entirely from a Chinese platform, meaning the exact timelines and interaction patterns might differ in other regional markets or on different streaming apps.
