What do we measure?
The webAI Delivery Delay Agent has been trained by ISTARI.AI to infer from companies’ websites their affectedness by delivery bottlenecks and delays. The agent also identifies non-deliverable products and products with very long delivery times. Companies affected by exceptional supply chain problems communicate this fact frequently. Based on this information, an individual Delivery Delay Intensity Score is formed for each company. This numerical indicator reflects how centrally the issue of delivery delays is communicated on the respective company website and how frequently products that are not available or only available for a very long time are listed.
The more acute this topic is for the company, the more significant it is for the company’s external communication. For example, a company whose entire product range can only be offered on a limited basis due to delays in global supply chains communicates very centrally on this topic. Also, many of the products offered by the company would currently be unavailable for order or have extremely long delivery times.
How do we measure?
Our webAI reads the website of the company under study and searches it for text sections (paragraphs) that deal with the topic of delivery. To do this, webAI first looks for keywords potentially related to delivery, then analyzes the paragraphs thus identified, determining whether delivery problems are communicated or whether the products are currently unavailable. If webAI has assigned a corresponding paragraph with a high probability to the topic of delivery problems, webAI remembers this paragraph and continues searching. Thus webAI searches the entire company website, or the particularly relevant “top-level” sub-webpages if it is a very extensive website with hundreds of sub-webpages (more information on this: https://doi.org/10.1007/s11192-020-03726-9).
WebAI thus finds a certain number of paragraphs per company website that deal with delivery problems and products that are currently not available. WebAI then relates this number to the total amount of text content read on the website. This is how webAI determines a Delivery Delay Intensity for the company.
How do you interpret the data?
The Delivery Delay Intensity calculated in this way would be 0.0 for a company with no communication on the topic and no products affected. In contrast, for the example described above of a company whose global supply chains are disrupted, the value could be 1.36. In contrast, for an e-commerce company with thousands of products on offer, only a normal number of which are unavailable, the value could be 0.15.
So, unlike simpler binary classifications (“delivery problem YES/NO”), webAI outputs a continuous score at the company level. This makes it possible to distinguish between companies for which delivery problems are only a marginal issue and those that are acutely affected. Users of webAI data can thus easily determine for themselves how severely a company must be affected by supply bottlenecks for it to be relevant to them.
As an example for the DACH region (Germany, Austria, Switzerland), it can be said that 40.4% of the companies examined have a Delivery Delay Intensity Score of greater than 0.0 and thus communicate the topic of delivery delays in some form on their websites or offer products that cannot be delivered. The average Intensity Score of these companies is 0.39, with half of the companies having a score greater than 0.22. The maximum score achieved is 6.60, with 8.45% of companies achieving a score of 1.0 or higher.
How do we ensure the quality of the data?
Like all our webAI agents, the webAI Delivery Delay Agent has been developed and validated together with independent subject matter experts. This ensures that we train the webAI agent with real expert knowledge and that it then generates expert-level results. For this agent ISTARI.AI collaborated with researchers at the Department of Geoinformatics – Z_GIS at the University of Salzburg. Z_GIS is an established competence center for GIScience, which is active in research and teaching in collaboration with academic and industrial partners.
Interested in our data?
Purchase this and other datasets for hundreds of regions in our Data Market.