Site icon Marketing Thought

Getting Beyond Text In Unstructured Data

Unstructured data has massive potential in marketing that is already somewhat being fulfilled. Text analysis has proved a fruitful resource. Using text mining, marketers can seek to uncover messages hidden in massive amounts of words. A typical usage would be attempting to uncover the themes in online reviews written about a product. Yet, unstructured data is any data that lacks the formal structure of a database. This means pretty much anything not arranged in the neat formats found in a firm recordkeeping system. This gives an idea of the potential and the challenge. Think of all the messy data there exists in the world that could potentially be useful to marketers. What can marketers do to use this data? How can they get beyond text in their assessments of unstructured data?

Get Beyond Text To Image, Audio, And Video

We recently published a paper on how to get beyond text in marketing analysis of unstructured data. This looks at the platforms available to do such analysis. The platforms we investigated were all from major tech players everyone would recognize — Google, Amazon, and Microsoft — so they are available to pretty much anyone with internet access. What is more, these platforms are relatively easy to use being designed to encourage even the not-so-technically gifted to employ them. Plus, they allow for free usage to varying extents. The commercial theory behind this is for the firms to allow all to do a little analysis using the big players’ AI tools and then charge those who get hooked and become heavy users. (Freemium type model). This is the same business model as many drug dealers, the big difference being that using these platforms will likely help your career (and may be less socially destructive?)

We considered three types of unstructured data.

Why Would You Ever Use Beyond Text Data?

The three types of data are widespread and can house interesting insights. Image data is all over many social media sites. Using the platform AIs can uncover the content of images. For example, is there a product in the image? Looking at the backgrounds in which products are shown can give insight into product usage. They can also pick out different objects in an image. What products tend to be used together?

Video data can be thought of in a couple of different ways. Video is, basically, just a lot of images put together. As such, many of the same image tools can be used. Trying to understand video you can measure changes between snapshots, i.e., how specific images at different points in time deviate from each other. You can also analyze the individual images. Again you can look for products and product usage. Video is becoming increasingly common on social media. As devices get more sophisticated and storage cheaper it is easy to imagine more and more video becoming available to the marketer.

Audio data can be available in call centers — “these calls may be recorded for training purposes”. There is also plenty of audio data on social media, not least in videos. Some of our prior research covered measuring vocal tones, e.g., the way things are said. This current paper largely concentrates on transcription services. How the AI can help by detecting the words that are said. Marketers (and others) can now access huge troves of audio data turning that into text data which can be analyzed using already well-known text analytics techniques.

Some Types of Unstructured Data To Use When Getting Beyond Text

Unstructured Marketing Data

With the ability to investigate a vast range of unstructured data marketers can investigate so much more. Indeed, the challenge could increasingly become knowing what data not to look at.

The world has been turned upside down. Now marketers have access to more data than could possibly be used in multiple lifetimes, with more shared every minute. The abundance of unstructured data beyond text, and the increasingly convenient analysis tools, have offered an opportunity for managers to monitor, structure, and enhance their offerings but how to do so?

Wang, Bendle, and Pan (2024)

If you want to try a bit of text mining this is a really useful and free resource.

For more on unstructured data see here, here, and here.

Read: Shane (Xin) Wang, Neil Bendle, and Yinjie Pan (2024) Beyond text: Marketing strategy in a world turned upside down, Journal of the Academy of Marketing Science. This is an open-access paper so it is free to read

Exit mobile version