How to Use AI for Market Research
Along with a helpful dose of math, python, data analytics, and some good old human decision making. We can't trust the AI too much, can we?
Market research can be both the most useful and the most limiting aspect of a young entrepreneur’s journey. Businesses built around market research charge large amounts of fees, sometimes upwards of 15-25k USD, an amount that is out of reach for many founders. On the other hand, if you work within Corporate America, much of the market research can be outsourced. The work is cheerily handed off to a company overseas or to a dedicated team within the company that would get it done for you.
And so, when the time came for me to do market research for a client with a near-zero budget and with no support teams in place, I had to get creative.
Fundamentally, market research can vary from industry to industry, from product to product and even changes depending on the person leading the research. To make sure that I would stay on track, I started with defining specific objectives that I wanted out of the research. What are the problems I needed to solve? Which were the questions my client needed to be answered?
Next, to make sure that I didn’t miss anything, I googled “What comprises a comprehensive market research for a new/overlooked industry?” I was hit with pay-walls and half-answers with promises of more (on payment ofc). All I wanted was a checklist, but unfortunately, even that is not easily accessible. It was frustrating to experience the lack of knowledge sharing and the commercialization of simple business techniques.
On a whim, I turned to ChatGPT. This is how our conversation on Market research started:
Here’s a shortened version of the rest of the answer if you’re interested:
1. Define Objectives and Goals
2. Identify Target Audience
3. Competitor Analysis
4. Market Segmentation
5. Trends and Technology Assessment
6. Survey and Interviews
7. Partnerships and Collaborations
8. Financial Assessment
9. Legal and Regulatory Compliance
10. Customer Feedback and Reviews
11. Industry Events and Conferences
12. Data Analysis and Synthesis
How great is that response??
1-6, 8-10, and 12 were all common steps. BUT I was impressed to see ChatGPT actually add in a “Collab” section which is highly appropriate for an individually fragmented industry like Digital Art. Besides that, “Industry Events and Conferences” was a good heads-up for me, as I had frankly assumed that this industry was not organized enough to conduct such events. To suit my consultancy style, I did add in a “Sustainability Analysis” section in which I explored ways to make a digital art business more circular and resource-friendly.
I.
Now that I had an outline, I started on secondary research, which basically meant snooping into any research / data others have previously compiled. I generally do my secondary research in three ways: looking up government studies and reports, previous market research done by consultancies on the same industry or a similar product, and insights from Industry Association or other community-specific reports.
What I found: ChatGPT’s summarization feature can help save tons of time here and be a valuable assistant, especially when you have no human ones.
How: I created a custom python script that would call on the ChatGPT API and use it for PDF summarization / web page analysis. Using that tool provided me with a quick look into existing research and helped me gather insights on where the industry is currently at. But not to worry, I didn’t fully trust my little AI assistant. I verified the docs myself and made sure that the outputs from GPT were free from hallucinations. On the plus side, the tool did save me time that would have otherwise been spent writing and summarizing those docs myself.
II.
Automated Competitive Analysis can be done using a blend of web-scraping + AI Analysis. One can feed in the competitor’s store URLs, personal portfolio websites, etc. and get back a response. I didn’t use this tool for every page though as I felt that GPT was missing out on some nuances. For example, some of the digital artists I had staked out as the relevant competitors come from affluent backgrounds which was easy for me to glean as a human (by going through their social media pages), but not so easy for an AI. However, the AI did give me a broad understanding of who a competitor is and what they are offering.
Pricing Analysis: I was able to experiment using similar tools, with some data analysis and statistical functions thrown in. For certain competitors, I did manually go through their product descriptions and gathered info on prices, shipping, and materials. Again, I wanted to make sure that I didn’t miss anything, and I know I should probably work on improving my model, but I enjoyed figuring out this piece of the puzzle by myself. AI helped with finding the mean/median prices, but it could not use intuition/secondary analysis to guess accurately at the reasoning behind the prices.
Sustainability Analysis: Now, this was a section I was very interested in. And AI showed how helpful it can be to build more People and Planet Friendly Businesses.
When I asked ChatGPT for ways to make an online stationery business more circular, it provided a list of reasonable ideas. Even though not all of them were feasible for a business that is just getting off the ground, the ideas were reasonable enough to keep stored for later.
IV.
What about Primary research you ask?
Primary Research involves talking directly to potential customers and analyzing lots and lots of data. Intuitively, using AI for sifting through the data makes sense. I had created a short survey to determine interest in Digital Art Products amongst Indians and collected the responses in a CSV file. I had less than 50 responses so I used basic math, Pandas and matplotlib (visualization) to get the information I needed out of the data. BUT I was curious to see if it's possible to use ChatGPT here… and yes, you can feed it a dataset and ask questions based on that.
V.
Grammar and paragraph checker:
The final use I got out of my Gen AI Assistant was to verify the document I had created and highlight any grammatical / spelling errors I may have inadvertently made. This is a core feature of Chatgpt and it proved to be extremely useful.
In Conclusion,
The general hype surrounding Gen AI and how its going to take over all our jobs appeared quite overblown to me after my experience tinkering around with it. ChatGPT is a tool and must be used in addition to the work done by a human. It should be treated as an assistant, and not a replacement. It can help when you don’t have the money or resources to get another human being to help you out, or it can help make your existing workers more productive. (We can do 4-day workweeks, people!) I also found that we shouldn’t fully rely on its accuracy. It’s important to verify its reasoning and outputs, especially if those answers could have a direct effect on human lives.
Thanks for reading this post! I would love to hear how you have used Generative AI to make your work life easier! Please comment and let me know your thoughts.