How to run a 3C analysis with AI, separating facts from hypotheses

If you ask ChatGPT or Gemini to "do a 3C analysis of this company," the customer, competitor, and company table fills in right away. A filled table and something you can use for a decision are different things.

The AI has not met your customers. It does not know a competitor's internal situation either. If you have it analyze without handing over materials, it sometimes fills the blanks with a generic picture of the industry, or with a plausible persona.

When you do a 3C analysis with AI, what matters more than how well the sentences are written is a mechanism that separates facts you could confirm, the AI's hypotheses, and what you still do not know. This article introduces a procedure and a prompt you can try in free ChatGPT or Gemini.

What you want to decide at the end of a 3C analysis is the overlap

A 3C analysis is a way to sort the state of a business or its marketing from 3 parts: Customer (market and customer), Competitor (competitors), and Company (your own company). Salesforce's explanation and Canva's explanation also introduce the basic form that uses these 3 elements.

Filling the 3 columns only ends as research notes. What you need in practice is to put the following overlap into 1 sentence.

What the customer is looking for
× What competitors are not satisfying enough
× What your own company can carry out
= The growth opportunity to test this time

To make this 1 sentence, you gather the materials to pass to the AI.

The materials you pass to the AI differ in kind across the 3

Before you let the AI imagine a customer, pass words that were actually used. For competitors, limit it to what you can confirm on the official site. Do not let it write weaknesses that cannot be known from public information, such as "support is weak" or "customer satisfaction is low."

For your own company too, pass what you can provide now, not the ideal. If you treat something you have no results for as a strength, the moves after that drift away from reality as well.

Put 3 labels on every statement

Have the AI's output attach the following labels.

• Fact: it can be confirmed in the materials you passed, or at a URL you specified
• Hypothesis: it can be considered from the facts, but it has not been confirmed yet
• Missing information: there is no material for a judgment

For a "fact," also have it write which input it used as the basis. If it says "competitor A puts weight on beginners," have it also write which expression on the pricing page or the top page it judged from.

You do not have to delete hypotheses. Once you know it is a hypothesis, you can turn it into "a question for the customer," "an appeal to try small in an ad," or "a page to look at next." What is dangerous is a hypothesis lined up with the same appearance as a fact.

A 3C analysis prompt you can use in ChatGPT or Gemini

Paste the information you have into the next prompt and use it. In an environment that cannot read URLs, paste the page text or the main points along with it.

You are an analyst who assists marketing research.
Using only the information below, create a 3C analysis.

Purpose:
(Example: next month, decide the 1 acquisition move to test first)

Information about your company:
- Product or service:
- Price:
- What you can handle:
- What you cannot handle:
- Your site, or the page text:

Information you have confirmed about customers:
- Words used in inquiries and sales conversations:
- Reviews and frequently asked questions:
- Search terms:
- Reasons a purchase or an application was completed:
- Reasons it was declined:

Competitors' public information:
- Competitor name and URL:
- What is written as the target customer:
- Product, price, and appeal:
- The path up to the application:

Rules:
- Do not add facts that are not in the input
- Attach a label of "fact," "hypothesis," or "missing information" to each statement
- For a "fact," write the input or URL that was the basis
- Do not guess a competitor's weaknesses, revenue, or customer satisfaction when they are not in the public information
- Do not invent customer attributes such as age or gender without a basis
- Do not fill an empty field with a generality

Output:
1. Customer: the trouble, the criteria for comparison, the reasons they hesitate
2. Competitor: target customer, appeal, price, path
3. Company: value provided, strengths, constraints, supporting evidence
4. A growth opportunity that can be considered from the overlap of the 3
5. Information to confirm next, in order to check the opportunity
6. 1 small test you can run this week

After the output, check the evidence first

Before you reread the AI's text from the top, check the count and the evidence for each label.

1. Does each "fact" have a basis in an input or a URL
2. Is a guess about internal circumstances mixed into the evaluation of competitors
3. Is the customer picture built from age or gender alone
4. Does the growth opportunity connect to all 3 of customer, competitor, and your company
5. Has the missing information turned into a question you can confirm next

If the evidence is only on the customer and your own company, you cannot yet decide a line that can win and that includes competitors. In that case the conclusion is "confirm the competitor information." An analysis is more useful when the next thing to look up is decided than when you force it all the way to a move.

Gathering the materials and URLs every time is a lot of work

You can try this method with a free AI as well, but every time you analyze you have to gather the business information, the customer's words, the competitor URLs, and the measured data again. If the previous hypotheses, and the moves you did not adopt, are also in another place, you repeat the same research.

Shoryaku AI sorts the market, the customer, and your own company, using the business information you registered and the data it was able to fetch as the material. In the "where you can win" chapter of the analysis result, it shows the growth opportunity put into 1 sentence, the material about your company that could be confirmed, and what could not be confirmed yet. It does not fill an item that has no material.

For competitors, it used to have a feature that had the AI read official sites and compare them automatically, but it was removed because no comparison was completed in production. That was so the competitor field would not be filled by guessing. The background is in Why automatic competitor-site comparison was removed. Checking competitors is more reliable if you do it 1 company at a time with the prompt in this article.

When you move from a 3C analysis to a move, How to ask AI for revenue ideas, and how Shoryaku AI does it is also useful. If you are explaining the analysis result to a customer or a manager, you can sort the structure in Seven ways to make a marketing deck easier to read with AI.

Summary

When you do a 3C analysis with AI, do not make filling the table the goal.

1. Pass different materials for the customer, the competitor, and your company
2. Separate everything into fact, hypothesis, and missing information
3. Attach an input or a URL as the basis of a fact
4. Make the growth opportunity from the overlap of the 3 into 1 sentence
5. Turn the missing information into a question to confirm next

At first, try passing only 3 pieces of customer language you have on hand, and 1 official competitor URL. A prompt that can leave missing information missing leads to the next piece of research and the next decision.

• Marketing
• AI
• Product