If you’ve ever tried validating an Ideal Customer Profile (ICP) with a huge stack of interviews, you know how exhausting it can be. Scheduling, recording, transcribing, analyzing… and by the time you reach number 40, you're not sure if you're getting new insights or just more noise.
In marketing today, ICP validation is no longer optional. Without it, campaigns get vague, offers miss the mark, and sales calls feel like cold outreach. But do you really need to talk to 50 people just to know who you're selling to?
Let’s look at why interviews still matter, and how to make them work faster, smarter, and with fewer conversations.
Why conduct interviews at all when you can just collect analytics?
Marketers love numbers. Behavioral dashboards, heatmaps, bounce rates, CLTV, they all look objective and convincing. But here's the problem: these metrics show what the user does, but almost never why they do it.
If you want to understand how to craft the right message, choose the right channel, and predict pre-click behavior, you don't just need numbers. You need context. And the fastest way to get it is through conversation.
Here's what you can learn in a 30-minute interview that no analytics can:
- What real words does the client use to describe their problem
- What actually triggered the search for a solution (it's almost never what you think)
- What alternatives they considered, and why they didn't work
- What the selection process looks like inside the company
- What criteria determine the final "yes" or "no"
These are not just details. This is the core of the ICP, on which everything depends: positioning, creatives, funnel, email campaigns, and even sales enablement.
If you build marketing without this knowledge, you build on guesswork.
As Jennifer Havice, author of Finding the Right Message, said:
“The words customers use are not just ideas. They are the structure of all communication. If you don’t know them, you are talking past them.”
And yet, 50 interviews are a pain. Especially if you have limited access to users or you work in a narrow niche. In B2B startups with a long deal cycle, collecting 50 relevant contacts is already a challenge. You also need to persuade them, conduct them, decipher, and analyze...
That's why today more and more teams are looking for a way to reduce the number of interviews - without losing depth. One way to do this is to first build an ICP hypothesis using data from open sources and generative tools like Elsa AI, and only then validate it on a smaller volume.
How this works and why it really saves time - we will analyze in the next block.
Why 25 interviews can yield more than 50, especially if you start with a hypothesis rather than from scratch
To be honest, validating an ICP without any preparation is still a reality for many teams. And it’s expensive.
When you start with a clean slate, interviews quickly turn into an endless race for patterns. You catch individual phrases, try to put together a mosaic, but in the end, everything looks disjointed. Fifty conversations, and you still don’t know who you really want to attract.
Sound familiar?
Now imagine another situation. You have a working hypothesis. Not perfect, but clear enough. You understand what a potential client looks like, what pain points they have, what prevents them from making a decision, and what triggers they respond to. Interviews stop being a search by touch, they turn into validation of specific blocks.
This is where Elsa AI really helps.
You set the basic input: industry, segment, client type, and role. The system compiles a draft ICP, including:
- audience goals and motivations;
- problems they are trying to solve;
- barriers that prevent them from taking action;
- decision-making criteria;
- content consumption channels.
This works as a starting point. You get a primary map, with which you can already go into the field. And yes, this allows you to reduce the number of interviews without losing depth.
In practice, it looks like this:
- Without a hypothesis:
You talk to 50 people, asking open-ended questions. Lots of repetition. Little structure. Lots of guesswork.
- With a hypothesis from Elsa AI:
You focus on key assumptions. Ask targeted questions. After 10-15 interviews, you have a verified ICP, collected on live data.
According to a recent survey among B2B marketers, more than 60% of those who started with a hypothesis were able to capture the main elements of the ICP after 20 interviews. For those who went without a structure, this figure shifted to 40-50.
Important: a hypothesis is not a dogma. It can be wrong. But it is the presence of a structure that allows you to quickly see this and adjust it in the process.
Personally, I prefer this approach: first, I put together a rough ICP in Elsa, then I put together a block of questions for each element (goals, problems, barriers, alternatives), and only then went to the interview. This saves not only time, but also energy - you know why you are communicating with each person.
If you work in a startup or a B2B product, where speed is more important than perfection, this path will definitely suit you.
How to build a working ICP hypothesis with Elsa AI, even if you are not a researcher
If you think that this requires experience in Custody, long forms or in-depth interviews with transcripts, relax. The system itself does 70% of the work for you.
It all starts with a basic description: who is your client, what segment are they in, what their company does, and what roles make decisions. The more clearly you describe the input, the more accurate the hypothesis will be.
Elsa does not just generate an ICP card, it collects a logically connected profile in which everything is interconnected. This is not a list of characteristics, but a decision-making map:
- What does a person want to achieve? (Goals)
- What problems are getting in the way? (Problems & Pain points)
- What can push them to action? (Triggers)
- What barriers can stop them? (Barriers)
- What has he already tried? (Alternatives)
- What does he look at when choosing? (Buying criteria)
Each block is not just described, it is provided with explanations of where the information came from. This can be an analysis of forums, social networks, public reviews, and interviews. You immediately have context.
What is especially convenient is that you can click on each element and get a list of questions to ask during the interview to confirm (or refute) the hypothesis. For example:
- Pain point: “Purchases take too long to be approved due to the uncertainty of metrics.”
→ Question: “Tell us what the process of approving a new tool usually looks like? What slows down the process?”
- Trigger: “A new team member has appeared who needs results quickly.”
→ Question: “Have there been situations when changes in the team forced you to look for new solutions?”
This structure makes interviews fast and focused. You don’t go off on a tangent, don’t catch abstract answers, and don’t waste time “just talking.” Everything is clear: you check block by block.
If a person confirms the hypothesis, great, you increase confidence. If not, you edit the model and continue. Flexibility in real time.
We at M1-Project have often observed how teams save 20–30 hours on preparing for custom development after generating an ICP in Elsa. They go to interviews with a ready-made structure and spend 2 times less time on analysis, because the structure already exists.
And this is only the first version. After each update, the model becomes more accurate: the more you verify, the more workable the profile becomes.
How to choose 25 respondents who will save you 25 interviews
Not every user will give you valuable data. Not every client can explain exactly why they bought. But if you choose the respondents correctly, even 10 interviews will give more depth than 50 random ones.
First, prioritization. At M1-Project, we recommend starting not with a cold audience, but with those who have already interacted with the product at different stages. Why? Because they have experience - and awareness. Here are three groups to focus on:
- Lost deals.
Customers who reached the final stage but did not buy. They will tell you about barriers, alternatives, and the weak points of your offer. The main thing is not to sell again. Just listen.
- New customers (up to 30 days).
They still have fresh memories of the selection process. They remember what prompted them to buy. These are ideal respondents to confirm decision triggers.
- Loyal customers (6+ months).
They have already received results. They have a language that can be used to write sales texts. These interviews often yield unexpected formulations of pain and motivation.
Instead of taking 50 people “just in case,” you simply select 5–8 from each group. This gives you breadth, depth, and contrast. And it also leads faster to insights that affect the funnel.
Here is a simple framework for selecting the right people:
- Stage: Where were they in the funnel? (lead, trial, demo, rejection, upsell)
- Intent: How motivated were they? (for example, did they sign up themselves or through advertising)
- Outcome: How did the interaction end? (purchase, leave, postpone)
The more diversity in these three criteria, the better. This gives you a complete picture: what works, where it breaks down, and why.
By the way, Elsa AI can automatically suggest a list of questions and frames for each respondent type. For example, for lost deals, focus on barriers and alternatives, for new clients, triggers and selection criteria.
If you combine the right sample with the logic of hypotheses, you get 3 times more signal and 5 times less noise. This is no longer research for the sake of a report. This is a real lever of influence on the product and marketing.
What to do with the data after the interview: turning insights into strategy
Okay, the interviews are done. The recordings are on disk. Someone even took notes. And this is where most teams start to freeze: “What next?”
The reality is that just “talking” is not enough. The interview is only half the journey. The other half is structuring the information and integrating it into marketing.
So, first of all, we clear out the chaos.
You can’t use 10 interviews if they are scattered across files, without structure, without clear conclusions. So step 1 is to put everything on the shelves.
Here’s how teams that know how to turn clients’ words into strategy do it:
We take the “as is” wording. People often say ready-made headlines and messages themselves. Phrases like “I just need something that won’t break in 3 months” are not “a quote on a sticker,” but real text for a landing page.
Mapping responses to the ICP structure.
Goals, pain points, triggers, barriers, alternatives, selection criteria — if you mark these blocks during the interview, you get not just a transcript, but a client decision-making map.
Looking for repetitions. If 7 out of 10 people say: “we are afraid to waste the budget in vain” — this is no longer a personal opinion. This is a hypothesis for advertising and a page with trust markers.
Identifying moments where emotions went off the scale. Where a person got irritated, sighed, laughed — there are real “keys” to motivation. This is not just “inconvenient” — it is “it drives me crazy every time I log into my personal account and don’t understand where the reports are.”
After this, you no longer look at the interview as “soft research.” You have structured data that can be directly invested in the strategy.
Do you want to understand what pain points to emphasize in the first section of the landing page? See what is mentioned most often.
Don't know what lead magnet format to make? Check what tasks people complain about.
Planning a mailing? Use customer phrases - they sound truthful because they are real.
And if you want to automate this step, Elsa AI is already designed for this.
Instead of manually extracting insights, you simply upload interviews and get a full portrait of the ICP - from goals to preferred channels. Plus: you are immediately offered hypotheses and wording for advertising, email, and landing pages. Everything in context, with minimal risk of making a mistake in tone or message.
This is how interviews turn into a strategy, and marketing into a predictable system, not a lottery.
Conclusion
If marketing research used to be associated with a long list of interviews and weeks of transcripts, today everything is changing. With tools like Elsa AI, you can act precisely: start with hypotheses, not from scratch, reduce the number of conversations without losing depth, and quickly translate insights into real marketing actions.
The key is not quantity, but structure. You can talk to 10 people and build a strategy that hits the mark. Or conduct 50 interviews and still remain guessing. It all depends on how clearly you know what you are looking for and how well you can read between the lines.
Marketing is no longer about intuition. It is about precise signals gleaned from live conversations - and turned into positioning, landing pages, content, and distribution channels. The sooner you integrate the ICP framework into your work, the faster you will get real results from each dialogue.
No magic. Just structure, attention to detail - and the right questions.
.svg)
