Most first-time founders describe their customer as "small businesses in India" or "any startup that needs what we sell." That is a market, not a customer. An Ideal Customer Profile (ICP) is a sharp description of the specific type of organisation that gets the most value from your product, can actually pay for it, and can be reached efficiently. Getting this right early is the highest-leverage decision in go-to-market, because it decides who you build for, who you sell to, and who you deliberately ignore.
ICP is not your market, and not a persona
Keep three ideas separate. Your total addressable market is the full universe of companies that could ever buy. A buyer persona is the individual human you sell to, such as a finance head or a factory owner. Your ICP sits between them: it is the profile of the company most likely to buy fast, succeed with your product, and stay. For a B2B startup, define the ICP (the account) first, then the personas inside it.
Firmographics: the "who"
Firmographics are the observable, filterable attributes of a company. In the Indian context, the most useful and verifiable signals include:
- Size band: Under the revised MSME classification effective 1 April 2025, a Micro enterprise has turnover up to Rs 10 crore, a Small enterprise up to Rs 100 crore, and a Medium enterprise up to Rs 500 crore. Which band you serve changes your pricing, sales motion, and support model.
- Formalisation: Whether a business is GST registered is a strong filter. GST registration is generally mandatory once turnover crosses Rs 40 lakh for suppliers of goods or Rs 20 lakh for services in most states. A GST number signals a business mature enough to buy software and to want proper input-tax-credit invoices.
- Entity type: A Private Limited company or LLP, each carrying a registration number on the MCA registry, behaves very differently from a sole proprietorship in budget, decision speed, and compliance needs.
- Sector, geography, and stack: Which industry, which cities or states, and which tools they already run.
Behaviours: the "how they act"
Firmographics tell you who a company is. Behaviours tell you whether it is ready to buy now. Look for:
- Trigger events: a new funding round, a compliance deadline, a GST or TDS filing crunch, rapid hiring, or expansion into a new state.
- Current workaround: the spreadsheet, agency, or manual process they rely on today, which reveals real pain and real willingness to switch.
- Buying behaviour: who signs off, how quickly they decide, and whether they buy self-serve or need a demo and a formal invoice.
The best ICPs combine both layers: a firmographic filter plus a behavioural trigger. "Small manufacturers in Pune who just crossed the GST threshold and still track invoices in Excel" is far more actionable than "SMEs in India."
Why a narrow beachhead wins early
A beachhead is one small, homogeneous segment you dominate before expanding, an idea popularised in Geoffrey Moore's book Crossing the Chasm. Narrow beats broad early for concrete reasons:
- Sharper product: one segment shares one set of problems, so you build depth instead of shallow features for everyone.
- Cheaper distribution: a tight segment shares the same WhatsApp groups, trade bodies, events, and language, so word of mouth compounds.
- Believable references: ten happy customers who look alike sell the eleventh far better than fifty scattered logos.
- Faster learning: concentrated feedback tells you clearly whether you are approaching product-market fit.
India is not one market; it is many. A beachhead lets you win a specific slice, for example D2C brands in one city or clinics of a single specialty, before you generalise.
Write your ICP this week
Draft one or two sentences covering industry, size band, location, entity type, the trigger that makes them buy, and the behaviour that shows they are stuck today. Then pressure-test it against your last ten sales conversations. The prospects who converted fastest and used the product most are your real ICP, not the ones you wish you had. Revisit it every quarter, and let data steadily replace assumption.