How to create my first ICP

An ICP (Ideal Customer Profile) is the description of who you are trying to reach. It is the yardstick the Leads module uses to score everyone who engages with your posts. Creating one takes a three-step wizard.

Step 1: Describe your ideal customer

Give the ICP a name and a description in plain language. The description is the only part the AI reads when it assigns a fit level, so this is where the accuracy comes from.

Be concrete about the role, the type of company and the context.

A good description:

> Head of Sales or VP Sales at a B2B SaaS company, mid-market, based in Europe or North America, with hiring or growth signals.

What to avoid:

• Too vague. "People interested in sales" tells the AI nothing.

• Contradictory. Asking for startup founders *and* enterprise CTOs in one ICP. Create two.

• Overstuffed. Keep it to the profile. Countries and company sizes belong in the filters at step 3.

MagicPost also suggests a few ICPs based on your own LinkedIn profile and your recent posts. Click one and it fills in the name and description for you, ready to edit.

Step 2: Add example profiles

Paste the LinkedIn URLs of up to 5 people who perfectly match the profile you just described. The AI uses them to calibrate.

This step is optional, but it is the single biggest lever on scoring accuracy. Use real customers, prospects you wish you had more of, or the best accounts of your competitors.

Step 3: Set your filters

Filters run before the AI scoring, they are free, and they save credits.

Always visible:

• Country: include or exclude.

• Industry: include or exclude.

• Company size: by employee bracket.

Under Advanced filters:

• City: include or exclude.

• LinkedIn profile type: Creator, Premium, Open to work.

• Follower range: a minimum and a maximum.

They do not all bite the same way, and it matters when you set them.

• Hard rejects, applied on the spot, no credit: any exclusion (country, industry, company size, city), the country include-list, the follower range, and the LinkedIn profile type.

• Softer: the industry, company size and city include-lists. A profile that misses them is not disqualified, it goes to a free borderline check that decides whether it deserves a scoring. LinkedIn's industry and city values are free text ("Greater Nantes Metropolitan Area"), too unreliable to reject someone on.

• A filter is skipped when the profile has nothing in that field. Someone whose profile lists no country will not be cut by your country filter.

Nobody costs you a credit at this step. How AI scoring works walks through what happens to everyone who gets past it: 1 credit per lead scored, per ICP, out of the 100 credits a month a paid plan gives you (10 during the trial).

What happens next

MagicPost analyses your ICP. It reads your description and your example profiles, then builds its own structured understanding of who you are after. This takes a few seconds. While it runs, the ICP card shows Creation in progress...; when it lands the card shows Ready and lists the key signals it will be looking for.

An ICP has to be Ready before you can run a detection with it.

At the same time, MagicPost generates the keywords used for the free pre-qualification pass. You can review and adjust them at any time from the ICP editor (see How to edit an existing ICP).

Your work is saved as you go

If you close the wizard halfway through, your draft is kept for 7 days. Come back and a banner offers to pick up where you left off. The draft lives in the browser you started it in, so it will not follow you to another device.

How many ICPs can I have?

Three active ICPs, on the trial and on paid plans alike. Archived ICPs still take a slot, so delete one to free space. The New ICP button tells you when you have hit the cap.

If your organisation has shared ICPs, you can adopt one instead of building your own. It counts toward your three.

FAQ

Can I duplicate an ICP?

Not directly. Create a new one and reuse the same description as a starting point.

Should I really bother with the example profiles?

Yes. An ICP with 3–5 good examples scores noticeably better than one built on a description alone.

What if I target two very different segments?

Build two ICPs. A single ICP pulling in two directions produces mush. Every profile is scored against each of your ICPs separately and can come out strong on one and weak on the other.

Need help?

Reach out via the in-app chat, or book a 30-minute onboarding call: https://cal.com/magicpost-team/contact-sales-team