Raw identified visitor data is noisy. Without an ICP filter, you'll see internet service providers, random enterprises, and irrelevant companies cluttering your feed. Your first ICP filter turns the noise off and makes the signal useful.
Your first filter should have no more than 3 conditions. Adding 8 attributes from day one makes it hard to diagnose why the filter is working or not working. Start with your most predictive attributes and expand from there.
For most B2B companies, the best 3-attribute starting filter is: industry match + employee count range + geography. These three conditions eliminate the vast majority of non-ICP traffic without requiring any behavioral data yet.
The most common mistake in the first filter is making the employee range too narrow. If your sweet spot is 100–500 employees, start your filter at 50–750 — a bit wider than your true ICP. As you see what's coming through and what's converting, tighten the range in subsequent iterations.
Once your firmographic filter is live and you've reviewed what it surfaces for a week, add one behavioral condition: require that the company visited at least 2 pages, or spent at least 3 minutes on site. This eliminates single-page bounces that are unlikely to convert even if they're ICP-fit.
Check your filter performance weekly for the first month. Look at: how many accounts pass the filter per day, how many you actually reached out to, and how many replied. If you're seeing more than 20 per day, tighten. If fewer than 3, loosen. The goal in the first 30 days is to find the setting that produces 5–15 actionable accounts per day.