Most ICP documents fail because they capture a moment in time\u2014customer mix, product, and positioning shift, so a static write-up stops matching who actually converts.<\/p>\n
You define your ideal customer profile<\/a> once, usually in a strategy meeting, then file it away. Three months later, your product mix shifts, your best customers look different, and the targeting brief no longer reflects reality.<\/p>\n A usable ICP research template is not a\u00a0static document. It’s an evidence-gathering workflow you can rerun when you need fresh targeting data. Store inputs and outputs in Google Sheets and schedule refreshes with PhantomBuster so the same steps run reliably each cycle.<\/p>\n This article shows how to build that workflow in PhantomBuster. Start from your best customers, resolve their LinkedIn URLs, extract profile and company data, normalize the fields, then turn the result into a structured ICP brief.<\/p>\n Use PhantomBuster’s Google Sheets integration to manage inputs and outputs, and the Scheduler to run the same flow on a cadence.<\/p>\n Static ICP definitions are often built on assumptions. Teams brainstorm job titles, industries, and company sizes, then turn that into a targeting document. That document usually describes who you think you sell to, not who consistently converts and stays.<\/p>\n That document also drifts as your customer base and positioning change. A document written once can become misleading within a quarter. One common mistake: recency bias. One recent win becomes the template, even if most revenue comes from a different role or segment.<\/p>\n You need one input: a clean list of your best customers. Export contact names, company names, and any LinkedIn URLs you already have from your CRM. Define “best” by retention, revenue, expansion, or product fit\u2014not just recent deals. Your output should be a structured ICP brief with:<\/p>\n Map functions to Sales Navigator seniority and function filters, apply the size band as company headcount, and seed lookalike audiences with the normalized titles and keywords from your brief.\u00a0Example: Function = Marketing + Seniority = Director\/Head; Headcount = 51\u2013200; Keywords = “attribution” OR “pipeline.”<\/p>\n Customer-led sampling removes guesswork. You already know these people converted and retained, so patterns from this group are easier to defend.<\/p>\n When teams start from broad LinkedIn searches, they often profile strangers who look plausible. That can produce a neat spreadsheet without proving who actually buys.<\/p>\n Starting from customers flips the logic. You begin with known outcomes, then work backward to find shared attributes. If your best customers cluster around Operations Directors at 100\u2013200 employees, treat that as a targeting signal. If they spread across roles and sizes, plan for separate ICPs<\/a>.<\/p>\n Export your best-customer cohort from your CRM. Include:<\/p>\n Remove duplicates and stale records before processing. Bad inputs create bad LinkedIn matches, which then contaminate the rest of the workflow. If you only have names and companies, run a resolution step first.<\/p>\n Use PhantomBuster’s\u00a0LinkedIn Profile URL Finder<\/strong> to match name plus company, or name plus professional email, to return the LinkedIn profile URL most consistent with the name and company or email. Spot-check 20\u201330 records to confirm accuracy before scaling.<\/p>\n As part of the same flow, use PhantomBuster’s\u00a0LinkedIn Company URL Finder<\/strong>\u00a0to resolve missing company page URLs before enrichment.<\/p>\n Tip: start with 20 to 30 customers and check the matches manually. A small bad match rate can distort your ICP analysis once you scale.<\/p><\/blockquote>\n Once you have profile URLs, extract the fields that show role, context, and language. Useful fields include headline, current role, location, company name, recent experience, skills, and About text. These are often enough to spot repeated role patterns and the words your best customers use to describe their work.<\/p>\n PhantomBuster’s LinkedIn Profile Scraper<\/strong> extracts public profile data from a list of URLs\u00a0as part of a single PhantomBuster workflow that starts with URL resolution and ends in a structured ICP brief stored in Google Sheets. Use it to pull the same fields each time, so every refresh is comparable. Treat pacing as a pattern question.<\/p>\n Keep activity steady during normal working hours and avoid spikes\u2014risk increases when behavior changes abruptly. Use PhantomBuster’s Scheduler to spread runs and keep volume consistent.<\/p>\n Start with 20\u201330 profiles per day. If sessions remain stable for a week, increase volume by 10\u201315% the next week.<\/p>\n Profile data alone isn’t enough. You also need company-level context. In the same PhantomBuster workflow, add company data (employee count, industry, website, HQ) with PhantomBuster’s LinkedIn Company Scraper<\/strong> so your profile attributes line up with firmographics in one dataset.<\/p>\n Use company data to answer basic questions:<\/p>\n Raw exports are messy. Titles vary. Industries are labeled inconsistently. Company sizes need grouping. Normalize before you analyze\u00a0so one-off titles and labels don’t skew your counts.<\/p>\n Titles:<\/strong> Map messy titles into functions such as Marketing, Sales, RevOps, Operations, Finance, or Customer Success. Use a controlled list in your Google Sheet (data validation) or a small PhantomBuster + Sheets script to map titles to functions consistently.<\/p>\n Company size:<\/strong> Bucket employee counts into bands, such as 1 to 10, 11 to 50, 51 to 200, 201 to 500, and 500+.<\/p>\n Industries:<\/strong> Merge similar labels when they mean the same thing for your go-to-market motion. Create a two-column mapping table in your Sheet (raw \u2192 normalized) and VLOOKUP or INDEX-MATCH it during refresh to keep results consistent.<\/p>\n You can do this manually, with spreadsheet rules, or with LLM-assisted categorization. If you use AI, keep a human review step for edge cases.<\/p>\n Exact titles are noisy. Two people can do the same job with different titles. Prioritize function over job title for targeting\u2014it’s a more reliable signal across companies.<\/p>\n Group roles like “Director of Demand Gen” and “Head of Growth Marketing” together when the scope is similar. Look at responsibilities, keywords, and context, not just title strings.<\/p>\n Use repeated terms like “pipeline,” “attribution,” and “demand gen” as primary targeting signals\u2014they reflect responsibilities better than titles.<\/p>\n Look for concentration. If many best customers sit in one headcount band, that’s often a stronger signal than industry.<\/p>\n Companies of similar size tend to share operating constraints. Industry still matters, but treat it as a messaging layer. If most best customers are in SaaS, you have an affinity market. If they spread across several industries, you likely solve a horizontal problem.<\/p>\n Skills and About sections help you understand vocabulary. Repeated tools, methods, or goals can improve outreach and landing page copy. If many customers mention Salesforce, HubSpot, attribution, hiring, or workflow automation, that tells you how they frame the problem.<\/p>\n If you use an LLM to summarize About text, keep the prompt narrow. Ask for common goals, constraints, and vocabulary. Then validate the themes against the raw profiles. Store your prompt in the Sheet next to the inputs so every refresh uses the same criteria.<\/p>\n Avoid building your ICP around one standout customer. Look for patterns that repeat across a meaningful share of the cohort.<\/p>\n If you sell multiple pricing tiers, your best customers look different by tier. If your SMB wins cluster around operators at 20\u201350 employees and enterprise wins around executives at 500+, define separate ICPs and briefs for each.<\/p>\n Don’t average those together. Create separate Google Sheets tabs by tier and run the PhantomBuster workflow on each tab.<\/p>\n If you sell multiple products, create one input sheet per product and run the same PhantomBuster workflow for each to produce distinct briefs. One product might sell to Marketing leaders. Another might sell to Finance or Operations.<\/p>\n One master ICP will blur those differences and weaken targeting. If the data shows distinct clusters, create separate briefs and separate targeting rules.<\/p>\n Don’t process your full customer base on day one. Start with 30 to 50 customers. Check whether URLs are correct, fields are populated, and company matches make sense. Once the workflow is stable, expand in steps. This reduces wasted runs and avoids sharp activity changes.<\/p>\n Use PhantomBuster’s Scheduler to spread runs across normal working hours\u00a0(e.g., 9am\u20135pm in your time zone). Avoid stacking multiple PhantomBuster LinkedIn Automations on the same account at the same time.<\/p>\n Watch for session friction\u00a0(forced re-auth, cookie issues, checkpoints). If you see them, pause runs, refresh your PhantomBuster session cookie, and reduce volume by 10\u201320% for the next week.<\/p>\n Consistency beats bursts. Build the workflow in layers first, then scale once it runs cleanly.\u00a0Save each step as a PhantomBuster Automation with the same input sheet and enable Scheduler once tests run cleanly.<\/p>\n Use PhantomBuster’s Google Sheets integration as your input layer. Update the sheet, then re-run the same PhantomBuster Automations manually or on a schedule with the Scheduler. A simple structure works:<\/p>\n Keep sensitive data out of shared sheets unless it’s necessary. For ICP research, names, companies, and URLs are usually enough.<\/p>\n Set a monthly refresh if your product or segment changes quickly; otherwise schedule a quarterly run with PhantomBuster’s Scheduler. Each refresh follows the same steps: export best customers, resolve URLs, extract profile data with PhantomBuster, extract company data with PhantomBuster, normalize fields, then update the brief.<\/p>\n Compare each new brief to the last one. If company size, role, or industry clusters shift, update targeting and messaging.<\/p>\n A LinkedIn ICP research template isn’t a slide deck. It’s a repeatable workflow that turns your best customers into structured evidence. Start from customers you’ve already won.<\/p>\n Resolve LinkedIn URLs, extract profile and company data, normalize the fields, then look for patterns you can defend. If the data shows multiple ICPs by tier or product, split the workflow. Separate ICPs usually produce cleaner targeting than one blended average.<\/p>\n For a deeper look at how to turn this data into action, see our guide on ICP list building<\/a>. Start your free trial<\/a><\/p>\n Start with 30 to 50 best customers. Use the first run to validate LinkedIn matches and data quality before scaling.\u00a0A small test batch lets you spot resolution errors and field gaps without wasting volume on bad inputs.<\/p>\n Run a resolution step first\u00a0with PhantomBuster’s LinkedIn Profile URL Finder (people) and LinkedIn Company URL Finder (companies). Match name plus company, or name plus professional email, to return the profile URL.\u00a0Spot-check the first 20\u201330 matches to confirm accuracy before scaling to your full list.<\/p>\n Set a monthly refresh if your product or segment changes quickly; otherwise schedule a quarterly run with PhantomBuster’s Scheduler. The refresh only matters if you compare it to the previous version and update targeting when patterns change\u2014otherwise you’re just collecting data.<\/p>\n Extract headline, current role, location, company name, recent experience, skills, and About text. These fields show role, context, and vocabulary. They’re enough to spot repeated patterns in how your best customers describe their work and responsibilities without needing a full profile dump.<\/p>\n Use PhantomBuster’s Google Sheets integration as your central input and output layer. Feed one Sheet to multiple Automations in sequence (URL Finder \u2192 Profile Scraper \u2192 Company Scraper), then pull all outputs back into the same Sheet for normalization. Schedule each Automation with time gaps so they don’t overlap.<\/p>\n Start with 20\u201330 profiles per day and keep activity steady during normal working hours. If sessions remain stable for a week, increase volume by 10\u201315% the next week. Risk comes from sudden behavior changes, not absolute numbers\u2014so ramp gradually and use PhantomBuster’s Scheduler to spread runs evenly.<\/p>\n Store a status column in your Google Sheet to track which records have been processed, which failed, and which need re-checking. Use data validation for tier, product, and function fields so new entries follow the same controlled vocabulary. Archive each refresh as a separate tab so you can compare patterns over time.<\/p>\nWhat a reusable ICP research template means in practice<\/h2>\n
Why static ICP documents fail<\/h3>\n
What you need as input, and what you should get as output<\/h3>\n
\n
Prepare your best-customer input list<\/h2>\n
Why start from customers, not prospects<\/h3>\n
How to clean a CRM export and resolve LinkedIn URLs<\/h3>\n
\n
What the workflow looks like: profile data first, then company data<\/h2>\n
Step 1: Extract profile-level ICP attributes<\/h3>\n
Step 2: Add firmographic company data<\/h3>\n
\n
Step 3: Normalize and structure the dataset<\/h3>\n
Interpret the output without overfitting<\/h2>\n
Titles vs functions: what matters for targeting<\/h3>\n
Company size clusters and industry patterns: how to read the signal<\/h3>\n
Skills and About text: use them for language and context<\/h3>\n
Handle edge cases: when one ICP is not enough<\/h2>\n
Tier-specific ICP divergence<\/h3>\n
Multi-product organizations<\/h3>\n
Keep the workflow responsible and repeatable<\/h2>\n
Small batches first, then scale<\/h3>\n
How to pace and schedule for account health<\/h3>\n
Make the template reusable across refresh cycles<\/h2>\n
Use a control center sheet for inputs<\/h3>\n
\n
How to schedule refreshes<\/h3>\n
Conclusion<\/h2>\n
FAQ<\/h2>\n
How many customers should I include in my first ICP sample?<\/h3>\n
What if I only have customer names and companies?<\/h3>\n
How often should I refresh my ICP research?<\/h3>\n
What fields should I extract from LinkedIn profiles for ICP research?<\/h3>\n
How do I chain PhantomBuster Automations into one repeatable run?<\/h3>\n
What’s a safe daily volume for extracting LinkedIn data?<\/h3>\n
How do I keep my dataset clean between refresh cycles?<\/h3>\n
How do I map my ICP brief to Sales Navigator filters?<\/h3>\n