Reply tagging to refine your ICP: a simple intent-based method

Why replies are the fastest reality check for your ICP

Your ideal customer profile (ICP) isn’t a slide deck. It’s a clear description of who you help, when the pain shows up, and why they’ll pay to fix it.

A simple ICP usually answers four questions:

• Who: the roles, company types, and team sizes you serve well
• When: the trigger that makes your offer feel urgent (a new hire, new tool, new goal)
• Why: the problem they want gone and the outcome they want instead
• Buying reason: why they choose a paid solution instead of doing nothing

Cold outreach is where that story meets real people. Your ICP can look perfect on paper, but replies show what’s true. Sometimes the segment you expected to care doesn’t answer at all. Other times, a different segment replies quickly because your message hits something real.

Replies beat vanity metrics because they cost attention and time. An open can be an accident. A click can be curiosity. A reply usually means the person understood what you wrote and felt enough friction or interest to respond.

Not every reply is a win, though. The fastest reality check is intent, not volume. The goal of reply tagging isn’t “more replies.” It’s spotting which replies signal a real problem, a real timeline, or a real next step.

A few examples:

• “Not now, we just signed a 12-month contract” is often more useful than “Thanks,” because it confirms they buy solutions like yours.
• “We don’t do outbound” is useful too, because it tells you your targeting or positioning is off for that segment.

Even a small sample can teach you something. Ten replies can reveal patterns like “founders respond, managers don’t” or “agencies push back, SaaS teams ask questions.” Tools that auto-sort replies can help you see patterns sooner, but the insight still comes from the words people use and what they’re trying to do next.

What counts as intent in a reply (and what doesn’t)

To use replies to refine your ICP, you need a clean line between people moving toward a buying action and people being polite. If you mix those together, your “best segment” becomes the one that replies the most, not the one that buys.

A practical rule: intent means the person is trying to progress the conversation, or they’re revealing buying conditions (timing, process, budget, fit).

High-intent signals (real progress)

These replies make the next step obvious and teach you something about fit:

• Asking for a call: “Can you share times?” “Next week works.”
• Asking about price: “What does this cost?” “Is it per seat?”
• Asking about timing: “Could we do this this quarter?” “We’re evaluating next month.”
• Requesting decision material: “Send a deck,” “case study,” or “one-pager.”
• Clarifying requirements: “Do you support X?” “Can you integrate with Y?”

Even a short reply can be high-intent if it includes a constraint you can act on, like: “Interested, but only if you work with companies under 50 people,” or “We already have a tool, what’s different?”

Low-intent signals (polite, vague, or non-committal)

These replies can sound positive, but they don’t move the deal forward. Tag them separately so they don’t inflate your results:

• “Not now” with no timing or trigger
• “Send info” with no question
• “Sounds interesting” with nothing specific
• “Maybe later” or “We’ll keep you in mind”

A simple test: if you reply once and still don’t know what to do next (schedule, qualify, or disqualify), it’s probably low intent.

Negative intent (a clear stop signal)

These replies are still valuable data. You just don’t try to convert them:

• Removal requests: “Remove me,” “Stop emailing,” “Unsubscribe”
• Hard objections: “Not a fit,” “We never use vendors for this,” “Don’t contact our team”
• Hostile replies: treat as stop, tag it, end the thread

Auto-replies and courtesy replies deserve their own bucket. Out-of-office messages, “I’m on leave,” or forwards to a generic inbox aren’t intent. They’re delivery and routing signals.

A tagging model you can keep consistent

Reply tagging works best when it stays boring. The fastest way to break it is to create 30 tags and let everyone invent new ones.

Aim for a small model (around 6 to 10 tags per layer) that takes under 10 seconds per reply. A simple structure is three layers:

A simple rule: intent is the outcome, segment is the context, reason is the driver.

Here’s a tight starting set most teams can stick to:

• Intent: Interested, Not interested, Referral, Out-of-office, Unsubscribe
• Reason: Bad timing, Not my role, Already has vendor, Too expensive, Not relevant

For reason tags, avoid vague labels like “not a fit.” Force yourself to pick something you can act on.

Segment tags should match how you actually target. If your lists are built by role and company type, use segments like “Founder-led SaaS” or “Ops leader in logistics,” not “Tech.”

Example: someone replies, “Not me, talk to our Head of RevOps.” You might tag it as Intent = Referral, Segment = Mid-market SaaS, Reason = Not my role. That single pattern tells you both which titles to target next and which ones to avoid.

If your inbox volume is high, automate the sorting where you can, but keep the same tag model. For instance, LeadTrain (leadtrain.app) can automatically classify common reply outcomes like interested, out-of-office, bounces, and unsubscribes, so your team can spend time tagging the segment and reason that actually shape the ICP.

A few guardrails help keep tagging consistent:

• Write one sentence for each tag (what counts, what doesn’t)
• Don’t add new tags for a week; park “new ideas” in a note
• If two tags feel similar, delete one
• Tag the reply, not your mood (use the words in the message)
• Review 20 tagged replies together once, then let the system run

Set up your tagging workflow before you send more emails

The biggest win is consistency. A perfect model nobody follows is worse than a simple one that gets used every day.

Start by choosing a few fields you always capture, even when you’re busy. These should answer: who replied, what kind of company they are, and why they responded.

Pick a few fields you can keep forever

Four to six fields is enough for most teams:

• Company type (industry or business model)
• Role (job title or function)
• Use case (the problem they mention, even loosely)
• Geography (country, region, or time zone)
• Company size (rough bucket: 1-10, 11-50, 51-200, etc.)

Write the values you expect to see (SaaS, agency, e-commerce) so people don’t invent new labels every time.

Decide where tags live (and make it easy)

Pick one place where everyone can find the latest tags: a CRM field, a shared spreadsheet, or inside your email platform. The best choice is the one that reduces copying and pasting.

Set a cadence that matches the speed of your outreach. Tag replies within 24 to 48 hours so the context is still fresh. If you wait a week, people forget what the prospect meant, and tagging turns into guesswork.

Finally, write a one-page rulebook that answers common debates:

• How to tag vague replies (“Maybe later”)
• What counts as a use case vs. a feature request
• How to handle multi-person threads (tag the decision-maker)
• When to mark a reply as “not ICP” and why
• What to do when you’re unsure (pick the closest tag and add one note)

Example: if a prospect says, “We already have an in-house SDR team,” your use case might be “outbound support” or “deliverability” depending on what they ask next. The rulebook makes sure two people tag that reply the same way.

Step-by-step: turn replies into ICP insights in one afternoon

You don’t need a data team. You need a small batch of real replies, a consistent tagging habit, and one simple count.

Pick a time box (90 minutes works) and a sample (50 to 200 replies). Use auto-sorting if you have it, but still skim the text so you don’t miss context.

1. Collect replies, then remove noise first. Pull the last 50 to 200 replies from recent campaigns. Set aside bounces and out-of-office messages. They matter for deliverability and timing, but they don’t tell you who wants the offer.

2. Tag in this order: Intent, then Reason, then Segment. Start with what the reply means (interested, not interested, referral, unsubscribe). Next, tag the main reason (already has a vendor, no budget, bad timing, wrong persona). Only then tag the segment (industry, company size, role, region). This order keeps you honest.

3. Count intent by segment using a simple table. Rows = segments, columns = intents. Add totals. You’re not aiming for perfection. You’re looking for which segments produce real “interested” signals and which produce polite brush-offs or role mismatch.

4. Read a small set of replies from the top segments. For the 2 to 3 segments with the most “interested” (or the strongest objections), read about 10 replies each. Highlight repeated wording, especially urgency and constraints.

5. Turn what you found into one ICP update and one targeting rule. Rewrite your ICP in plain language using the words prospects used. Then adjust targeting rules so you get more of what worked and less of what didn’t.

Example: you might find that “IT managers at 200 to 500 employee firms” aren’t rejecting you, they’re saying “we can’t change vendors mid-contract.” Meanwhile “RevOps at 20 to 80 person SaaS companies” replies with “can you share pricing” and “we need this this quarter.” That’s an ICP signal, not just a messaging tweak.

From tags to decisions: what to change in targeting and messaging

Tagging only matters if it changes what you do next. Once you have 30 to 50 replies, turn the labels into a simple decision system.

Start with a basic intent score that matches how close someone is to a meeting:

• High intent: asks about pricing, timing, setup, or says “yes” to a call
• Medium intent: mild interest but needs one key detail (fit, region, budget, use case)
• Low intent: polite reply with no next step
• Negative: not a fit, already has a vendor, angry, unsubscribe, wrong person

Map each level to one default action so your inbox doesn’t turn into a guessing game:

• High intent: propose 2 time options and confirm the goal of the call
• Medium intent: ask one qualifying question (only one) and offer a call if yes
• Low intent: pause follow-up for a set time or move to a lighter nurture sequence
• Negative: stop outreach and record why (so you don’t target that profile again)

When you review results, don’t look only at reply rate. Track time-to-meeting by segment (title, company size, industry, use case). A segment that replies a lot but takes 12 days to book is often a distraction compared to a segment that books in 2 days.

Use reply content to add disqualifiers. If many replies say “we’re agency-only” or “we don’t buy in EMEA,” that’s targeting guidance. Add those as stop rules in your lead filters.

Common mistakes that make reply-based ICP work misleading

Reply tagging only works if the data stays comparable. Most teams tag for a week, then quietly change the rules. After that, you’re no longer learning about your ICP. You’re learning about your tagging habits.

Mistakes that usually ruin reply-based insights:

• Letting tag meanings drift. If “Interested” sometimes means “asked for pricing” and later means “any positive tone,” your trend lines become meaningless.
• Creating too many tags. When people have to choose between 14 labels, they stop tagging or they guess.
• Treating reply rate as success without checking intent. Replies full of “not a fit” or “remove me” are a warning, not a win.
• Ignoring negative-intent patterns. Lots of “already have a vendor,” “no budget,” or “wrong role” is a strong signal your targeting is off.
• Overreacting to one loud reply. Look for patterns across 20 to 50 replies.

Lock your tag definitions for one campaign cycle. If you truly need a new tag, add it, but don’t redefine old ones mid-stream.

Example: one week of replies that changed the target segment

A small HR software vendor sold a tool that helps teams handle onboarding paperwork and policy sign-offs. Their first ICP was mid-market operations leaders (Director of Operations, VP Ops) at 200 to 1,000 person companies.

They sent a 5-day sequence to 320 contacts. The pitch was simple: reduce onboarding admin time and keep every document in one place.

By Friday, they had 41 replies. The tags made the pattern obvious:

• Interested: 6
• Bad timing: 9
• Not my role: 18
• Other (out-of-office, unsubscribe, unclear): 8

“Not my role” looked like a dead end until they read the text. Many ops leaders replied with some version of: “HR owns this. Try our People Ops manager.” A few even forwarded the email to HR.

The surprise showed up inside the “Interested” tag. Four of the six interested replies were from People Ops and HR Operations roles at 150 to 400 person companies, especially in professional services (agencies, consultancies, accounting firms). Ops leaders at the same types of companies mostly responded with “not my role” or “bad timing.”

They updated the ICP in two ways:

• Shifted the target role to People Ops Manager, HR Ops Lead, and Head of People (with ops leaders as optional secondaries)
• Narrowed the segment to services firms where onboarding happens constantly due to project staffing and churn

One message tweak followed immediately. Instead of leading with “ops efficiency,” they opened with “new hire paperwork without chasing signatures,” then added: “If HR owns onboarding for you, I can send this to the right person.” That single sentence reduced “Not my role” replies the next week and increased clean handoffs.

Quick checklist: keep the system simple and useful

If reply tagging starts to feel like extra admin, it’ll get dropped. The goal is a small system you can run every week without a meeting and without a spreadsheet that grows forever.

Use this as a quick check:

• Keep the tag set small: a handful of intent tags plus a few reason tags. If you can’t remember the tags without looking, you have too many.
• Tag fast while context is fresh: aim for within 48 hours.
• Review weekly, not monthly: once a week is enough to spot patterns before you waste another 1,000 sends.
• Be willing to stop: if a segment keeps producing negative intent, pause it and test a different slice.
• Save the buyer’s words: copy 1 to 2 lines of exact phrasing and reuse that language in subject lines, openers, and CTAs.

Keep the workflow lightweight: decide who tags replies (one person per inbox), when it happens (a daily 10-minute block), and where tags live. If the output doesn’t change targeting, messaging, or the offer, you’re collecting labels, not learning.

Next steps: run a small test and tighten your ICP

Treat this like a short experiment, not a big rebuild. Keep the offer the same and compare two segments your tags suggest behave differently.

A simple 2-week test plan:

• Pick 2 segments and pull equal-sized lists (same country, same titles, similar company type).
• Keep the offer constant and send the same sequence to both segments.
• Tag every reply using the same intent model.
• After week 1, adjust list filters only.
• After week 2, adjust only the first email based on the biggest tag differences.

Keep the loop tight. Tighten targeting first. Then tweak messaging. Changing copy, targeting, and offer at the same time makes results hard to trust.

Make sure your reply data stays clean by keeping deliverability healthy. If inbox placement is unstable, you end up measuring spam filters, not market intent.

A deliverability baseline:

• Use proper SPF, DKIM, and DMARC.
• Warm up new mailboxes before sending at scale.
• Keep volumes steady during the test.
• Remove bounces and honor unsubscribes fast.

If you want less time spent on inbox triage, an all-in-one setup like LeadTrain can help by handling domains, mailboxes, warm-up, sequences, and reply classification in one place. The key is still the habit: consistent tags that turn replies into decisions.