Most lead scoring criteria lists look the same. Job title, company size, email clicks, form fills, maybe a pricing page visit. Those criteria matter for demographic fit and marketing engagement. But if you're running a product-led motion, they miss the strongest signals you have.
Product usage data is where conversion behavior actually lives. A user who invited three teammates and completed an integration last week is telling you something that no form fill ever will. The challenge is figuring out which product actions actually predict conversion and which ones are just activity. Not all usage is signal. Most of it is noise.
This guide gives you a framework for identifying the product usage milestones that matter, validating them against your own data, and operationalizing them for scoring. Whether you build this yourself or use a tool like TrailSpark to evaluate milestones alongside demand gen and fit signals, the framework is the same.
Why Most Product Signals Don't Predict Anything
The activity trap
Teams that invest in product analytics often have the opposite problem from teams with no data: they have too much of it. Every click, every page view, every session gets tracked. The temptation is to pour all of that into scoring, because more data should mean better predictions. It doesn't.
Raw event counts rarely correlate with conversion. A user who logged in 47 times might be deeply engaged, or they might be confused and unable to find what they need. A user with 200 API calls might be building something real, or they might be running automated tests that will never turn into a paid use case. Volume alone doesn't tell you which.
" Teams track everything and score nothing useful. Raw event counts rarely correlate with conversion. "
The problem with vanity metrics
"Daily active users" is a product health metric. It's not a scoring signal. High DAU can mask low value realization. A user might log in every day but never complete setup. Another might visit once a week but have rolled the product out to their entire team and integrated it into their workflow.
The same logic applies to session duration, feature usage counts, and most aggregate metrics. They tell you something about product engagement in the abstract. They don't tell you who is ready to pay.
What you're actually looking for
The milestones that predict conversion share a few characteristics:
- They represent value realization - The user experienced what makes your product useful, not just what makes it clickable
- They indicate commitment - The action took effort or involved a decision, like completing setup, inviting a colleague, or connecting an integration
- They correlate with retention - Users who hit these milestones don't just convert. They stick around
Activity tells you someone is present. Milestones tell you someone is progressing.
The Milestone Framework: Four Categories That Matter
Not all milestones carry the same weight or appear at the same stage. This framework organizes them into four categories based on what they signal about the user's journey.
Activation milestones
Activation is the moment a user first experiences core value. This is not signup. Signup is just the starting line. Activation is the first meaningful action that demonstrates the user understood what your product does and did something real with it.
Examples vary by product:
- Completed onboarding - Finished the setup flow, not just started it
- Created first core object - First project, first report, first workflow, first whatever your product's central unit is
- Connected an integration - Linked an external tool, which signals intent to use your product in context
Users who never activate rarely convert. Activation milestones are your first filter.
Adoption milestones
Activation tells you someone got started. Adoption tells you they came back and went deeper. These milestones separate trial users from people building a habit.
- Returned 3+ times in the first 14 days - Repeat usage in a compressed window is a strong signal
- Used a feature more than once - First use is exploration. Second use is adoption
- Created multiple objects - One project could be a test. Three projects is a workflow taking shape
The specific thresholds (3 visits, 14 days, etc.) will vary for your product. The principle holds: repeated, deepening engagement within a time window indicates adoption.
Collaboration milestones
This is where milestones get interesting for B2B scoring, because collaboration signals organizational buy-in. An individual user exploring your product is useful data. Multiple people at the same company using it is a buying signal.
- Invited a teammate - They're endorsing your product to a colleague
- Shared a project, report, or artifact - The work they're doing in your product has an audience
- Added users to a workspace or team - They're building your product into their team's workflow
" When a user invites a teammate, they're telling you something no form fill ever will: this product is worth sharing. "
Collaboration milestones often correlate more strongly with conversion than individual usage depth. A single power user with heavy activity might never convert if the organization isn't involved. Three users with moderate activity are a much stronger signal.
Expansion milestones
Expansion milestones indicate that a user or account is approaching or exceeding the limits of what's available for free. These are directly tied to revenue triggers.
- Hit usage limits - Storage, seats, API calls, or whatever your free tier caps
- Enabled a premium feature - Tried or toggled a feature that requires a paid plan
- Created multiple workspaces or teams - Usage is spreading beyond the initial use case
Expansion milestones are the closest thing to explicit purchase intent that product data can give you. A user hitting your free-tier limits isn't browsing. They're bumping into the paywall because they're using the product for real.
How to Find Your Milestones
The framework tells you what categories to look for. Finding the specific milestones that matter for your product takes some analysis. Here's how to approach it.
Start with converted users
Pull a cohort of users or accounts that converted to paid in the last 6-12 months. Work backward through their product activity. What did they do before converting? Look for the actions that show up consistently.
Example: Working Backward from Conversion
You're not looking for the one action that every converter took. You're looking for patterns: actions that converted accounts did at meaningfully higher rates than accounts that didn't convert.
Compare to non-converters
The comparison is what makes a milestone predictive. If 80% of your converters completed onboarding but 75% of your non-converters also completed onboarding, onboarding completion isn't a useful scoring signal. It's just a common action.
The milestones that matter are the ones where the gap between converters and non-converters is wide. Look for actions where converters are 3-5x more likely to have completed them.
Look for common sequences
Conversion often follows a path. Maybe most converters activated, then invited a teammate, then created multiple projects, all within a two-week window. The individual actions might be common, but the sequence in a compressed timeframe is rare and highly predictive.
This is where event-level data with timestamps becomes essential. You can't identify sequences from aggregated counts. You need to know what happened and when.
Validate with qualitative input
Data tells you what happened. People tell you why. Talk to your sales team and customer success managers about what they observe in accounts that convert:
- What does a "ready" account look like when they pick up the phone?
- What patterns do they notice in accounts that close quickly?
- What's different about accounts that seem engaged but never buy?
These conversations often surface milestones that aren't obvious in the data, like a specific configuration step that signals serious intent, or a workflow pattern that indicates the product is being embedded into daily operations.
Test and iterate
Start with your best hypotheses based on the analysis. Define 4-6 candidate milestones and track their correlation with conversion over 30-60 days. Some will hold up. Others won't. Refine and test again.
Milestone definitions are not permanent. They're hypotheses you validate and adjust as your product, pricing, and customer base evolve.
The Data Spectrum: What Level of Detail You Need
The quality of your milestone scoring depends directly on the level of product data you can access. Not all product data is created equal, and most scoring systems that claim "product data support" are working with less than you'd expect.
Level 1: Attribute flags. "Feature enabled: true." This tells you a setting exists. It doesn't tell you when it was enabled, how often it's used, or whether it changed anything about the user's behavior. Limited scoring value.
Level 2: Aggregated counts. "Projects created in last 30 days: 12." Better than a flag, but aggregates hide the differences that matter. Were those 12 projects created in a burst yesterday, or spread evenly over a month? Was it one user or four? You can't tell from the number alone.
Level 3: Event-level data. "User X created project Y at time Z." Now you can see what happened and when. You can identify sequences, measure recency, and distinguish between a spike and a slow build. This is the minimum level needed for meaningful milestone scoring.
Level 4: Contextual sequences. "Users typically convert after they complete steps A, then B, then C within a 14-day window." This is where pattern recognition across your full dataset starts producing insights that humans wouldn't find manually. AI scoring systems operating at this level can identify and weigh behavioral progressions, not just individual events.
TrailSpark accepts real-time events through flexible webhooks, so you can send activation, adoption, collaboration, and expansion milestones as they happen at event level. That matters because milestone scoring built on aggregated counts will always be fuzzier than scoring built on timestamped events.
Operationalizing Milestones for Scoring
Identifying milestones is the research. Operationalizing them is where the scoring actually happens.
Pick 5-10 milestones max
More milestones does not mean better accuracy. It often means the opposite. Every milestone you add introduces noise and complexity. A focused set of high-signal milestones will outperform a sprawling list of marginal ones.
Choose the milestones with the widest gap between converters and non-converters. If you're unsure, start with one per category (activation, adoption, collaboration, expansion) and add from there only if the data supports it.
" Pick 5-10 milestones max. More milestones does not mean better accuracy. Often it means the opposite. "
Weight by predictive value
Not all milestones are equal. In most PLG businesses, collaboration milestones (inviting teammates, multi-user activity) predict conversion more strongly than individual usage depth. Your data may show something different, but don't assume a flat weighting across categories is right.
Test the relative predictive power of each milestone against your conversion data. The ones that most separate converters from non-converters should carry the most weight.
Add time decay
A milestone hit yesterday is more actionable than one from 60 days ago. Recency should be a factor in how much weight a milestone carries in scoring.
Simple Recency Rule
Roll up to accounts
Individual milestones matter. But multiple users at the same account hitting milestones matters more. One person activating could be a solo evaluation. Three people adopting and collaborating is an account-level buying signal.
Rolling up product milestones to the account level requires connecting product users to CRM records. TrailSpark handles this through cross-system identity resolution, matching users to organizations even when they signed up with different emails or don't exist in your CRM yet. Without that connection, your milestone data stays siloed in product analytics and never makes it into scoring.
Combine with fit signals
Milestones without fit context can mislead. A free-tier user hitting every milestone from a 5-person startup outside your ICP is probably not worth sales time, no matter how engaged they are. A user with moderate milestone progress from a mid-market SaaS company that matches your ICP segment is a much stronger candidate.
Full-context evaluation means scoring milestones alongside firmographic fit and demand gen engagement, not in isolation. Product behavior tells you what someone is doing. Fit tells you whether they're someone you can close and retain.
Common Mistakes
-
Treating all product events as equally important - A login is not the same as an integration setup. Weight milestones by what they actually predict, not by how often they occur
-
Scoring on totals without considering recency - 50 lifetime logins doesn't mean much if the last one was 3 months ago. Recency should always be a factor
-
Ignoring account context - A single power user with deep activity looks impressive but might be a solo hobbyist. Multi-user engagement at the same account is the stronger signal
-
Setting milestones once and never revisiting - Your product changes. Your pricing changes. Your customer base shifts. Milestones need to be re-validated quarterly at minimum, and after any major product or GTM change
-
Copying another company's milestones - Every product has different activation moments, different collaboration patterns, and different expansion triggers. Framework is transferable. Specific milestones are not. Validate against your own conversion data
Maintaining Milestones Over Time
Product changes break milestones
When your product team ships a new onboarding flow, deprecates a feature, or restructures how workspaces work, your milestone definitions may silently break. The event names might change, the user paths might shift, or an action that used to be rare becomes default.
Build a process for reviewing milestones whenever the product changes. A quarterly check is the minimum. A review triggered by any major product release is better.
Conversion patterns shift
As your customer base evolves, the behaviors that predict conversion may change. Early adopters behave differently than mainstream buyers. Enterprise accounts activate differently than SMB accounts. What worked six months ago may not work now.
Re-validate your milestones against fresh conversion cohorts at least quarterly. If the gap between converters and non-converters narrows for a milestone, it's losing its predictive value.
Feedback loop with sales
The most direct signal on milestone quality comes from the people using the scores. Are milestone-based PQLs actually converting? Are sales reps finding that high-scoring accounts feel ready when they reach out? What are AEs seeing in conversations that the data isn't capturing?
Build a lightweight feedback process. It doesn't need to be heavy. A monthly check-in where sales flags the best and worst scored accounts is enough to keep milestones calibrated.
Quick-Start Checklist
- Audit your current PQL criteria - Are they actual milestones (value realization moments) or just activity counts (logins, page views)?
- Pull conversion cohort data - What did your last 100 converted accounts do before converting? What patterns emerge?
- Define one milestone per category - Pick one activation, one adoption, one collaboration, and one expansion milestone as your starting set
- Check your data level - Do you have event-level data with timestamps (Level 3+)? If not, that's the first gap to close
- Confirm account-level mapping - Can you connect product users to CRM accounts? Without this, milestones stay siloed in product analytics
What to Read Next
If you haven't already, read Why Rules-Based Lead Scoring Breaks Down for context on why product milestones can't be captured effectively in a points-based model.
For the full framework on AI-powered scoring, including how milestones fit alongside demand gen and fit signals, see the 2026 Guide to AI Lead Scoring.
TrailSpark evaluates product usage milestones alongside demand gen signals and ICP fit in a single full-context assessment, with real-time event ingestion and cross-system identity resolution. Sign up free →