TL;DR
- In SignalRaven data, about 1 in 30 people who engage with a watched LinkedIn post become a qualified signal. The other 29 aren't discarded by one filter.
- SignalRaven works through a 14-stage process that checks the post, identifies and enriches the person, evaluates ICP fit, removes staff and sellers, scores urgency, adds context and routes the signal.
- The biggest drop happens early. The pre-screen alone removes about 45% of engagers before company research begins.
- If your signal feed feels too thin, start with your sources and target titles before loosening ICP thresholds.
A LinkedIn reaction or comment isn't automatically a useful signal.
In SignalRaven data, about 1 in 30 people who engage with a watched LinkedIn post become a qualified signal. That's intentional. The rest may be staff at a company you track, people at competing vendors, students, people outside your ICP, or engagers on a post that wasn't relevant enough in the first place.
SignalRaven works through 14 stages between discovering a post and delivering a signal. Some stages filter. Others research the person and company, score fit and urgency, add context or route the signal to your team. That process is what separates useful signals from everything else.
Here's what happens between LinkedIn engagement and a signal in your inbox.
Fictional example. Imagine a 200-person payroll software company monitoring a competitor's LinkedIn page. Four people comment on one post. One works for the competitor. One is a student. Another works for a different payroll vendor. The fourth is an operations director at a 400-person logistics company that fits the ICP. After qualification, that fourth engagement becomes the signal.
The 14-stage qualification process
SignalRaven calls these stages the 14 qualification gates, and they run in a fixed order. Not every stage is a filter. Some decide what gets through. Others enrich, score, explain or deliver the signal. Read in order, they trace the whole path: a relevant post, its engagement, a person identified, the person and company enriched, staff and sellers sorted out, ICP fit evaluated, urgency calculated, context and talking points added, a signal delivered.
Find the right engagement
Everything starts with the post. A relevant post is the only kind worth collecting engagement from.
| # | Stage | What it does | Who drops out |
|---|---|---|---|
| 1 | Post relevance | Parks low-engagement posts and scores the rest Strong, Relevant, Tangential, Weak or Irrelevant | Posts scored Tangential or lower (the default cutoff) |
| 2 | Post summary | Summarizes each qualified post once | Nobody |
| 3 | Engagement harvesting | Collects comments and reactions | Nobody |
Understand the person
Next, SignalRaven works out who engaged and removes the people it can rule out from the profile alone.
| # | Stage | What it does | Who drops out |
|---|---|---|---|
| 4 | Title check | Reads a reactor's headline against your target titles and personas | Off-target or blank headlines |
| 5 | Person enrichment | Pulls the LinkedIn profile | People with no findable profile |
| 6 | Staff filter | Checks the employer against every company page you track | Their staff (yours too if you track your page) |
| 7 | Pre-screen | Reads the profile's title (and location, if targeted) | Clear mismatches |
Qualify the person and company
With the person identified, SignalRaven researches the employer, classifies the commercial relationship and scores ICP fit.
| # | Stage | What it does | Who drops out |
|---|---|---|---|
| 8 | Company enrichment | Researches the employer | Nobody |
| 9 | Commercial classification | Matches the employer against your tracked companies and your own, then classifies it as a competing vendor (a seller), an influencer or a target company | Sellers; influencers when excluded |
| 10 | ICP scoring | Scores title and role, market affinity, industry, company size and location | Anyone below an enabled dimension's minimum |
Turn qualification into an actionable signal
The remaining stages add the intelligence a rep reads and route the signal to where your team works.
| # | Stage | What it does | Who drops out |
|---|---|---|---|
| 11 | Urgency score | 1 to 10: an AI base plus increases for earlier signals from the same person or company, a Strong post and an employer classified as a target company | Signals below urgency 8 on people you follow |
| 12 | Talking points | Writes them for signals at urgency 5 or higher | Nobody |
| 13 | Why it matters | Explains the person's role, company and commercial fit | Nobody |
| 14 | Delivery | Routes signals by your destination rules | Nobody |
Qualification tells you who fits. Urgency tells you what deserves attention now.
Passing the ICP doesn't make every signal equally important. The urgency score also weighs the context around the engagement, including post relevance and previous signals from the same person or company. A single engagement can be useful. A pattern of relevant engagement is more meaningful.
Where most engagement gets filtered
In SignalRaven data, each filter removed this share of all engagers:
- Pre-screen: about 45%
- ICP scoring (including commercial classification): about 18%
- Staff filter: about 8%
Reactions arrive with just a headline, so only reactions face the title check. It removes about 3 in 10 reactors (27% to 36% by source).
The biggest source of noise isn't sophisticated ICP scoring. It's much simpler: the person's role doesn't match who you're trying to reach.
Competitor engagement is noisy by default
About 3 in 10 people reacting to a competitor's company-page post are the competitor's own employees. About 1 in 5 reactors on a company's own page are its staff. Comments carry less staff noise: about 1 in 6 commenters on competitor pages work there.
That's why raw engagement counts can be misleading. SignalRaven evaluates the person behind the engagement before creating a signal. Judge a competitor page by its signals, and expect its reactions to include the rival's own staff.
Why someone who looks like a fit can still be filtered out
SignalRaven classifies someone as a seller when their company offers a direct substitute for your product. Sellers are filtered out. When the classification isn't clear, the person continues through qualification.
Influencers are handled differently. Consultants, agencies, systems integrators, recruiters and advisors may still be relevant to your market, so SignalRaven labels them rather than automatically removing them. You decide whether to include them. The Exclude influencers toggle is off by default and open on every plan.
At ICP scoring, each dimension you've switched on must clear its own minimum. A fictional VP of Operations with strong title, market affinity and company size scores still drops if her employer's industry misses your minimum. The average only ranks people who pass.
How to read the dashboard funnel
| Stage | What it counts |
|---|---|
| Discovered | Posts found (low-engagement posts too) |
| Qualified | Posts scored Strong or Relevant |
| Watching | Qualified posts still being rechecked |
| Engagement | Every comment and reaction captured |
| Signals | Signals created from those engagements |
Quiet posts wait below the engagement floor and are rechecked if engagement picks up, so Discovered normally runs well ahead of Qualified.
Start with post relevance. If most posts are Tangential, Weak or Irrelevant, your sources are the problem. Each source's page shows the same breakdown.
If posts are relevant but engagement isn't turning into signals, look at who is engaging. You may be attracting peers, employees, sellers or people outside your target roles.
Only after the sources and titles look right should you consider changing ICP thresholds.
What to tune when the funnel is too tight or too loose
Sources decide who shows up, titles decide who gets through, and thresholds decide how close a near miss can be. That order matters because of what each change admits. A title edit adds only the roles you name. A lower threshold admits everyone who sat just under the old line.
When signals run thin
- Check your sources. Are you finding the right conversations?
- Check your titles. Are you targeting the right people?
- Check your thresholds. Are good matches falling just below qualification?
Here's what each check means in the portal:
- Sources. Rewrite or replace sources whose posts score Tangential or below. See choosing sources and how to write keyword searches that find your target audience.
- Titles. Widen target titles and personas, and check your locations. An exclude title as broad as "manager" can knock out the ops managers you sell to.
- Thresholds. On the Scale plan, lower one dimension threshold at a time and review the new signals before the next change.
A noisy feed needs the reverse. Turn on Exclude influencers, add exclude titles and add classification examples for resellers, partners and adjacent vendors. Swap a busy company page for the "A company's posts, filtered by topic" source. On the Scale plan, raise the minimum on the dimension letting noise through.
Once the feed looks right, the three talking point playbook covers what to say to the people who come through.
FAQ
What percentage of LinkedIn post engagers become signals?
About 3%, or roughly 1 in 30, become qualified signals in SignalRaven data. This only includes engagement on posts that already passed the post relevance stage, so it isn't 3% of all LinkedIn activity.
What are the 14 qualification gates in order?
Post relevance, post summary, engagement harvesting, the title check (reactions only), person enrichment, the staff filter, the pre-screen, company enrichment, commercial classification, ICP scoring, the urgency score, talking points, why it matters and delivery. Not every stage filters. Posts drop at post relevance. People drop at the title check, person enrichment, the staff filter, the pre-screen, commercial classification and ICP scoring. Influencers also drop when excluded. The urgency score ranks each signal and holds back signals below urgency 8 on people you follow. The other stages collect data, write what your rep reads or route the signal.
Why was someone who looks like a fit filtered out?
Walk the stages in order. A reactor fails the title check when the headline is off-target or blank. Anyone who works at a company whose page you track drops at the staff filter. A title or location that clearly misses your targets fails the pre-screen. That's the most common drop. An employer classified as a seller drops the person at commercial classification. With company size switched on, anyone at a company far outside your employee range drops before that check. And a dimension you've switched on can fall below its minimum even when the average looks strong.
Does SignalRaven filter out my own employees and my competitors' employees?
Yes. The staff filter drops anyone who works at a company whose LinkedIn page you track. That covers every competitor page you watch and your own page if you track it. Commercial classification then matches employer names against your tracked companies and your own company name, and an AI review settles close matches.
What's the difference between a seller and an influencer?
A seller works for a company that offers a direct substitute for your product. Sellers are filtered out. An influencer works for a company that advises, implements, integrates or recruits for companies in your market: consultancies, agencies, systems integrators, recruiters and advisors. Influencers arrive with an influencer label unless you turn on Exclude influencers. The toggle is off by default and available on every plan.
Can I change the ICP thresholds?
Yes, on the Scale plan. On every plan, admins and managers can edit target titles, exclude titles, industries, locations and the other basic ICP fields, and they can use the Exclude influencers toggle. Setting target locations switches on a high location minimum automatically. SignalRaven sets the engagement floor and the post relevance cutoff. Neither sits in the ICP editor.
Why are reactions judged on the headline first?
A reaction gives SignalRaven only the reactor's headline to go on. The title check reads that headline against your target titles and personas before any profile research, and a blank headline fails. Commenters skip this stage and go straight to person enrichment. In SignalRaven data, the title check removed about 3 in 10 reactors.
Sources
- SignalRaven data, aggregated across customer workspaces
- SignalRaven pricing: ICP editing rows