The story of whittling fake pipeline noise from 90% contamination down to near zero, and what HubSpot, Salesforce, and every serious sales tech company has quietly been fighting for years.
“Wbel Buzfe wants to become a customer.”
That’s what our system told a sales rep. Then it told them about “Ezbegvafm.” And “Babag Fubu.” And 13 more just like them — all in the same week.
Try the Easy Button™ Sales Intelligence System Free for 30 Days
No risk. No long-term contract. We’ll prove ROI before you spend a dollar — or you walk away. Also includes:
- →Free Market Snapshot — See how many ideal-fit companies exist in your target market
- →1,000 Free Verified Prospect Contacts — Real decision-makers matching your Ideal Customer Profile. Yours to keep.
- →30-Day Proof of ROI — Walk away if it doesn’t work
We’d built a real-time sales intelligence engine for B2B companies. The idea was simple: watch who opens your emails, clicks your links, and visits your website, score their behavior over time, and the moment they cross your hot-lead threshold, create a deal in their CRM and fire an instant Slack notification to the sales rep: this person is warm, call them now.
The system worked. Until it didn’t.
The Problem: A Pipeline Full of Ghosts
Within the first few weeks of running our signal engine for a logistics company with over 100,000 email subscribers, the alerts started firing. Deals were being created. Notifications were going out. The system was working.
Except when the sales team went to make calls, the “hot leads” were gibberish. Deals with names like “Cebael Sonictoolsusa” and “Muhagfe Rohmsemiconductor.” Real company domains, sure. But zero actual human beings behind them.
Sixteen junk deals created in a single week. The pipeline was contaminated.
Here’s what was actually happening: bots. Not the kind that try to hack your server — the kind embedded in corporate email security stacks that automatically scan every link in every email your subscribers receive to check for phishing and malware. Microsoft Defender, Proofpoint, Mimecast, Barracuda. When these systems scan an email, they click every link. They open it. They follow redirects. They look like a very enthusiastic human being.
Your scoring system, which was built to reward engagement, rewards them.
You’re Not Alone — HubSpot and Salesforce Have Been Fighting This for Years
This isn’t a niche problem. HubSpot’s own community forums are full of marketing teams reporting that over 50% of their email engagement data is bot-generated. One user wrote that they had to strip opens and clicks entirely out of their lead scoring model because the noise was so bad it made the scores meaningless.
HubSpot’s own bot filtering documentation acknowledges what they call a “game of Whac-a-Mole” — the moment you block one bot behavior pattern, the security scanners update and find a new way around it. HubSpot evaluates interactions against behavioral patterns rather than static IP lists because the IP lists change constantly.
Salesforce Einstein, their AI-powered lead scoring engine, flags the same issue: scoring based on email opens and clicks is “very unreliable data.” Their recommended alternative? Weight web page visits more heavily — because bots don’t usually stick around on pages long enough to count as real dwell time. But as we learned, even that isn’t enough on its own.
The people building the biggest sales intelligence systems in the world are actively fighting the same thing you’re dealing with in your pipeline.
Our Journey: Four Layers of Defense, Iterated in Real Time
We didn’t solve this in one shot. Here’s how the filter evolved, layer by layer.
Layer 1: Burst Scanner Detection
The first generation of bots we encountered were obvious: same email address triggering 3+ identical events (all opens, or all clicks) within a 5-minute window. Classic email security scanner behavior — they open the email and click every link in rapid succession.
We added an individual scanner detection rule: if a contact fires 3+ events of the same type within 5 minutes, mark them as a bot and never score them again. Then we added a cluster rule: if 10 or more contacts from the same list fire the same event type in the same 5-minute window, mark the whole cluster.
This caught a lot of bots. Not all of them.
Layer 2: Pixel Dwell-Time Gate
The smarter bots — the ones embedded in corporate security stacks — don’t just scan emails. They follow links to the actual landing page and check the website too. Our pixel (a lightweight tracking script installed on the client’s website) was picking these up as “website visits” and scoring them.
We added a dwell-time gate to the pixel: a visit only counts if the browser stays on the page for at least 8 seconds and doesn’t appear to be a headless crawler. Bots following redirect chains rarely stick around. This helped significantly on the website visit side.
Layer 3: Tightening the CRM Push Filter
Even with dwell-time gating, some bot traffic was still making it through — visits spread across multiple days, accumulating enough score to trigger deal creation. We tightened the CRM push filter to require either multiple email clicks or multiple verified pixel visits with a time spread of more than 4 hours between first and last activity. Pure single-session activity got filtered out.
Better. Still not clean.
Layer 4: The Final Fix — Email Engagement as a Deal Creation Gate
When we pulled the database records for all 16 bot deals created in a recent week, we found a pattern we hadn’t expected:
clicks_count: 0 | opens_count: 0 | visits_count: 3-7
Every single bot contact had zero email engagement. No opens. No clicks. They were scoring entirely from pixel visits that accumulated over days — website crawlers, security scanners following links, automated browsers. They passed the time-spread filter. They passed the dwell-time gate. But they had never once opened or clicked a real email.
The fix was obvious in hindsight, and it’s the same principle HubSpot calls “identity-anchored engagement” and Salesforce builds into their multi-signal confirmation requirement:
A contact must show at least one real email interaction — an open or a click — before they qualify to create a deal in your pipeline.
One line added to the database query that gates deal creation:
AND (opens_count > 0 OR clicks_count > 0)
Website visits still count toward scoring. Pixel tracking still runs. Contacts that only visit your site still get identified and tracked. But they don’t create Pipedrive deals or fire sales alerts until they’ve engaged with your email — proving they’re a real person who has actually received and interacted with your communication.
Why This Works (And Why It’s the Industry Standard)
Think about what real email engagement actually proves:
- The email reached a real inbox. Spam traps and bot accounts don’t open emails — they accept delivery and discard them.
- A human made a deliberate choice. Opening an email or clicking a link requires intent. Security scanners click links automatically before the human even sees the message. A human-triggered click comes later, from a different IP, with different behavioral patterns.
- The contact has a real relationship with your brand. Someone who opened your last three campaigns is not a ghost. They know who you are.
This is why 6sense, Bombora, and the other intent data platforms that power enterprise sales intelligence require “identity anchoring” — they only attribute anonymous website behavior to a known contact if there’s a prior email or form interaction on record. They discovered years ago that website visits alone are too noisy to be actionable.
Our Watch Tower dashboard lets clients see exactly who is engaging, what their score is, and how the filters are working in real time. No black box. Full transparency into every lead that makes it into the pipeline — and every one that doesn’t. We’ve now implemented the same standard as the enterprise platforms, in a system small and mid-size B2B companies can actually afford.
The Result: A Pipeline You Can Trust
After implementing all four layers — burst detection, dwell-time gating, visit spread requirements, and the email engagement gate — our signal engine now runs close to zero bot contamination. Deals created in Pipedrive represent contacts who have:
- Opened at least one email (proven inbox delivery, human receipt)
- Crossed a meaningful engagement score threshold (not just one click — a pattern of behavior over time)
- Shown up on the website with genuine dwell time (not a scanner following a redirect)
When a deal is created and a Slack notification fires saying “Call NOW — Score 25 — 3 email clicks, 4 site visits over 11 days,” your sales rep can pick up the phone with confidence. That’s a real person. They know who you are. They’re warm.
What This Means for Your Sales Team
Noise is the enemy of sales momentum. When your reps learn to ignore pipeline notifications because half of them are garbage, you lose the speed advantage that makes real-time lead scoring valuable in the first place. The whole point of a sales intelligence system is that when it fires, your team moves immediately.
That only works if the signal is clean.
We built this system the hard way — by watching it fail in production, diagnosing each failure mode in the live data, and adding progressively smarter filters until the noise was gone. The result is a standard that matches what the biggest sales platforms in the world have arrived at through their own experience: email engagement is the identity anchor. Website behavior is the amplifier. Both signals together, over time, identify real human buyers with real intent.
Here’s the Offer. No Catch.
Your email list already contains people who want to buy from you. (Don’t have a list yet? We can build one.) They’re opening your campaigns. Clicking your links. Visiting your site. You just don’t know who they are — and right now, nobody is calling them.
We’ll install our signal engine on your email list and website, connect it to your CRM (we handle setup end-to-end), and in 30 days you’ll have a ranked list of your hottest prospects — the real ones — showing up in your pipeline automatically, with a Slack or email alert the moment they cross the threshold.
You don’t pay for setup. You don’t pay until you see it working.
If it doesn’t generate at least one legitimate hot-lead opportunity you didn’t have before, you owe us nothing.
That’s it. No discovery call theater. No 12-slide deck about our “methodology.” You tell us your list size and your CRM, we get it running, and you start calling real people who already know who you are.
Most of our clients book their first legitimate meeting from a signal-engine alert within the first two weeks.
The question isn’t whether your list has buyers in it. It does.
The question is whether you have a system to find them before your competitors do.
Sources referenced:
– HubSpot Community: Are Bots Affecting Your Email?
– Bot Filtering in HubSpot Marketing Email Analytics — Insidea
– Salesforce Einstein Behavior and Lead Scoring Overview — Trailhead
– Salesforce Einstein Lead Scoring Guide — Peeklogic