Too Many Signals & All Our (Intent Signal) Knives are Dull af
It's all fun and games until you have too many signals, how do we cut thru the noise?
In this substack
Why signal prioritization matters in scaled outbound – The mistakes I made at Unify sending 1.4M+ emails – How we fixed issues with reply rates, fatigue, and targeting – The future of outbound systems
Scaling Automated Outbound as a Channel
Running Automated Outbound is a lot like going to the gym - you see initial traction relatively quickly and then you need to hunker down and begin to get methodical to see incremental progress and momentum. In the 13 months at Unify, my work had sent north of 1.4 million automated emails via the Unify platform. Without a comprehensive tool like Unify (or other solutions), a feat like this is rarely possible.
If you were to send each one of these emails, manually, one after the other with no sleep, eating, or breaks with just a 10 minute delay between each send, this would take one person roughly 9,700 days to accomplish.
The pipeline the growth team and I generated while at Unify, using Unify was nothing short of spectacular. In my 13 months at Unify, we generated north of $12 million dollars in pipeline. This pipeline came from a substantial number of intent signals and plays that we built using Unify, integrated with other tools, and more - which gave us a diverse set of sources through which we booked meetings from. Every day, we ran 40+ plays, which funneled into 100+ sequences and the list continues to grow.
To name just a few of our plays:
Website visits (normal website visits, UTM / paid traffic retargeting, individual intent, content downloads, etc)
Linkedin engagement (followers, likers, commenters, profile viewers)
Email engagement (link clicks, replies, sentiment, etc.)
G2 intent
New hires & champion tracking, Promotions
AI agent qualification with infinity signal for recurring account research for
CRM-based plays (Closed Lost Revisit, Follow ups, etc.)
Fundraised
Website Tags
RSS feeds (the hidden gem of outbound IMHO)
and much, much more
We’ll walk through 2 chapters
The roadblocks I hit as we scaled automated outbound
What is intent signal prioritization and how to account for it
Unify’s Automated Outbound Roadblocks
Hitting the breaking point
Over the first 9 months as we scaled our automated outbound program, we rarely ran into problems - it was common for us to easily accommodate our email backlog from the email send capacity we had. Additionally, each week we saw a diverse selection of meetings being generated from outbound plays, with little / no need to focus on optimizing for any particular one.
After all, if it ain’t broke - dont fix it, right?
Unfortunately, before we knew it, we had fallen into a trap of our own doing. As the program continued to grow and we became more bullish on email, we actually ran into a few fascinating problems. Notably:
How do you know when to prioritize a signal at the person level or a signal at the company level? Moreover, all things being equal, how does your outbound system prioritize what to write in an email if more than one signal is present for that person?
These things tend not to affect smaller automated outbound motions - but for us reaching this point was inevitable. The learnings I share below will hopefully help you build your framework for prioritizing signals and avoid the same mistakes I made!
When (and how) I realized I screwed up
On Monday February 10 (coincidentally my birthday), we had our best day ever at Unify. Just Unify’s automated outbound program alone booked a large number of high quality meetings with top prospects at target companies. As we continued to ship new signals and product improvements, it was impossible to imagine we would hit a huge roadblock. Due to the success of the program, I began to take a number of risks all at once without trying each one incrementally. A few of the things that I did were:
I removed booking links from our automated email signatures in the hopes of improving deliverability even further.
I A/B tested across website views sequences by having no links in any email outside of the opt-out link
With our public launch of AI agents and the infinity signal, I moved many of our recurring scrapes that occurred outside of Unify, directly onto the Unify platform - including signals like
Recent or Future Event mentions on website
Hiring for certain jobs via job listings
Recent or Future Webinar
Recent of Future Earning Announcement
In the same light with our launch of email classifications and enrollment results in Audience builder, I began to build sequences for new permutations of email enrollment results
What were the consequences of our massive outbound program?
The sheer number of new agent scraping we had running caused us to generate a lengthy email backlog
The large number of signals we were monitoring on every company meant we were running out of people to email (lol)
I (at the time) was not monitoring reply rates by mailbox to determine which domains and mailboxes were showing signs of fatigue and poorer deliverability
I didn’t prioritize dashboards to manage results and progress in real time
Not prioritizing person level signals versus Company level signals
1. Removing Booking Links in the Email Signature
Lesson: While it is widely considered problematic to include links in emails (especially if they differ from the domain you are sending from), providing a booking link actually makes it far more convenient for prospects to book time with you. If you don’t prospects will have to respond to an email, work thru back and forth of scheduling, and more. This significantly reduces the barrier to prospects learning more about your solution.
While no show rates may be higher and conversion on a opportunities without an actual email thread may be lower, this resulted in a drop in meeting generation.
Learning: I added back signatures in our emails to lower the barrier to booking a meeting for prospects.
2. A/B Testing for Other Links in Emails
Lesson: Similar to the above, removing links to unifygtm.com from our emails meant that prospects could not easily access the website and do their own research and discovery. While keeping emails light on links / HTML is key for deliverability, I believe the sacrifice in deliverability is occasionally worth it so that prospects can easily learn more about the product you might be pitching.
3. AI Agents for Every (Lower Converting) Company Signal
Lesson: Launching AI agents was a huge milestone for Unify - giving our customers the ability to qualify or disqualify any number of companies (at scale) based on criteria they set that is both dynamic and intelligent. However, what I failed to realize initially is that just because I could now do this, that didn’t mean that I should. Of course the team wanted to prioritize more signals, more emails, more meetings, more pipeline, but we didn’t think through the process of what could occur.
Within days we had enrolled thousands of peoples into outbound sequences that were company-level signals, and our composition of enrollments moved from predominantly person-led signals to company-led signals, which are understandably lower-converting.
This makes sense because if we discover something about an individual person, this is likely a better signal than a signal in relation to the company they work for. This over enrollment of people into lower converting signals had a number of downstream consequences
The people we emailed could not be emailed for other intent signals that may be of higher quality because they were in this sequence already
The people we emailed could not be enrolled into other sequences until their cooldown was completed. On Unify you can set your own cooldown period (a timeframe under which someone could not be emailed again) - we recommend 30 days
Lower quality signals result in higher opt-out rates or likelihood to opt-out, meaning you can’t email them again.
Additionally, based on high level cuts of data, person-level signals tend to have much higher reply rates and booked meeting rates when compared to company-level signals. Empirically, we have observed as much as a 2X+ improvement in reply rates for person level signals over company level ones.
Interestingly, we also track peoples clicks and opens in old emails (what we call email engagement). Since we had enabled users to filter audiences based on how people engaged with emails, I decided to email those who had never previously responded to an email with a highly customized email to get them to re-engage. While this felt like a smart idea initially, it is generally not a good idea to do this as they are unlikely to engage if they haven’t engaged already!
Learning: We quickly realized we needed to reevaluate which company-level signals were worth automating and which to hand to BDRs.
4. Creating a Lengthy Backlog from Too Many Signals
Lesson: Additionally with the release of AI agents, we now had the ability to monitor for a near infinite number of signals. This meant that we began enrolling thousands of people into sequences, creating a tremendous backlog of emails to be sent. At Unify, we recommended each mailbox sends no more than 25 emails per day. If we were to generate 200 emails for one mailbox in one day, this would take roughly 8 business days to work through the backlog assuming no new enrollments are made and excluding the affect of more messages associated with those enrollments (email 2, 3, etc.)! This is extremely problematic, as certain signals need to be prioritized sooner - you don’t want a signal to go stale! The last thing you want is an email for someone visiting your pricing page to go out several days after an email for someone who was recently hired at a new company!
Learning: We moved to create mailbox groupings based on the level of the intent, and dynamically calculated the sending volumes we would need for each one:
Instant Send - a mailbox grouping used for the highest intent signals like people viewing pricing pages or product demos, people who fail to book meetings, and more
High Intent - people who view other high intent website pages, view or engage with Linkedin content, etc.
Medium Intent - people who engage with us or competitors on G2 or on Linkedin, regular website views, and others.
Low Intent - used for cold or mostly cold emails
5. Running Out of People to Email
Lesson: If you try to email everyone, you will. And then there will be no one left to email! This comes back to some of the earlier points, but because of the sheer number of signals we were emailing for, we began to return zero people from prospecting because we either couldn’t find additional people at a company or the people we did find were in other sequences.
Learning: Again, this forced us to re-evaluate which company level intent signals are most important to us, which ones we should outbound based off of, and how many people to email for each one!
5. Not Monitoring Mailbox and Domain Reply Rates
Lesson: In any large scale outbound motion, it is critical to monitor reply rates as a heuristic for performance and thus deliverability. At the time, we were using one massive round-robin to enroll people into sequences. We did not have any focus on the reply rates of individual mailboxes, domains, and more. This type of setup causes the poor deliverability of mailboxes to compound and continue to worsen.
Learning: We moved to monitor the reply rates of every mailbox being used on a regular 2 week cadence. This had allowed us to reactively pause sending for lower performing mailboxes and allow them to recover, and proactively move mailboxes that are trending in the wrong direction into warming or a pause state to allow them to improve.
6. Metrics, Dashboards & Monitoring
Lesson: When I joined the startup world, I found it hilarious how sales people hated updating and using CRM / Salesforce. They told me it’s so much easier to function based off gut and intuition. My version of not updating Salesforce was creating and monitoring dashboards day-in and day-out. On a small scale, its easy to act based on intuition or as I like to say a finger in the wind, but inevitably, things quickly begin to slip thru the cracks. Every team needs a comprehensive way to monitor successes, but most importantly failures.
Learning: Throughout March and April of this year, I spent a considerable amount of time writing SQL queries to monitor all of our outbound metrics in close to real time and ensure that we had an active pulse on what was going on. Much of this effort was what led me to build Galaxy.
Intent Signal Prioritization
What is intent signal prioritization?
Intent signal prioritization is the ability to dynamically stack rank signals based on the likelihood that outbounding based on that signal will convert into a meeting, a qualified opportunity, and hopefully a closed-won deal. In the early days of outbound (and because email is such a low CAC channel), prioritization is typically not a top priority for GTM engineers - early outbound automation is focused on a) identifying the relevant signals and b) building process for outbounding based on these signals consistently. As the scale of your outbound systems grow however; you will inevitably encounter this problem. This happens for a few reasons:
Sending Constraints
You have a constant number of mailboxes with a limited number of email sends per day
You are running an increasing number of outbound experiments which reduces your available send capacity
You must be careful how many emails you send and which emails you send so as to not hurt email deliverability
Generating a high backlog means intent signals will go stale
Company Constraints
There are a limited number of people you can email at any given company
Intent Signal Constraints
Intent signals tend to occur at totally different times
The conversion rate from any two signals may differ substantially (someone taking a meeting that originated from an offer for a Gift Card will likely convert substantially lower than a meeting that is booked from a Competitor teardown campaign).
Before you know it, you went from being able to email freely with no constraints on how or why you outbound, to needing to optimize for the the best signal. Without a focus on prioritization, you may actually email people for lower converting signals, and this caused us to re-evaluate the framework through which we conduct our outbound motion, with strategy, timing, and more at its core. As detailed above, we faced this problem firsthand in March of 2025!
The importance of signal prioritization
In the world of outbound at scale, intent signal prioritization is the name of the game. To start small, let’s pretend we are the GTM engineer for Salt.com and Alan (who works at Pepperman.com 🫑) is a prospect that we want to book a meeting with. With Unify, we can surface a large number of intent signals from Alan at Pepperman.com. Some of these signals might be:
Has Alan visited our website
Has Alan visited high intent pages on our website (pricing page, sign up for a demo, our integration docs)
Has Alan downloaded a piece of content
Has Alan checked out our product walkthrough
Is Peperman.com hiring for GTM engineers
Has Alan or Pepperman.com recently hired a GTM engineer
Are target personas at Pepperman engaging with my Linkedin page by liking, commenting, or viewing my profile
Has Pepperman.com been mentioned in the news
Without an automated outbound solution, GTM teams with this number of signals typically resort to one of a few strategies, all of which are not optimal. This may be in the form of BDRs just outbounding based off of what they see first, drip campaigns for all of them, and more. Typically none of these are the right solution…
Company vs Person Level Signals
How do you know when to prioritize a signal at the person level or a signal at the company level?
As we dug deep into the data, we had a few interesting findings, that our composition of sequence enrollments shifted from predominantly person-level intent based to company-level intent based. What’s more is that we found historically that company-level intent based sequences showed roughly 2X worse reply rates than person-level intent based sequences. These were two fascinating realizations that informed how we reworked our process for running outbound and how to optimize it.
How do other platforms prioritize signals?
There’s a wide spectrum of outbound tools on the market. In relation to signal prioritization, I wanted to take a moment to delineate what each one of them does, and how we differentiate from them
Email Tools
In email tools, there simply is no signal prioritization, meaning you have to manually choose between one signal or another in all cases. This is fairly common for tools like Apollo, Outreach, Salesloft, and many more. You can drop someone into a cold email sequence for this, or for that.
Warm Outbound
In warm outbound tools, the platforms typically do not even have enough intent signals to actually justify any type of signal prioritization or framework. They tend to just involve website intent, and so there is no need to do any type of prioritization
Intent Signal Aggregators
intent tools provide you tons of intent signals but no comprehensive way to score / optimize for which ones to email off of, especially since the outbound (and thus conversion) does not live in the tool.
Clay and other Lead Workflow Tools
While they don’t want to admit it, the majority of people using Clay are actually using it for cold outbound with account research that makes that outreach slightly warmer. With a few exceptions, prospects that are emailed from Clay haven’t necessarily shown interest in your product so there isn’t a massive prioritization component that is critical to success.
However, for the minority of people, in order to achieve some form of signal prioritization, you need to build one table that has dozens of signals integrated to it across hundreds of columns. To do this requires building a comprehensive scoring system for each signal, and a waterfall of signals to determine which signal will be the one that is emailed about. This is extremely difficult to build (and maintain) and even more difficult for most people to wrap their heads around. More importantly, a complex Clay table like this creates a ton of tech debt and key person risk associated with whoever built it.
Additional Intent Signal Optimization Problems
There are a handful of criteria that make this an objectively difficult set of problems to solve. Lets look at few of the ones that were not covered in the Unify Roadblocks section above, with a few examples. Let’s again pretend that we are the GTM engineer for Salt.com and Alan (from Pepperman.com) is a prospect that we want to email.
Consecutive actions
In the case of Alan interacting with Salt.com (our hypothetical website), there are a number of user flows that might occur
Alan simply visits the website and then leaves
and then Alan downloads a piece of content or an ebook (like this one)!
and then Alan visits the product walkthrough (we use Navattic for this)
and then Alan tries to book a meeting but doesn’t (we use Default for this)
and then Alan successfully books a meeting and an opportunity is created in CRM.
Each one of these outcomes alone, requires its own unique email campaign.
But, if they all happen one after the other (which is fairly common!), we need to wait to make sure to choose the highest intent signal to email for!
This could look like:
If someone just visits our website we wait to see if they did anything else
If someone downloads a piece of content we wait to see if they did anything else
If someone tries to book a meeting but doesn’t we wait to see if they were able to book a meeting
If someone books a meeting and an opportunity is generated, we shouldnt email them (asking them to)!
This type of logic is not present in other outbounding tools, but with Unify’s Play builder, you can easily use logic workflows and time delays to account for the actions that are relevant to you and how to prioritize them. This means that using Unify you can make sure that consecutive actions like these are accounted for and that this prospect will be emailed for the greatest intent signal that is displayed by Mr Alan from Pepperman.com
Timing
Signals never present themselves all at the same time. Let me show you what I mean. Alan from Pepper.com is not capable of doing all of these things at the same time:
viewing our website
viewing our product demo
tries to book a meeting
likes my Linkedin post
moves jobs
Rather, these things tend to happen one after the other, in no particular order:
Alan from Salt.com views the Pepperman.com website on Monday at 11AM.
Alan views the product demo at 2PM on Tuesday Play Prioritization - EBook 16
Alan tries to book a meeting but doesn’t on Tuesday at 6PM.
Alan likes my LinkedIn post at 6PM on Friday.
If you took a snapshot of Alan’s activity at a certain point in time, say Tuesday at 1:30PM, all you would see is that he viewed our website, even though he is going to view the product demo in <30 minutes! In an optimal scenario, would it make sense to email him the second that intent is shown, or wait to see if other intent appears?
The future of intent signal prioritization
The future of automated outbound systems need to account for this type of chronological but disparate sources of intent to ensure you are emailing someone for the best signal across that time frame, and take into account any type of predicted behavior that may occur in the near term!