Parts 1 and 2 covered the signal-based model and the signals themselves. This part is about execution: the volume and depth math, how channels actually divide the work, why programs really fail, and how to run the system as an organization.
Volume and depth: the two numbers that govern outcomes
Signal-based does not mean low volume. It means the volume is warmer. Two numbers decide whether a program produces pipeline:
- Volume: enough at-bats. Programs need roughly 5,000+ leads flowing through them for the math to compound. Thin tracking lists and over-filtered audiences starve a program before it can prove anything.
- Depth: 5+ activities per prospect. One email is not an attempt; it is a formality. Customers who work at least 1,500 prospects to 5 or more activities each generate roughly 650% more revenue and 1,200% more pipeline than those who do not. The benchmark to manage toward is 8 to 10 median activities per actioned lead.
Volume without depth is spam. Depth without volume is a hobby. Working programs do both.
Channels: each one has a different job
The channel mix works when you assign each channel the job it is actually good at:
- Email is your awareness channel. It scales, it carries the signal-based personalization, and it puts something true and relevant in front of the buyer. Judge it as awareness, not as a direct conversion machine: a relevant email that never gets a reply still registered an impression with a buyer who now knows you exist and why you are relevant. Buyers research independently and increasingly consult AI before ever replying; your email's job is to get on the radar and break through the clutter.
- Phone drives the highest reply rates. Nothing converts attention into a conversation like a live call, especially one that references the signal ("saw you just stepped into the VP role..."). The phone is where awareness becomes pipeline.
- LinkedIn adds the human layer. A face, a mutual connection, social proof around the same moment.
The sequencing research backs this division of labor. Salesloft's analysis of 3.4 million cadences (with TOPO's sales development benchmark) found that 80% of the top 100 performing cadences opened with a call followed by an email, and that the best cadences front-load their touches into the first 10 business days before spacing out. Salesloft's multi-channel research similarly found that sequences combining email, phone, and social touches see significantly higher engagement than single-channel approaches. Translation: lead with the phone, reinforce with email, concentrate effort early while the signal is fresh.
Then there is the multiplier: coordinated coincidence. When media and outbound are aligned, the buyer sees your ad this week, gets a relevant email the same week, and hears from a rep days later. To them it feels like you are suddenly everywhere at exactly the right moment. That is not luck; it is the same signal firing your ads sync and your sequence together. Outbound lands dramatically better on an account that marketing has already warmed, which is why the strongest programs run sales and marketing off one signal engine instead of two calendars.
And one non-negotiable under all of it: deliverability. Cold outbound on secondary domains, 50 to 150 sends per mailbox per day, SPF, DKIM, and DMARC on every sending domain, bounce rate under 5%. The best messaging in the world converts nothing from a spam folder.
The real reason outbound fails: tasks don't get done
Here is the uncomfortable truth, and it is rarely the data, the tool, or the messaging. Mark Kosoglow, CRO at Docebo and the operator who built the entire outbound motion at Outreach, put it bluntly on stage at a UserGems event: "The number one reason outbound fails is because reps don't do their tasks!" Signals are only as effective as how well they are actioned.
The math from earlier is unforgiving, and independent research says the follow-through gap is the norm, not the exception: RAIN Group's research on sales follow-up found that 80% of sales require five or more follow-ups, yet 44% of salespeople give up after just one. The standard is 5+ touches on every lead the system surfaces. What actually happens in struggling programs: tasks sit overdue, sequences get skipped, and leads receive 1 to 3 touches before being marked dead. Every GTM motion fails by those standards, signal-based or not. A perfect signal, resolved to a verified buyer, with a drafted email, is worth nothing if the sequence never runs its course.
This is why execution telemetry belongs in your weekly review right next to pipeline: overdue task counts, median touches per lead, sequence completion rates. And it is why automation of enrollment and drafting matters so much: every step a rep does not have to do manually is a step that cannot be skipped.
Messaging: train the brain, don't chase perfection
The most predictable friction in any AI-assisted program is messaging defensiveness. Reps take pride in their words, as they should; personalizing is their craft. So the first AI drafts get picked apart, and teams stall for weeks polishing prompts before sending anything.
Reframe it. AI messaging is not about perfection out of the gate: the brain needs to be trained, and the trajectory is the point. Early on, reps should edit; their edits and feedback are what teach the system. The goal is that reps need to do less and less customizing over time, not that draft one is flawless. And keep the denominator honest: even before a single message is polished, finding the right people from signals, backing them with reliable data, serving them up prioritized, and drafting sequences that are 80% of the way there saves hours per rep per week. Do not sink that dividend into chasing the last 20% by hand on every email. Buyers are not grading your prose; they are deciding in seconds whether the message is relevant to them.
Measure what matters
Traditional metrics mislead in a signal-based motion: open rates are inflated by mail privacy proxies, clicks measure curiosity, and activity volume rewards busywork. The framework that works:
- Reply rate, the clearest resonance signal: how do reply rates compare to other cold outbound you've run?
- Signal-to-opportunity conversion, tracked by signal type, the most important number in the system: which signals actually become deals.
- Capacity per rep: qualified conversations, not emails sent.
- Process time from signal to coordinated action: every day of delay is a day a competitor can get there first.
- Pipeline per signal type, connecting signals to dollars so you know where to invest next.
- Execution telemetry: median touches per lead and task completion, because a program can only perform as well as it is actioned.
Close the loop: log signal, outreach, response, and outcome, and feed results back into scoring so the system learns from your history.
Attribution: nitpicking credit is a waste of time
Attribution deserves its own fundamental, because more GTM energy is wasted litigating it than almost anything else. Start from the structural facts:
- The models disagree by design. First-touch, last-touch, and multi-touch produce different answers from the same data, and none is objectively correct. Reconciling them is not analysis; it is arguing about which fiction to prefer. (See UserGems Attribution for the full breakdown.)
- Most of the journey is invisible anyway. Gartner's research puts 70 to 80% of B2B buying activity before the buyer ever engages a sales rep, in channels no attribution software instruments: peer conversations, private Slack communities, podcasts, and increasingly AI assistants. Chris Walker of Refine Labs, who popularized the "dark funnel" concept, has argued for years that attribution software structurally cannot measure communities, word of mouth, or podcasts, and that the honest supplement is simply asking buyers how they heard about you.
- Even the tracked portion is wrong. Refine Labs measured a roughly 90-percentage-point gap between what attribution software reported and what customers self-reported for web search alone, across a 12-month, 620-conversion study. And a Demand Gen Report 2025 survey found only 21% of B2B marketers say they can measure marketing ROI with confidence. The tools are precise; they are not accurate.
- Deals close from shortlists built early. 6sense's research found 95% of B2B deals close with a vendor that was already on the buyer's day-one shortlist. The work that gets you onto that shortlist (relevant, signal-timed awareness) is exactly the work attribution software cannot credit. This is why judging signal-based email on reply rate alone undercounts it: an unanswered relevant email is shortlist-building the model will never see.
The conclusion is not to abandon measurement; it is to measure the right thing and stop polishing credit splits. Anchor on one question: are the leads the program creates turning into new pipeline and revenue? UserGems operationalizes that with three CRM-based buckets: Sourced (a UserGems contact created, contacted, and added as the first opportunity contact), Assisted (added to an open opp that already had contacts), and Suspected (5+ activities on an account with an open opp, first activity before the opp existed). Sourced and Assisted are conservative by design; review Suspected regularly, because those contacts often belong on the opp and were never added. Answer the pipeline question, review the buckets, and put the reclaimed hours into inputs and follow-through, which is where results actually move.
Run it as a system, not a side project
The last fundamental is organizational. A GTM engine cannot be siloed to a single person: it touches sales, marketing, ops, and CS, and it feeds the entire business, so it needs the entire business feeding it back. Executives need to stay involved, and alignment needs to happen continuously, not once at kickoff: which signals matter, what the swim lanes are, how reps are held to the actioning standard.
Can you set and forget the system and still see positive results? Yes. Will that be the best version of your program? Absolutely not. GTM operations are about experimenting and iterating: test a new signal, refine a persona, tune a sequence, review what converted, repeat. Perfection is the enemy of progress, and at the speed most sectors are moving, over-engineering the program before launch, or leaving it siloed with one owner, is a recipe for disaster. Ship the reasonable version, watch the numbers, and improve it every month. That habit, more than any single configuration, is what separates programs that print pipeline from programs that stall.
Sources: Measuring signal-based outbound: the metrics that matter and Why AI outbound is failing: the volume-first trap, UserGems Blog; UserGems customer benchmark data on engagement depth, deliverability, and program volume. Quote from Mark Kosoglow, CRO at Docebo and former SVP Sales at Outreach, speaking at a UserGems event. External research cited inline: Salesloft cadence analysis (3.4M cadences, with the TOPO Sales Development Benchmark); Salesloft multi-channel engagement research; RAIN Group sales follow-up research; Gartner B2B buying research; Chris Walker / Refine Labs dark funnel and self-reported attribution work; Demand Gen Report 2025 survey; 6sense buyer research. See also UserGems Attribution.