I built the system that ran my team’s GTM busywork. Then I handed it to agents.
My best closer spent a Tuesday morning building account briefs by hand, twenty minutes an account. By lunch he had researched six companies and talked to none. So I stopped making the busywork faster and deleted it. Here is the whole system, end to end, and the version that embarrassed me first.
a forecast built on a Deal Confidence Score, not a guess
20 min → 0
the account brief, written before the rep asks
60+
self-updating reports, so someone else can run it
source to sent
sourcing, scoring, enrichment, sequence
01 · SOLVE · DRAW THE LINE AROUND THE PROBLEM
Most of go-to-market is five jobs, done by hand, forever.
Pick the account, enrich it, score it, write the brief, guess the forecast. Every one ran by hand, and three were broken. Account selection scored old wins that don’t repeat, so it pointed good reps at logos that were never going to buy. The forecast sat at 65 percent, one call in three wrong, held up by whoever said their number the loudest. And the whole motion only moved when someone pushed it. A motion that only moves when someone pushes it is not a system. It is a person with a to-do list.
THE REFRAME
Not a GTM Engineer to hire. A workflow to build.
A headcount is a delay. A workflow is buildable. That one swap is the unlock.
The false starts, because they are the point. First I tried to make the busywork faster: better templates, a tighter cadence. A twenty-minute task on a nicer template is still twenty minutes, and it still stops when the person stops. Then I rebuilt the scoring model and it embarrassed me. It scored big companies high and small companies low, and when I pulled it apart it had simply learned that big companies are big. It was size, wearing fit’s clothes. That was the lesson that changed everything after it: never trust a score you cannot explain.
02 · THE ICP AND THE SIGNALS
Score on data reps trust, with the reasons attached.
I threw out the history and rebuilt the score on ICP-only data, the accounts that actually look like our wins, not every closed-won in the table. Octave is the ICP brain behind that: who fits, and the positioning for why. Then I split the score into three signals and kept them apart, so one could not hide behind another.
Fit
Does the account actually look like a win, on ICP-only data, not every closed-won in the table.
Expansion potential
Room to grow inside the account, kept separate so it can't hide behind fit.
Product usage
Usage that is actually happening. The loudest, most honest signal in the model.
A black box gets ignored the first time it’s wrong. A score a rep can audit is a score a rep will run.
So every score carries a Signal Library: open an account, see exactly why it scored. That is the difference between a model people use and a model people override.
INSIDE THE SCORE
THE MODEL, ADDITIVE, NOT A WEIGHTED AVERAGE
LAYER 140-70 base
Fit qualifies
In-ICP accounts get a base of 40 to 70. Below the bar, you do not score.
LAYER 2loudest
Product is the engine
Real usage is the loudest signal in the model. Power adoption alone adds +18.
LAYER 3cap +20
Buying signals boost
Hiring, funding, competitor and stack signals are capped add-ons. They only ever help.
QUALIFY ON ICP + PRODUCT · TIER ON INTENT
WHAT PREDICTS A WIN, SIGNAL BOOST
Product adoption (Power) · loudest signal+18
New VP Sales / CRO hire+6
Funding round+6
Competitor tool in stack+5
CRM in stack · 1.64x win-lift+3
Decision-maker found+2
BUYING SIGNALS ONLY EVER ADD · CAP +20
03 · STACK · BUY THE PLUMBING, BUILD THE EDGE
Who found the signal, and what I built on top.
Most of this is tooling a growth-stage GTM org already pays for. The work was not buying more tools. It was wiring the ones I had into something that runs itself. Buy the plumbing. Build the edge.
Octave
The ICP brain. Who actually looks like a win, and the positioning behind why.
Salesforce
System of record. The truth lives here.
Snowflake
The data layer. Where history and signal get centralized.
Common Room
Product and community signal.
Deepline
Enrichment and signal routing.
FullEnrich
Verified email and cell for the buying committee.
Claude agents
The autonomous layer that scores, enriches, and writes the briefs.
THE METHODOLOGY, IN THE OPEN
Composite Scoring v7
VIBECODED
Composite Scoring v7
The channel-aware, evidence-backed scoring methodology behind every account.
How Your Leads Are Scored
VIBECODED
How Your Leads Are Scored
The rep-facing signal stack: one number, three layers.
2026 Master ICP Profile
VIBECODED
2026 Master ICP Profile
Who we sell to, built from 12 months of won and lost data.
AND THE TOOLS I PUT IN REPS’ HANDS
Sequence Lab
VIBECODED
Sequence Lab
Drop a domain, get a ready-to-send sequence, pre-filled from the site.
Signal Playbook Builder
VIBECODED
Signal Playbook Builder
Describe a prospect; the AI picks the signal stack and writes the follow-ups.
Proposal Intelligence
VIBECODED
Proposal Intelligence
Wrap a proposal in a tracked viewer; know the moment a prospect is ready.
GTM University
VIBECODED
GTM University
Structured learning tracks and AI roleplay coaching. 150+ modules.
04 · SPLIT · CUT THE DRAG, KEEP THE JUDGMENT
The machine runs the execution. The people keep the call.
I packaged the enrichment, the scoring, and the brief engine as autonomous agents. They run on their own, so the system stopped being something I operate and became something that operates. The brief that took my closer twenty minutes now exists before he asks. But the judgment stays with the rep: they audit the score, work the account, and attest the part of the forecast the machine cannot see.
THE FORECAST, WITH A NUMBER BEHIND IT
The forecast stopped being a vibe. Part is automated from signal: activity, engagement, product usage, the stage math. Part is attested by the rep, because the human knows things the system cannot see. Together they make a number you can defend in a room, line by line.
A forecast you can’t explain is a guess in a nicer font.
Salesforce history + sequencer engagement + Gong calls. This is not a cold account.
SALESFORCEClosed Lost, chose the incumbent2022
SALESFORCEClosed Lost, no budget and wrong timing2024
MIXMAX6 emails, 2 meetings over 3 years, last touch 8mo agoongoing
GONGLast call: revisit when we replatform. They just did.call
Two prior swings, both lost on timing and the incumbent. Not a first attempt. Lead with what changed since 2024, don’t pitch from zero.
THE PLAY · YOUR FIRST MOVE
Lead with the sequences gap and what changed since 2024. Open on the 3 power users by workflow; the new VP plus fresh funding is the mandate the last two swings were missing. Move this week.
ACTION · THE REP TAKES IT FROM HERE
Draft the sequence in ClaudePush to the sequencerMCPDraft the intro emailSave to my accounts
This is the Split. The machine did the research, scoring, enrichment, and history. The rep clicks once and keeps the one thing that matters: the conversation.
THE EVIDENCE · WHY IT SCORED
THE VERDICT
Fit (base)55
Product usage · power adoption+18
Buying signals · hire + funding+14
Every point is auditable in the Signal Library. No black box.
THE PRODUCT STORY
WAU 28 → 42
3 power users, all in send tracking. Sequences never activated, the wedge. Adoption rising 12 weeks straight.
Cut the drag to the machine. Keep the judgment with the people.
The cut is the efficiency. The keep is the point. It is why a lean team gets stronger, not just cheaper.
05 · WHAT BROKE
The receipts include the failures.
A build log that only lists wins is a brochure. Here is what broke, because the scars are where the method came from.
The score learned the wrong thing
My rebuilt model scored big companies high and small low. It had learned that big companies are big: size, wearing fit's clothes. It taught me to never ship a score I can't explain, which is why the Signal Library exists at all.
The agents assumed a human was present
Run a build engine on a schedule and you find which steps quietly need someone logged in. My first scheduled run failed on a step that needed an interactive session. The fix was hunting down every place the system assumed a person, and removing it.
Bad data healed with bad data
Early on I let a model heal its own gaps from a source that was itself unreliable. Confident bad data. I turned the auto-heal off until the source was rewired. A system that fixes itself from a bad source just gets more sure of itself.
06 · WHAT OPERATORS SAY
“I worked for Heath as VP of Revenue for several years, and he stands out as one of the best sales leaders I've worked with across multiple industries. At Mixmax he led the transition from a sales-led to a product-led-sales motion, wearing countless hats and handling all of them expertly, and he brought the team along with him. Driving short-term growth is one thing; building an engine that delivers consistency is what Heath is incredible at.”
Morgan Wible · Mixmax
“Heath led Revenue, I led Product and Growth. Those roles are traditionally destined to fight. We understood we could build together. He started our Product Channel Tiger Team, pushed us to define PQLs together, and kept showing up to the product and data conversations that most sales leaders would skip. Heath is honest in a way that's rare, he says the uncomfortable thing, and he does the work that falls between org charts. I learned something almost every time.”
Jakub Tutaj · Mixmax
“I had the pleasure of working with Heath at Mixmax, and what I appreciated most was how data-driven he is as a leader, which made it easy to align on initiatives and make decisions we could actually stand behind. He also stays close to the work, happy to dig into the details himself and quick to unblock whatever's in the way. On top of that, Heath is someone you want to work with: supportive, easy to be around, and genuinely invested in his team's growth. Any team would be lucky to have him.”
Viola Melis · Mixmax
“I had the privilege of working with Heath at Mixmax. He creates an environment where everyone feels comfortable speaking their mind, asks the right questions, and challenges you to think differently without ever making you feel small. He encourages people to grow, trusts them to own their work, and celebrates their wins along the way. I'm grateful to have learned from his leadership and would highly recommend him to any team looking for someone who leads with both empathy and intention.”
Diana Calucer · Mixmax
See where I ran this.
This engine sits under a revenue number. Here is what it moved, and the rest of the fifteen-year arc.
