GTM Engineering
Assessment
The goal is a working outbound engine the team owns by day 90, not a system that breaks when we leave.
The read
Extract has a PLG product, a mobile/gaming customer base most ELT tools can't touch (EA, King, Playtika, Scopely, Miniclip, SciPlay, WB), and an outbound motion that is "evolving toward precision and signal-driven targeting." Three things matter:
- 01.The product event stream is the unfair advantage. Every signup, source connection, and extraction is a first-party intent signal nobody else in the category has. Most outbound is cold. Extract's can be warm by default if it gets plumbed in.
- 02.ICP is the only place to start. Targeting wrong accounts with great signals and great copy still loses. The model has to be built first, then everything else activates it.
- 03.The risk is over-building. Clay can be a black hole. We ship the worst usable ICP model by week 3 and iterate weekly against SDR feedback, not in isolation.
The three bets
Bet 1. Build the ICP scoring model.
Take action from it. The model has two input layers and one output:
- Fit: vertical, stack, headcount, has a warehouse, uses the connectors we cover.
- Timing: hiring data/BI/analytics-eng roles, fundraise, public Fivetran/Airbyte cost complaints, tech-stack change.
- Output: scored in Clay, written to CRM. We can then decide the action to take: campaign outreach or signal-based outreach.
v1 live by week 3. Everything else downstream depends on this.
Bet 2. Plug PLG events into the ICP model to prioritize action & campaigns.
Get product events out of the warehouse and into Clay + CRM. Three plays to start:
- Signup, no source connected in 7 days.
- Source connected, no extraction in 14 days.
- Multiple signups from the same company domain (route to AE, not SDR).
Specific outbound copy per play because we know exactly what they tried. This is the bet only Extract can make.
Bet 3. Stand up the execution loop so the engine gets used.
Boring infrastructure that decides whether the first two bets matter:
- Sequencer-CRM-Clay glue.
- Persona-and-signal templates so reps spend 30 seconds prepping, not 10 minutes.
- Other infrastructure projects: inbound plays, signal-based plays.
Everything documented so a named in-house owner can run it without us by day 90.
What we're consciously not doing
- No enterprise account-intelligence platform (6sense, Demandbase). Six figures for capability we can build for ~$3k/mo.
- No generic third-party intent data as a primary input. Too noisy. We have better signals in the product.
- No AI BDRs, voice agents, or multi-channel orchestration. Earn it after email and LinkedIn are dialed.
- No custom-built outbound platform. Clay does it.
- No CRM or sequencer replatform. Six-month black hole.
The pattern: nothing that builds capability before there's signal it's needed, nothing that buys vendor-flavored intelligence when first-party intelligence is sitting unused, nothing that creates dependency on us.
How we'd know it's working
ICP model v1 live in CRM, one SDR using it in cadence, weekly refresh running. Product event access confirmed or escalated. Funnel baseline captured. One experiment per week is the cadence, not the exception.
Three PLG plays live and producing meetings. Personalization templates live for 3 signal plays. Three SDRs using the scored list. First measurable lift on accept rate by score band. Named in-house owner identified for handoff.
Top-quintile accounts converting to meetings at ~2x historical baseline. PLG-sourced accounts ~25% of meetings with higher win rate than cold. Meetings per SDR per week up ~30%. Documentation handoff-ready. In-house owner running the experiment cadence with us reviewing weekly.
What would make us change course.
- SDRs not using the scored list in 30 days: integration or trust problem, not model problem.
- Product event access blocked past day 14: escalate. Past day 30: shrink Bet 2.
- A play converts at <1.5x cold baseline: retire it. Defend results, not bets.
What we'd need from you
Sequencer and CRM standardized. One named owner who can approve ICP weighting. Confirmation we're going first-party plus targeted external, not blanket intent.
Product event stream (non-negotiable, day 1). One named SDR partner and one AE partner for 90 days. CRM admin. Weekly hour with Demand Gen and Product Marketing.
Protect from one-off "build me a dashboard" asks. Public leadership signal that PLG events belong to GTM. No pipeline-attribution forecast at day 30.
Keep us on staff to run more of our GTM engineering playbook. Lots of automations for automated follow up, collateral creation, multi threading.
Our goal is to partner, identify white space and execute and implement systems that solve for it.
What this brief should have been
The brief gave us the company and the role. To make sharper choices we'd have wanted operational reality:
- Current SDR numbers: headcount and meetings-per-rep-per-week.
- The actual tool stack: CRM, sequencer, enrichment, warehouse, what the product emits and where it lands.
- Win rate by source: PLG self-serve vs outbound vs inbound.
- Two SDRs we could talk to for 30 minutes each.
A one-page ops snapshot and one named customer reference would let us defend the bets with evidence instead of inference.