A teardown of the revenue engine this role runs: the HubSpot-centred RevOps stack, where revenue leaks before it's booked, the KPIs I'd run, a plan against all 11 responsibilities, and proof — the agentic AI automations I already build (§17). Sourced from the JD + public info (Jun 2026); targets illustrative.
RevOps L2 is the operating system of CoinGecko's Sales team: keep HubSpot clean, the pipeline honest, the forecast trustworthy, leads routed fast, commissions correct — and, increasingly, automate all of it with AI. This JD is unusually explicit about that last part: it names Zapier, n8n, Claude, Gemini and Google Apps Script, and asks for "agentic" routing logic and AI-assisted insight automations. That's exactly what I build — autonomous AI agents and workflow automations (§17) — and I'm crypto-native, so CoinGecko's product and customers aren't foreign to me. The honest stretch: I haven't held a titled "Sales Operations" seat; I've done the analytical, ops and customer-facing work around it (§18).
CoinGecko is a Malaysian-founded (TM Lee & Bobby Ong, 2014), bootstrapped, no-VC crypto-data leader. The part that matters for this role: the Sales team's product is the commercial CoinGecko API & data licensing — paid tiers from Analyst/Lite/Pro up to Enterprise (custom SLAs, redistribution & white-label) — sold to crypto projects, exchanges, wallets, fintechs and institutions, plus advertising. So "revenue" here is B2B API subscriptions and data deals, and RevOps L2 keeps that motion running on HubSpot. Being Malaysia-based and crypto-native — a daily CoinGecko user across past protocol roles (deBridge, BOB, Aztec) and builder of web3wagmi.com — I'm close to both the company and what it sells.
Sources: founders/bootstrapped history from public profiles; platform numbers (10k tokens, 400+ exchanges, 45B calls, 300M views, 100+ countries) from the JD; commercial API tiers from coingecko.com/api/pricing (Jun 2026). See §19.
The JD's eleven responsibilities, each given its own section on this page — the sidebar maps one-to-one. Here's what each means in week-to-week RevOps work.
| # | JD responsibility | What it means in practice | On this page |
|---|---|---|---|
| 1 | Sales Process Optimisation | Maintain process docs & playbooks; find high-impact automation in the funnel. | §05 |
| 2 | Revenue Analytics & Insights | Maintain dashboards; ship AI-assisted automations that generate insights to leadership. | §06 |
| 3 | Sales Strategy & Planning | Provide data/analysis for strategic planning & the annual ops roadmap. | §07 |
| 4 | Sales Forecasting & Pipeline Mgmt | Validate data for weekly forecast; raise accuracy with signal-tracking & pipeline-hygiene triggers. | §08 |
| 5 | Commercial Ops & Standardization | Ideate, build & execute scalable sales plays and growth strategies. | §09 |
| 6 | CRM & Tech Stack Management | Be the go-to for day-to-day HubSpot support; use AI tools to build logic, troubleshoot and automate admin. | §10 |
| 7 | Sales Enablement | Update training materials; maintain the sales collateral library. | §11 |
| 8 | Lead Management | Run lead assignment/routing; build "agentic" routing that adapts to seller capacity & lead quality. | §12 |
| 9 | Data Integrity & Governance | Regular cleanup; advanced queries/scripts to auto-detect inconsistencies before they hit reporting. | §13 |
| 10 | Commission Administration | Pull performance data & prepare reports for incentive-plan calculation/review. | §14 |
| 11 | Programme & Project Management | Complete assigned tasks, track progress, give status updates on ops projects. | §15 |
"What we look for": 2–3 yrs sales/business ops (≥1 yr customer-facing); HubSpot, strong Excel/Sheets, exposure to Xero / B2B data tools / marketing platforms / BI; Zapier, n8n, Claude, Gemini, Google Apps Script; analytical, organized, entrepreneurial, web3 interest. Level L2 · RM9,938–RM10,932 (CoinGecko publishes transparent salaries). Fit mapped in §18.
Reconstructed from the JD's named tools and how a B2B API-sales motion typically runs. A dollar of revenue travels a pipeline — lead in → enriched → qualified → forecast → closed → invoiced → reported — and every handoff is a place revenue leaks or data rots. RevOps owns that whole pipe, and the JD is clear it wants it automated with AI.
Threaded across the whole pipe is the automation layer the JD names: Zapier / n8n for orchestration, Claude / Gemini for AI-assisted insights and "agentic" routing, and Google Apps Script for the glue between Sheets, HubSpot and finance. The modern RevOps shift this JD is hiring for is moving from manual admin to automated, AI-assisted operations — exactly the seam I work on.
The JD asks, in its own words, for experience "building and launching programs with Zapier, n8n, Claude, Gemini and Google Apps Script" and for "agentic" routing logic and AI-assisted insight automations.
This is the half of the JD I'm strongest on — I ship autonomous AI agents and workflow automations (§17). For me this isn't a "nice to have"; it's the day job I already do.
The ten ways a B2B revenue engine quietly loses money or trust before a dollar is booked. For each: what leadership or a seller actually says, the real root cause, and the first move — with the AI/automation fix the JD is reaching for. This is the map I'd run §05–§15 against.
| Leak mode | What they say | Real root cause | First move (the fix) | Sev |
|---|---|---|---|---|
| Dirty CRM data | "The reports don't tie out" | Duplicates, missing/free-text fields, no validation — every downstream number inherits the error. | Dedup pass; required-field validation; an automated inconsistency check (Apps Script / n8n) that flags before reporting. (§13) | High |
| Pipeline leakage / stale deals | "Forecast keeps slipping" | Deals rot in a stage with no activity; no hygiene trigger catches them. | Stale-deal trigger: auto-flag deals past N days in stage → nudge owner; weekly hygiene review. (§08) | High |
| Forecast inaccuracy | "We missed the number again" | Commits are gut-feel; no signal-tracking on what actually predicts close. | Validate the forecast data weekly; add automated signal-tracking (engagement, stage age) as the JD asks. (§08) | High |
| Slow lead routing | "Leads go cold before first touch" | Manual assignment, no capacity/quality logic — speed-to-lead dies. | Agentic routing on seller capacity + lead-quality signals; alert if unactioned past SLA. (§12) | Med |
| Attribution gaps | "We don't know what's working" | Broken/blank lead-source & UTM capture; can't tie revenue to channel. | Enforce source stamping at capture; reconcile source → closed-won in BI. (§06) | Med |
| Manual reporting drift | "Every dashboard says something different" | Copy-paste reporting, no single definition of a metric. | Govern metric definitions once; auto-refresh dashboards from source, not hand-built decks. (§06) | Med |
| Commission errors | "My commission is wrong" | Manual data pulls + spreadsheet math — one slip and seller trust is gone. | A reproducible commission pipeline: same query, same source, audit trail every cycle. (§14) | High |
| Process drift / no playbook | "Everyone sells differently" | Undocumented process; new reps improvise; plays don't scale. | Maintained playbooks + standardized sales plays as living docs. (§09) | Med |
| Enrichment gaps | "These records are thin" | No automated enrichment; reps hand-research accounts. | Zapier / n8n enrichment flow on new leads; fill firmographics at capture. (§10) | Low |
| CRM ↔ Xero handoff breaks | "Closed-won isn't invoiced" | Manual handoff to finance; revenue booked late or missed. | Reconciliation check between closed-won and Xero; flag mismatches automatically. (§10) | Med |
RevOps leaks compound silently. A 2% dirty-data rate becomes a wrong forecast becomes a mis-sized hiring plan — nobody sees the bug, only the missed number. That's why the JD repeatedly asks to automate detection "before it impacts reporting" and to raise forecast accuracy with signal-tracking rather than gut feel. The job is to make the pipe observable and self-correcting, not to manually mop up after it.
This taxonomy is illustrative of the model I'd run, built from the JD's named responsibilities and standard B2B RevOps practice — not internal CoinGecko data. Severities show how I'd triage. No private systems were accessed.
Not hypothetical — I built all three against a realistic crypto-API pipeline (HubSpot · n8n · Claude · Apps Script), runnable on a Google Sheet today. See the working demo →
An Apps Script / n8n job that flags duplicates, missing required fields and stale deals in HubSpot and routes them to the owner — kills the §04 leaks that poison every report, automatically.
Pick the contested metrics, write one governed definition each, auto-refresh from HubSpot/Xero — so "every dashboard says something different" stops on week one.
A small Claude/n8n flow that routes inbound on seller capacity + lead-quality signals and alerts on anything unactioned past SLA — the JD's "agentic routing," shipped as a v0.
At KIP I oversaw the sales process for a B2B engagement with a Thai tertiary-education provider (bringing AI to their textbooks) — re-architecting the intro → pilot → contract funnel so stalled deals moved. I authored process docs at BOB, and built the §17 demo (stale-deal, routing, AI digest) against this exact RevOps funnel — so "find the high-impact automation" is something I've already done, not theory.
Maintain the dashboards, define each metric once, and — the JD's exact ask — ship AI-assisted automations that generate insights to leadership, not just raw data. Below: the loop, and an illustrative script that pulls deals, flags inconsistencies, and drafts the "so what."
// Illustrative: weekly pipeline-hygiene + AI insight digest // (Apps Script / Node style) — runs on a schedule, no manual deck const deals = await hubspot.getDeals({ stage: "open" }); // 1 — flag the data leaks before they hit a report const issues = deals.filter(d => !d.amount || !d.owner || // missing fields daysInStage(d) > 30); // stale deal // 2 — let the model write the "so what", not just the "what" const brief = await claude.summarize({ prompt: "Weekly pipeline health for leadership: " + "call out forecast risk & the top 3 actions.", data: { coverage, conversion, issues } }); postToSlack(brief); // reviewed before it goes wide
Illustrative of the pattern, not production config. Built & runnable in the §17 demo. The principle — govern the metric, automate the pull, let AI draft the insight — is exactly the JD's "AI-assisted automations to generate insights to leadership."
Analyst at KIP — structured reasoning over ambiguous, incomplete data. First-principles, and genuine crypto-market context: I know the segments the strategy targets (funds, exchanges, wallets, institutions) because I've worked across them and I'm in that market daily.
The metric set I'd run for pipeline health and forecast accuracy. The point isn't the dashboard — it's that every metric has a definition, an owner, and a trigger. Values below are illustrative targets to show the model, not CoinGecko data.
Data completeness, deal-age and stale-deal count are leading — they predict a forecast miss before it lands. Forecast accuracy and won revenue are lagging. The JD wants accuracy raised via "automated signal-tracking and pipeline-hygiene triggers" — i.e. manage the leading indicators daily and the lagging ones follow.
Analyst discipline from KIP — I built eval harnesses and scorecards in Python and ran money-on-the-line work where a wrong number was costly. Validating forecast data and turning "gut commit" into "signal-backed commit" is the analytical core I'm strongest in. The §17 demo ships a working stale-deal trigger.
At deBridge I did ecosystem/BD — pitching wallets and protocols (e.g. 1inch) to integrate us, which is a B2B partnership pipeline: identify the target, make the case, drive the integration. I ship the play and the automation that runs it, and I read crypto GTM (CT, communities, conferences) so the plays fit how this market actually buys.
HubSpot Revenue Operations Certified, and I build and debug automations and agents daily — workflow logic, scripts, API glue (§17) — so HubSpot's automation surface is the same shape of problem I already solve. Honest note: hands-on HubSpot admin is newer for me, but for someone who ships n8n/Apps Script automations it's a weeks-not-months ramp, and I'd over-index on it early.
I authored the docs/FAQs at BOB that measurably cut repeat queries — the exact "turn recurring questions into self-serve content" loop. I write for clarity (5M+ Quora views, podcast host, Toastmasters club president) in EN / 中文 / BM, and being crypto-native the collateral stays technically accurate.
Administer day-to-day assignment/routing cleanly first, then build the JD's "agentic" routing logic that adapts to seller capacity and lead-quality signals — a sense → decide → act loop: read the new lead's signals, check who has bandwidth and fit, assign, and escalate anything left cold past SLA. I built a working v0 in the §17 demo.
This is literally what I build. Base Yield Agent (wagmi/viem, ERC-8004 #34601) senses on-chain state and acts — the same sense→decide→act shape as capacity-aware routing. And I know where CoinGecko's leads come from — I'm embedded in Crypto Twitter & the web3 community, so I understand how crypto B2B and institutional pipelines source and convert (projects, exchanges, funds).
At KIP I built an AI-QA failure scorecard — classifying bad automated outputs into clean categories (wrong / missing / no-match / partial). That's the same discipline as classifying data-quality defects and gating on them. Add a security mindset (CompTIA Security+) and comfort writing queries/scripts in Python & JS, and the §17 demo's hygiene detector is exactly this.
Detail-oriented and reproducible by default — analyst-grade accuracy at KIP, and comfort scripting the pull (Python / Apps Script) so the numbers are the same every cycle and auditable, not hand-keyed.
I ran operations, comms and project tracking for global communities at FrodoBots and Humanode across daily-shifting scope. I default to documented, reproducible work over heroics — exactly what operational project work needs.
This JD asks for Zapier, n8n, Claude, Gemini, Apps Script and "agentic" logic by name. For most applicants that's a wish-list; for me it's the day job.
/api/stats, /api/exchange-rate, /api/health…) on scheduled refreshes. So I've sat on both sides of the API economy this Sales team sells into: the buyer's "is it worth the tier?" call and the producer's uptime/versioning.| JD asks for | What I bring |
|---|---|
| AI & workflow automation — Zapier, n8n, Claude, Gemini, Apps Script, "agentic" logic | My strongest area. I build autonomous AI agents (Base Yield Agent, ERC-8004 #34601) and workflow automations; this is the day job, not a course I took. |
| Analytical skills — gather, organize, report accurately | Analyst at KIP; built AI-QA scorecards & Python eval harnesses; money-on-the-line rigor on the data↔revenue seam. |
| CRM (HubSpot), strong Excel/Sheets; Xero / BI / B2B data exposure | HubSpot RevOps Certified; fluent in Sheets & scripting; I build the automation layer that sits on top of these. Hands-on HubSpot admin is my ramp item (§10). |
| 2–3 yrs sales/business ops, ≥1 yr customer-facing | Sales-process work at KIP (re-architected the funnel for a B2B Thai education deal — AI for their textbooks) and BD pipeline at deBridge (pitching wallets like 1inch to integrate); years customer-facing (Aztec, BOB, FrodoBots). Honest gap: not a titled "Sales Ops" seat — see below. |
| CoinGecko product & crypto/blockchain understanding | A daily CoinGecko user across past protocol roles (deBridge, BOB, Aztec) and builder of web3wagmi.com (crypto news / on-chain data / DeFi yield) — plus a live data product (Malaysia4U) that buys & ships APIs. I know CoinGecko's product, its domain, and its customers from the inside. |
| Entrepreneurial, organized, communicates clearly, web3 interest | Self-initiated this teardown; ran ops/projects for global communities; 5M+ Quora views; Chevening Scholar; deep web3 interest. |
The AI/automation half the JD emphasizes, analytical rigor, crypto-native domain knowledge, Malaysia-based, and a builder who ships.
No titled "Sales Operations" role and HubSpot admin is new. My answer: I've done the analytical/ops/customer-facing work around RevOps, I'm unusually strong on the AI-automation side this JD leads with, and I ramp on tools fast — I learn systems by building on them.
Documented & reproducible over heroics, direct, detail-obsessed, GMT+8 and ready to hybrid into a Malaysian-founded company I genuinely rate.
Items tagged JD come straight from the job description (the 11 responsibilities, named tools — HubSpot/Xero/Zapier/n8n/Claude/Gemini/Apps Script, platform numbers, and the L2 salary band RM9,938–RM10,932). Items tagged confirmed are public company facts (founders, 2014, bootstrapped) and the commercial API tiers from coingecko.com. The RevOps stack diagram (§03), the leak taxonomy (§04), the KPI targets (§08) and the code (§06) are illustrative of the model I'd run — built from the JD + standard B2B RevOps practice, not internal CoinGecko data. My projects (Base Yield Agent, ERC-8004 #34601, KIP scorecard) are real and verifiable via edwardtay.com. No private systems were accessed.
Role: CoinGecko — Revenue Operations Associate (L2), Lever
Company: coingecko.com
Commercial API tiers: coingecko.com/api/pricing
Career levels: CoinGecko careers / level structure
Me: edwardtay.com
My data product: data.malaysia4u.com (consumes & ships APIs)
My crypto hub: web3wagmi.com (news · on-chain data · DeFi yield)
Agent standard referenced: ERC-8004 (onchain agent identity)
Claims are split into JD / confirmed (sourced) and illustrative (the stack, leak taxonomy, KPI model & code) throughout. Code is illustrative of approach, not production config. This is unsolicited interview homework — happy to walk through any section live.
Prepared by Edward Tay · for the CoinGecko Revenue Operations Associate (L2) role · Jun 2026 · edwardtay.com · Edwardtay7@gmail.com