Back to Work

Case Study: Relay's New Prospecting Email Machine

•Nicholas Wilson
Case Study: Relay's New Prospecting Email Machine

TL;DR

Relay, a leading network of local social media creators entering their 3rd city, built a full-stack outbound engine that (1) scrapes ICP-fit accounts via Apify, (2) enriches and decision-maker maps with Clay, and (3) sends 2,000 highly personalized cold emails/month through Instantly. The system has delivered up to a 3.8% positive-reply rate (booked meetings), producing up to 24 new meetings/month and a repeatable top-of-funnel for city expansion.

Context

Relay helps great local businesses grow with a community of creators and a tight, ROI-driven playbook. To scale into new markets efficiently, Relay needed a predictable way to identify right-fit prospects, personalize at volume, and book meetings with decision-makers—without ballooning headcount.

Objectives

  • Precision targeting: Only message businesses that match Relay’s ICP by category, revenue band, and geography.
  • True personalization at scale: Reference signals that matter to each company’s likely objectives.
  • Throughput with deliverability: Sustain 2,000+ emails/month while protecting domain health.
  • Closed-loop learning: Feed outcomes back into targeting, copy, and channel strategy.

Tech Stack

  • Apify – Scrapes companies meeting ICP filters (profile traits, financial performance proxies, geo).
  • Clay – Enriches firmographics, pulls decision-makers, and generates personalized insights tied to assumed goals.
  • Instantly – Multi-inbox sending, warmup, rotation, sequencing, throttling, and deliverability guardrails.
  • CRM/Data Hub (optional) – Syncs meetings, outcomes, and suppression lists.

Workflow (End-to-End)

  1. Define ICP & Filters
    • NAICS/category, local city/ZIP radius, revenue/employment bands, review velocity, ad presence, social activity, location count.
  2. Data Acquisition with Apify
    • Scrape target lists from curated sources.
    • Normalize names/URLs; dedupe against CRM and suppression lists.
    • Output: company_id, name, url, city, state, phone, source, confidence_score.
  3. Clay Enrichment & Insight Generation
    • Append exec contacts (Owner/GM/Marketing/Ops).
    • Pull signals: recent promos, hiring, reviews, menu/site updates, content cadence.
    • Insight builder: “Given this company’s category, size, and signals, what’s their likely Q4 objective?”
      • Examples: “Lift weekday breakfast traffic,” “Stabilize staffing costs,” “Increase review volume before holiday season.”
  4. Message Assembly & QA
    • Personalization tokens: {FirstName}, {City} intake trend, {1-line compliment/observation}, {Assumed Objective}, {Tiny Win we can deliver first}, {Proof/creator stat}, {2-slot CTA}.
    • Human spot-check on the top 50 per batch before scale.
  5. Instantly Sending & Deliverability
    • Rotating warmed inboxes + domain pool.
    • Daily caps, randomized send windows, inbox health monitoring, auto-pauses on spikes.
    • Automatic unsubscribe handling and hard-bounce suppression.
  6. Reply Handling & Meeting Bookings
    • Positive replies routed to calendar link or human hand-off.
    • Auto-tag reasons for “not now,” “already have,” “wrong contact” for learning.
    • Push all outcomes to CRM with source/batch IDs.
  7. Learning Loop
    • Weekly review: subject performance, hook lines, objective hypotheses that converted, industry segments to scale/stop.
    • Update ICP, angle library, and suppression rules.

Results

  • Volume: 2,000 cold emails/month (stable without reputation damage).
  • Quality: Up to 3.8% positive-reply → scheduled meetings up to (27 meetings/month).
  • Consistency: Performance sustained across waves by enforcing data hygiene and messaging QA.
  • Focus: Higher meeting density in categories where assumed objectives matched real operator priorities (e.g., weekday day-part lift, review velocity lifts before seasonal demand).

Personalization That Mattered (Patterns)

  • Local proof + micro-win: “Pittsburgh breakfast lift play” coupled with a specific micro-pilot offer.
  • Operational empathy: Referencing staffing realities, review cadence, or menu rotation—no generic “growth” talk.
  • Two-slot CTAs: “Tue 3:30 or Thu 10:00?” beat open-ended asks.
  • Short, non-salesy tone (Relay’s style).

Sample first touch (3 lines):
“Hey {FirstName}—noticed {Location} is pushing weekday mornings. We help neighbors like {Peer} lift breakfast traffic with creator-driven reels + reviews (fast, low lift). Worth a 12-min look Tue 3:30 or Thu 10:00?”

Governance & Guardrails

  • Compliance: Clear opt-out, honoring do-not-contact, and source transparency; respect local email laws.
  • Data Ethics: Use only publicly available/compliant data; promptly delete upon request.
  • Deliverability: SPF/DKIM/DMARC, dedicated domains, consistent warmup; suppress bounces/complaints immediately.
  • Brand Safety: Human QA on insights to avoid incorrect assumptions or awkward references.

What Made It Work

  • Right data → right angle: The “assumed objective” hypothesis unlocked credible, specific value props.
  • Tight message constraints: 2–4 sentences, a single micro-outcome, and a binary CTA.
  • Relentless hygiene: Dedupe, suppress, and audit. Bad data kills domains and confidence.
  • Weekly iteration: Keep the 3–5 best hooks, retire the bottom half.

Next Experiments

  1. Multi-channel boosts: Add LinkedIn profile-view nudges or a 15-second “why now” IG reel for warm touches.
  2. Event-based triggers: Auto-spin campaigns when reviews spike, a location opens, or hiring surges.
  3. Offer testing: Swap “quick audit” vs “creator sprint” vs “review surge” to find category-winners.
  4. Pricing pathways: Pre-price good/better/best in follow-ups to accelerate self-selection.
  5. Post-meeting nurtures: Auto-send proof packs (case clips, creator rosters) if no-show or “not now.”

Implementation Notes (Reusable)

  • Data schema: company_id, domain, category, geo, revenue_band, review_velocity, signals[], dm_name, dm_title, dm_email, assumed_objective, personalized_hook, batch_id.
  • Quality gates: A/B subject caps, 10% manual review/sample, and “yellow-flag list” for risky angles.
  • Reporting: Daily deliverability dashboard; weekly cohort report (opens, replies, positives, meetings, wins).