GTMhub / Work / Case Study

How we rebuilt outbound for a Series B fintech — and 3.4x'd reply rate in 90 days.

Industry: Fintech · Series B Engagement: Build + Operate Duration: 12 weeks build · 6 months operate Published: May 12, 2026
Reply rate
3.4x
+ from 0.7% → 2.4%
Across cold + warm sequences combined
Meetings booked
2.1x
+ 47 → 98 / month
Senior buyer meetings (Director+)
Pipeline opened · 90d
$1.8M
+ vs $0.6M baseline
From the rebuilt outbound motion alone
Cost / meeting
−62%
$420 → $160
Fully loaded · agency + tooling + ops
01 · Context

$28M ARR, plateauing outbound, and a board meeting in 90 days.

A Series B fintech in the embedded-payments space came to us in mid-Q1. They'd grown to $28M ARR on a strong inbound motion — content, SEO, and a few enterprise referrals — but outbound had stalled. Their 8-person SDR team was doing 800 dials a week and booking 47 meetings a month. Reply rate sat at 0.7%. Cost per meeting was $420 fully loaded.

The team had tried three vendors in 18 months. Each one promised AI-personalization at scale. None of them changed the underlying numbers. The Head of GTM had two months to show a meaningfully different outbound motion before a board check-in.

02 · The problem

It wasn't an AI problem. It was a signal problem.

Our two-week diagnostic sprint surfaced what was actually broken — and it wasn't what the team thought.

  • The SDR team was working a stale ICP list refreshed quarterly. Half of it was no longer in-market.
  • "AI personalization" meant first-line LinkedIn one-liners. The middle of the email — the actual pitch — was identical across all 12,000 contacts.
  • There was no signal-detection at all. No funding rounds, no leadership changes, no product launches, no tech-stack moves. Outbound went out on a calendar, not a trigger.
  • Most painful: the data warehouse had 80% of the signal the team needed. It just wasn't being read.
"The data was sitting in Snowflake. We just didn't have the operating layer to turn it into a sequence."
Head of GTM · Series B Fintech
03 · Our approach

Build the signal layer first. Sequences second.

We didn't start with AI. We started with the data. The first four weeks were a signal-and-routing build: every account in their CRM got tagged daily across 14 signal types — funding events, leadership changes, product launches, ad-spend spikes, tech-stack additions, hiring patterns, news mentions, and seven more.

Only then did we layer in AI workflows: an account-research agent that prepared a one-page brief on every triggered account, a sequence composer that pulled from a library of 40 narratives matched to signal types, and a routing engine that handed the right account to the right SDR at the right hour.

Every output was eval-gated. Nothing went out without passing our quality harness — a small evaluator model + a senior human review for 5% of outputs sampled randomly. We caught hallucinations early. We caught tonal drift early. We caught off-positioning early.

04 · Architecture

Inside the system we shipped.

The whole thing runs on their existing stack — Salesforce, Snowflake, Outreach — with our operating layer sitting alongside. No tool replacement. No data exfiltration. Their security team signed off in week two.

▾ The stack we shipped
01 · Signal layer
14 daily signals
funding · leadership · tech · spend
02 · Research agent
One-page brief
cited · eval-gated · < 12 sec
03 · Sequence composer
40-narrative library
matched to signal type
04 · Routing
SDR + time-of-day
based on prior conversion
05 · Eval harness
QA at every step
5% sampled · human-reviewed
06 · Console
Observability
runs · costs · audit log

The architecture is intentionally boring. Six clear stages, each independently observable. When something starts to underperform — and something always does — we know exactly which stage to tune. No black box.

05 · Results

Numbers, plain.

At 90 days post-launch, the rebuilt motion is doing what the previous motion couldn't at 5x the contact volume:

  • Reply rate: 0.7% → 2.4% (3.4x). Driven mostly by signal-relevance, not by AI copy.
  • Meetings: 47/mo → 98/mo. Quality is up too — senior buyer titles are 64% of meetings, up from 38%.
  • Pipeline opened: $1.8M in 90 days from outbound alone, vs ~$600K in the prior comparable period.
  • Cost per meeting: $420 → $160 fully loaded. The team is doing less, more effectively.
  • SDR retention: zero churn since launch — a first for this team. The work is harder but more meaningful.
"The board meeting went well. We didn't pitch AI. We pitched a working signal layer with auditable outputs."
VP RevOps · Series B Fintech
06 · What's next

From outbound to full GTM.

We're now 4 months into the Operate phase. The same signal layer is being extended into inbound routing (Q3) and expansion plays for existing accounts (Q4). The team has moved from "we need an AI vendor" to "we have an operating layer." That shift is the actual win.

If you're seeing the same pattern — strong inbound, plateauing outbound, board pressure to "do AI" — talk to us. The diagnostic sprint is 2 weeks and $7.5K. You'll know within that whether the problem is what you think it is.

Same problem?

Let's diagnose yours.

2-week Sprint. $7.5K fixed. You walk away with a roadmap whether you build with us or not.

Book a Sprint See all services