A 20-person B2B SaaS team was sending 5,000 cold emails per week and getting 400 replies. Their close rate was fine, but the top of the funnel was starving. 90 days later they were getting 1,600 replies per week on lower volume. Here's exactly what changed.
Starting state: the volume trap
The team was doing what most B2B sales orgs do. They'd picked Apollo, bought 50K contacts, dropped them into Outreach.io with a 5-step sequence, and hit 'send.' Their metrics looked like this:
emails sent per week
reply rate
weekly replies
of replies became SQLs
On paper it looks fine. In reality, it was grinding the team into the ground and producing diminishing returns. Every week they'd burn through more contacts. Deliverability was slowly degrading because of rising unsubscribes and spam complaints. And critically — most replies were 'not interested' or 'unsubscribe,' not real conversations.
The 4 changes that moved the needle
Change 1: cut volume 40%, raise list quality 10x
First move: stop buying broad Apollo lists. Instead we built custom ICP lists using a combination of Apollo + LinkedIn Sales Nav + manual firmographic filtering + intent signals (companies with hiring posts for relevant roles, recent funding rounds, specific tech stack). List size dropped from 50K/month to 12K/month. Quality was night and day.
Counterintuitive insight: volume was hurting reply rate because bad-fit contacts mark you as spam, which hurts deliverability for your good-fit contacts.
Change 2: AI research layer for every single lead
For every contact in the reduced list, we ran an automated research pass: latest 3 LinkedIn posts, company press releases in the last 60 days, website headline, recent job change. A Claude agent extracted 2-3 specific personalization hooks per lead.
Cost: $0.02 per lead in LLM fees. For 12K leads/month, that's $240/month — trivial compared to the lift.
Change 3: rewrote the emails around research, not templates
Old emails were template-based with variable substitution ('Hi {name}, saw you work at {company}'). New emails were AI-generated per lead using the research hooks, grounded in the team's offer, in a consistent brand voice.
Old template approach
- Hi [name], saw you lead [function] at [company]
- We help companies like yours [generic value prop]
- Are you open to a 15-min call next week?
- Reply rate: 8%
AI-personalized approach
- References specific recent post/launch/news
- Connects research hook to measurable outcome
- Asks for a 1-sentence reply, not a meeting
- Reply rate: 32%
Change 4: deliverability infrastructure
The single most underrated change. We rotated to a portfolio of 12 secondary sending domains (all warmed for 4 weeks before use), implemented SPF/DKIM/DMARC properly, throttled sending volumes per inbox to deliverability-safe levels, and added automatic pause logic when any inbox hit spam complaint thresholds.
The 90-day results
reply rate (up from 8%)
pipeline generated vs 90 days prior
incremental qualified pipeline
reduction in avg response time
What would've failed
Doing any 1 or 2 of these changes would have barely moved the needle. The compounding effect is what worked. We've seen teams do 'AI-personalized emails' on bad lists and get 10% replies. We've seen teams do 'better lists' with template emails and get 12%. It's the full rebuild — better lists + research + AI personalization + deliverability — that breaks the 30% ceiling.
See the full B2B sales ops case study
FAQ
Rebuild your outbound the same way.
We rebuild B2B outbound sales systems. List quality, research layer, AI personalization, deliverability — all the pieces that separate 8% reply from 32%. Free outbound audit below.
