Case studies

Numbers, methods, and screenshots.

Long-form stories from recent engagements - what we shipped, what moved the metric, and what we'd do differently. No marketing fluff. No anonymized "Fortune 500 client." Just the work.

Read the stories Be the next one
42
Engagements shipped
2.6×
Median pipeline lift
94%
On-time delivery
11d
Median launch time
7
Industries served
Browse by industry
SaaS · AI · Pipeline
01
IndustryB2B SaaS · 14 people
EngagementMarketing site rebuild + AI pipeline
Timeline11 weeks · Q3 2024
StackNext.js · LangGraph · Postgres · HubSpot

Northwind: 2.4× pipeline by replacing 11 manual sales-ops steps.

A 14-person B2B SaaS team was drowning in inbound - but losing 60% of leads in handoff. We rebuilt the marketing site around the intents that converted, then routed every inbound through an AI pipeline that does the boring work.

Before/after - 14-day pipeline traffic
↑ launch
2.4×
Pipeline value
−63%
Sales-ops hours
11
Steps removed
11d
Time to first ship

What we did

  1. Audited the pipeline. Sat with three reps for two days. Found 11 manual steps between "form submitted" and "demo booked" - every one a drop-off.
  2. Rebuilt the marketing site around three high-intent paths instead of one homepage funnel. Cut copy by 40%, added live calculators, dropped the page weight to 280KB.
  3. Built an AI router. LangGraph agent reads the form, scores intent, drafts the response, books the meeting, updates HubSpot, slacks the rep. Human approves anything ambiguous.
  4. Closed the loop with evals. Every routed lead is graded after the demo. Misroutes feed back into the prompt and the eval set. Errors trended down from 12% to 2.8% over 8 weeks.

"We used to lose 6 hours a week to lead routing. Now we lose 20 minutes. The pipeline is 2.4× and the team is calmer."

- Head of Growth, Northwind
Talk to us about your pipeline →
D2C · Conversion
02
IndustryD2C · Home goods
EngagementPDP rebuild + landing experiments
Timeline4 weeks · Q4 2024
StackShopify · Klaviyo · Hotjar

Aster Mills: from 1.4% to 3.9% conversion on a single landing page.

A founder-led home-goods brand with stunning product photography but a 1.4% conversion rate. The hero was selling the brand. The page needed to sell the product. We rebuilt around the buy decision.

Conversion rate, weekly
+178%
Conversion rate
0.9s
LCP
−34%
Bounce
2.3×
Email captures

The unlock

We moved the price + ship-by date + reviews block above the fold. We replaced the brand-story video with a 3-second product loop. We rebuilt the PDP gallery with a sticky add-to-cart that survives scroll. We pulled UGC from Klaviyo into the gallery automatically. Boring fundamentals, executed.

"Our designer wanted to keep the moody brand intro. The data wanted the price. The data won. We're not going back."

- Founder, Aster Mills
Audit my e-commerce site →
SaaS · AI Agent
03
IndustryVertical SaaS · 80 people
EngagementTier-1 support agent + evals
Timeline10 weeks · Q1 2025
StackAnthropic · LangGraph · Linear · Zendesk

Halcyon Support: a tier-1 agent that resolves 71% autonomously.

Their support team was answering the same 30 questions, 600 times a week. We built a docs-grounded agent with a real eval harness - so it could be trusted to reply without a human reading every output.

Resolution path · last 30 days
Ticket in1,820
→
Agent first reply1,820
→
Auto-resolved1,294 · 71%
Human escalated526 · 29%
71%
Auto-resolved
−4.2min
First-response
98%
CSAT (auto)
3.2×
Throughput

The hardest part wasn't the agent

It was the eval harness. We built 240 test cases from real tickets, ran every prompt change against them, and gated every deploy on regressions. The agent isn't smart because the model is smart - it's smart because the team can iterate on it without breaking trust.

"The first time it auto-resolved a ticket about a niche permission edge case, I thought it was a bug. It wasn't. It just read the docs."

- Director of CX, Halcyon
Build my support agent →
Marketing · SEO
04
IndustryHR-tech · Series B
EngagementTopical authority strategy + 6mo retainer
Timeline6 months · Q4 2024 → Q2 2025
StackWebflow · Ahrefs · Linear · Notion

Verdant: +312% organic traffic in 6 months.

No tricks, no programmatic SEO, no AI-generated junk. We picked three topical clusters where Verdant had real expertise and built the best resource on the internet for each one. Internal linking did the rest.

Organic traffic · monthly
+312%
Organic traffic
−42%
CAC
8
Top-3 rankings
26
Pieces shipped

What we shipped

  • Three topical clusters: each with one pillar piece (4,000+ words) and 6-9 spokes (1,200-2,000 words).
  • An internal linking spec - automated in CMS - so every spoke points to the pillar with intent-matched anchor text.
  • An interactive calculator on the highest-intent page. Doubled time-on-page, became the #1 conversion path.
  • A monthly experiment: one new SERP feature target per month (FAQ, video, snippet).

"It felt slow for the first 8 weeks. Then it compounded. Now organic is our biggest channel and our cheapest."

- VP Marketing, Verdant
Get an SEO audit →
Mobile · Wellness
05
IndustryWellness · D2C
EngagementCross-platform app + launch ops
Timeline14 weeks · Q1 2025
StackReact Native · Supabase · RevenueCat · Mixpanel

Pathline Wellness: 4.8★ at launch, 38k downloads in 60 days.

A small wellness brand wanted an app that felt as good as their print magazine. We shipped iOS and Android in 14 weeks - including App Store launch ops, ASO, and the first 30 days of lifecycle hooks.

App store · daily downloads (60 days)
4.8★
Avg rating
38k
60-day downloads
14%
D7 retention
2.1%
Trial → paid

What made it stick

The first session is the whole game. We obsessed over the first 90 seconds - onboarding, first content piece, first "win" moment. Then we layered three lifecycle nudges in the first 7 days: each one tied to a behavior, not a clock.

"I've shipped six apps over my career. This was the calmest launch I've ever had."

- CEO, Pathline
Build my app →
Industries

Where we've shipped before.

We're not generalists. We're industry-experienced generalists - meaning we've shipped enough in each of these to know the playbook, and humble enough to learn the parts we don't.

⌨

B2B SaaS

14 engagements. Marketing sites, dashboards, AI agents, growth retainers.

⌗

D2C / E-commerce

9 engagements. Shopify, headless storefronts, PDP rebuilds, lifecycle.

⌬

Fintech

6 engagements. Compliance-aware, design-system heavy, observability obsessed.

✚

Healthtech

5 engagements. HIPAA-aware, patient-side and clinician-side products.

✦

Pro services

4 engagements. Workflow automation, client portals, internal tools.

◑

Wellness / DTC mobile

4 engagements. Cross-platform apps, content-led acquisition, retention loops.

Methodology

Why our case studies have real numbers.

01

Scope by outcome, not by feature

Every engagement starts with a metric we're going to move. The deliverable list comes after - and gets cut whenever it doesn't serve the metric.

02

Instrument before we ship

Analytics, evals, dashboards - built into the project plan. No "we'll add tracking later." Later is never.

03

Verified, not vibes

Every number on this page came from a client dashboard or a tool screenshot. We share the methodology in scoping calls.

04

Failures, on request

We've had projects that underperformed too. Ask on a call - we'll be candid about what happened and what we learned.

FAQ

Case-study questions, answered.

Are these numbers verifiable?+

Yes. Every metric on this page came from a client tool - Mixpanel, GA4, HubSpot, App Store Connect, Ahrefs. We share the source dashboard with serious leads on scoping calls.

Can you connect us with the clients featured?+

Yes - once we're past the first scoping call. We don't hand reference contacts to cold leads, but we're proud of every project here and happy to introduce you when there's mutual fit.

What was the engagement model for these projects?+

Mostly fixed-scope projects (8-14 weeks) followed by monthly retainers for iteration. A few were retainer-only from day one. The pattern: engagement size matched the metric we were committed to.

Do you only work with companies this size?+

No. We've shipped for 4-person seed startups and 200-person Series Cs. The case studies skew mid-market because that's where most of our work clusters - but the playbook scales both ways.

What if my industry isn't on the list?+

Tell us about it on a call. If it's adjacent to one we know, we'll move fast. If it's truly novel, we'll be honest about the learning curve and what that means for timeline and risk.

How do you decide what to feature here?+

Three filters: client willing to be named, results we can verify, and a story that teaches something. About one-third of our work makes it onto this page.

Be the next case study.

Tell us where you are and where you're stuck. We'll send back a real plan in 48 hours - and a number we'd be willing to commit to.

Start a project See our services