The homepage sells on "per-second billing, sub-5-minute deployment, no commitments" and "no sales calls". But the main "Deploy now" button opens a form whose Term field offers only 1, 2, 3 or 5 years, asks for storage in petabytes, and has no on-demand option. The pricing page's four "Rent now" buttons go to Runpod's generic homepage, where a visitor has to find FarmGPU capacity alone. The developer who came for an H100 by the second hits a procurement form or loses the trail. Growth is the funnel, and this is its first door.
First action
Change the "Rent now" links from runpod.io to a tracked link to FarmGPU's cluster listing.
How to measure it
Tracked on-demand starts
- Now
- –
- Target
- 100 a month
- By
- 60 days
Steps
Map the paths
List every Deploy, Rent and Book button and where each lands today.
Marketing Manager 1 day
Agree the link with Runpod
A referral or deep link to FarmGPU Instant Clusters.
CBO and Co-Founder 1 week
Build the on-demand page
Three steps, per-GPU price, a short video.
Marketing Manager 3 days
Fix the form
Add an on-demand term and make storage optional.
Marketing Manager 1 hour
Relabel the buttons
"Start on demand" and "Reserve capacity" across all pages.
Marketing Manager 1 day
Tag the clicks
Google Analytics events on both paths.
Marketing Manager 1 day
Review weekly
Starts, forms and demos by path.
Global Sales Manager weekly
Why it works
Split the button into two clear paths. "Start on demand" goes to a FarmGPU page that walks through launching a FarmGPU Instant Cluster on Runpod (a tracked link to the exact cluster page, a 3-step guide, the per-GPU price) so FarmGPU sees every click. "Reserve capacity" keeps today's form, with "On-demand / month-to-month" added to the term choices. Measure both paths separately.
Buyers at the top of the funnel want to try first and talk later; the promise on the page should be one click away. Tracked hand-offs to Runpod also tell FarmGPU how much on-demand demand its marketing creates, which it cannot see today.
Ready to use
ChecklistTwo-door deploy checklist
[ ] "Start on demand" page with per-GPU prices [ ] Tracked link to FarmGPU capacity on Runpod [ ] "Reserve capacity" form with an on-demand term option [ ] Storage field optional [ ] Same two buttons on home, pricing, platform, performance [ ] Analytics event per button [ ] Weekly report: starts, forms, demos
Risks and how to handle them
Runpod cannot pass conversion data back
Use a dedicated referral code and compare usage monthly.
Large buyers skip the form
Keep "Reserve capacity" beside the on-demand path on every page.
Evidence
- Your websiteHomepage: "Stop overpaying for GPU compute. Up to 70% cheaper than hyperscalers on the same NVIDIA GPUs. Per-second billing, sub-5-minute deployment, no commitments." "Spin up dedicated GPU instances on-demand through RunPod. No quotas, no approval queues, no sales calls." The comparison strip shows an H100 hourly rate of "~$2.99" against "$7+". Main buttons: "Deploy now" and "Book a demo".Source
- Your websiteThe "Deploy now" page ("Break ground.") is a form: name, work email, company, end user yes/no, GPU type (H100 to Vera Rubin), workload, backend network, storage type, "Storage Capacity (PB)", location, and a Term field whose only choices are "1 year", "2 years", "3 years" and "5 years". There is no on-demand choice and no link to start right away.Source
- Your websitePricing page: the table says H100 SXM "Per GPU/hr ~$3.29" vs hyperscalers "~$7-13"; the cards read "H100 Cluster $3.29 per hour" (8x H100), "H200 Cluster $4.59 per hour", "B200 Cluster $6.79 per hour" and "B300 Cluster $7.89 per hour", each above an "8x" spec list. Every card's "Rent now" button links to www.runpod.io/, the partner's generic homepage. Reserved 1-5 year rates are quoted on request.Source
- Your docsdocs.farmgpu.com (Mintlify) is company documentation: snapshot, thesis, hardware and software stack, data center specs, security, legal, team, culture, brand. Its llms.txt index lists no quickstart, API reference or "launch your first cluster" guide; Instant Clusters are launched "from the RunPod dashboard with a few clicks". Two citations on that page link to the blog with "?utm_source=chatgpt.com" appended.Source