Bringing the world’s products into one catalog
Channel3 is working to make every product sold online searchable by apps and AI agents. Its catalog brings together product details, prices, availability, and links to buy, so developers can build shopping experiences without collecting and maintaining that information themselves.
Making that catalog useful means recognizing the same product across stores, even when each retailer describes it differently. Channel3 uses AI to read product pages, match listings, connect options such as sizes and colors, and fill in missing details. Information from retailers and partner networks becomes consistent data that apps can search and compare.
The work keeps coming as products are added, listings change, and prices and stock levels move. As Channel3 expands its catalog, it has more information to process and keep current.
When looking for better cost monitoring, CTO George Lawrence described running AI prompts millions of times a day on different inputs, consuming billions of tokens. Tracking spending across AI providers and platforms was time-consuming, and reports could take a couple of days to arrive. At that volume, a bug would run up costs long before it showed up in a spend report.
A simpler way to track AI costs
George was looking for a way to track millions of AI requests a day and see usage and cost across providers in one place.
Datadog metrics had broken twice on Datadog’s side, according to George. PostHog’s AI Observability seemed better suited to examining individual AI agent traces than to monitoring metrics across billions of requests.
Channel3 already had tools to check AI response quality and inspect a sample of requests, but they needed a reliable and scalable system to track their overall usage and cost across different workflows, providers, models, and other dimensions.
Channel3 adopted SuperPenguin’s SDK to send AI usage data in the background, while requests continued going directly to its providers. The team gained a shared view of costs across an AI operation running billions of tokens a day.
Very easy to set up. Every other solution we have pursued has been more heavy-handed.
A live view of AI spend
Channel3’s team now regularly uses SuperPenguin’s dashboard to manage AI costs. It brings spend across providers into one shared, live view, with support for both open-source and closed models.
With live cost breakdowns, the team can see which workflows are driving spend as its workload changes. George uses that visibility to choose where engineers should focus their cost-cutting work, without waiting days for provider reports.
- The starting point
- Reports could take a couple of days. The team needed an overall view of millions of daily AI requests across providers.
- With SuperPenguin
- Live cost breakdowns help the team identify expensive workflows and focus engineering effort on reducing their cost.
Turning cost visibility into savings
Recognizing that much of Channel3’s workload is not as time-sensitive as realtime chat, SuperPenguin’s optimization agent recommended moving some of Channel3’s heavier AI workloads to batch processing. For those jobs, batching allowed Channel3 to submit AI requests together and receive the results later at a much lower price.
SuperPenguin helped quantify the potential savings so Channel3 could decide whether cost improvements from recommended actions justified the engineering time. Channel3 chose which jobs could wait and made the changes to move them to batch processing.
Batching was one part of Channel3’s broader AI cost-cutting effort. To investigate other spending, George also used his AI assistant, connected to SuperPenguin’s cost data through MCP.
Managing AI costs as Channel3 grows
Channel3 used SuperPenguin to track cost per AI request and see whether its changes delivered the expected savings. The savings from batching exceeded the team’s expectations.
On selected workloads, SuperPenguin recorded a drop of more than 50% in average cost per AI request.
At billions of tokens a day, cost is a constraint on Channel3’s growth. The team uses SuperPenguin’s live dashboard regularly to manage that spend. During its cost-cutting effort, the team also relied on SuperPenguin to coordinate the work, from choosing changes worth making to checking the results.
SuperPenguin is the central place for managing our AI spend.