Product Talk · Free post · Discovery & customer research · Metrics, data & experimentation

Product in Practice: Getting Started with Several APIs (And the Challenges I Faced)

Teresa TorresApr 16, 202524 min
SourceProduct Talk
KindFree post
PublishedApr 16, 2025
Originalproducttalk.org ↗
N:

Teresa Torres describes how her course business was drowning in manual admin work across several tools, and how a mistake-prone admin process led her to automate using APIs. She explains her approach: pick a single system of record (Airtable, chosen because its API was well documented) and learn each tool's API through a repeatable process covering authentication, endpoints, error handling, and rate limits. Along the way she documents many usability gaps: missing auth docs, undocumented business rules, incomplete endpoints, and silent failures. Her core argument is that API teams need to do far more discovery, such as pairing with customers, testing assumptions, and usability testing, to find these issues before customers hit them.

01Key takeaways

  • Automate repetitive, error-prone admin tasks that software can handle better than people.
  • Choose one primary system of record as the single source of truth before wiring tools together.
  • Learn each API's authentication, endpoints, and error behavior in a consistent order before building.
  • Test API assumptions and pair with customers to surface undocumented business rules early.
  • Document hidden dependencies between fields, since undocumented rules create costly trial-and-error for users.

02Key sections

Automating the manual admin work
Torres was manually syncing student data across six tools, leading to errors. Reading a book on software led her to realize humans should not be doing tasks that computers can automate.
Choosing a single source of truth
Teachable lacked an API, so she moved student data to Airtable, which had a well-documented API, and built around it as her system of record.
Learning each API's authentication and endpoints
Authentication varied widely across tools, with Slack requiring a dummy app and Google Calendar lacking server-to-server support. Endpoint coverage also varied, from Teachable's missing user lookup to LearnWorlds' comprehensive API.
Request formats, business rules, and undocumented logic
Inconsistent request formats, hidden business rules (such as LearnWorlds' price and access interaction), and confusing linked-record behavior in Airtable made integration difficult despite documentation.
Error handling and rate limits
Blank fields, overloaded server errors, and silent failures made error handling hard, while rate limits were rarely a constraint but occasionally caused problems with concurrent tasks.

03From the post

“A few years ago I was drowning in administrative work. To deliver high-quality online courses we were patching together several different tools to create a good student experience. We used Teachable as our course platform, Slack for our community, Zoom for our live sessions, Google Calendar to send out”

“I was having humans (my admin) complete tasks that a computer was much better suited to do.”Teresa Torres · Product Talk
“API teams need to be doing far more discovery than they are clearly doing.”Teresa Torres · Product Talk
“Error messages are how APIs communicate during times when things go wrong.”Teresa Torres · Product Talk

04Frameworks mentioned

Summary and takeaways written by PM Atlas; quotes are short excerpts. © the original author.