CASE STUDY · API MANAGEMENT & INTEGRATION (IPAAS)
Building an API gateway that cut integration time for new partners by half.
- 8 weeksPartner onboarding down from 8 weeks to 4
01 · PROJECT OVERVIEW
The client and the brief.
An anonymised mid-market B2B services company. Name withheld under confidentiality.
An API gateway, documentation and a sandbox. Partners can read the docs, request keys and test against realistic data without waiting for staff.
- 1Partner onboarding took 8 weeks.
- 2A new partner's developers emailed questions, waited for answers, received test credentials by hand and often built against outdated information.
- 3The company's own engineers were pulled away to explain the same things again.
- 4Some partners gave up before going live.
- 5Every integration was slightly different, so support was hard.
02 · WHAT WENT WRONG
What went wrong, and how we fixed it.
Every engagement hits problems. These were the ones that mattered here, and what we changed so they did not come back.
- PROBLEM 01The pilot partners were able to use the sandbox, but several then failed in production
The sandbox was more forgiving than the real system: it accepted records that production validation rejected. We aligned the validation rules in both environments, added the real error cases to the sandbox, and wrote a short go-live checklist covering the differences that remain. The second pilot partner went live without escalations.
- PROBLEM 02The second pilot partner went live without escalations
- PROBLEM 03A second hurdle was the quota policy
We first set one rate limit for all partners. A large partner running a nightly batch hit it and thought the platform was broken, while smaller partners never came close. We moved to per-partner quotas with a documented process for asking for more, and added a clear response that tells the caller when they can retry.
- PROBLEM 04Complaints about throttling stopped
03 · THE PROCESS
How we ran it.
We began by reading the email history and talking to the engineers who handled onboarding, to find where the weeks went. Most delay was waiting, not building. We then worked in four phases: clean up and version the API surface, put the gateway in front of it, build documentation and the sandbox, and pilot with two partners before opening it up. We met the engineering lead weekly and reviewed the documentation with a partner-facing manager, since that person knew the questions partners actually asked.
- 01Step 1
We began by reading the email history and talking to the engineers who handled onboarding, to find where the weeks went.
- 02Step 2
Most delay was waiting, not building.
- 03Step 3
We then worked in four phases: clean up and version the API surface, put the gateway in front of it, build documentation and the sandbox, and pilot with two partners before opening it up.
- 04Step 4
We met the engineering lead weekly and reviewed the documentation with a partner-facing manager, since that person knew the questions partners actually asked.
04 · RESULTS
What changed.
Measured against the way the work ran before the engagement began.
- NOW 01
Partners start in the sandbox and rarely need a call.
- NOW 02
API changes follow a versioning and notice process.
- NOW 03
Documentation is updated with each release and treated as part of the product.
- NOW 04
Support can see a partner's recent errors before replying.
- 8 weeks
Partner onboarding down from 8 weeks to 4
05 · WHAT WE BUILT
How it works.
- Gateway: a single entry point handling authentication, rate limiting, request validation and logging.Partners get individual keys with scoped permissions.
- API contract: an OpenAPI description kept in version control that drives documentation, validation and sample requests.
- Versioning: breaking changes require a new version, with a published deprecation notice period.
- Sandbox: a separate environment with synthetic data and the same behaviour as production, including common error cases.
- Documentation: a getting-started guide, authentication walkthrough, worked examples for the most common flows and a list of error codes with causes.
- Observability: per-partner dashboards showing request volume and errors, so support can see what a partner sees.
06 · QUICK ANSWERS
The short version.
- What problem did the client have?
- Partner onboarding took 8 weeks. A new partner's developers emailed questions, waited for answers, received test credentials by hand and often built against outdated information. The company's own engineers were pulled away to explain the same things again. Some partners gave up before going live. Every integration was slightly different, so support was hard.
- What did Brandsmashers build?
- An API gateway, documentation and a sandbox. Partners can read the docs, request keys and test against realistic data without waiting for staff.
- How was the work run?
- We began by reading the email history and talking to the engineers who handled onboarding, to find where the weeks went. Most delay was waiting, not building. We then worked in four phases: clean up and version the API surface, put the gateway in front of it, build documentation and the sandbox, and pilot with two partners before opening it up. We met the engineering lead weekly and reviewed the documentation with a partner-facing manager, since that person knew the questions partners actually asked.
- What were the results?
- Partner onboarding down from 8 weeks to 4.
YOUR TURN
Have a similar requirement?
Tell us what is stuck. We will come back with a plan and matched, pre-vetted engineers.
