Case study, 2026
Billing and attribution for Konga’s retail ads platform
Konga’s retail media network sells sponsored placements across search and product listings, priced per click. I built the part that turns clicks into money: billing and spend, click-to-impression attribution, deduplication, ad slot allocation, and spend control across campaigns running at the same time.
- Role
- Senior software engineer
- Where
- Konga Group
- When
- 2026
- Stack
- NestJS, MySQL, BullMQ, ElasticSearch, Redis
The problem
On a cost-per-click network, every click is an invoice line. Sellers fund a campaign, their products appear in sponsored slots on search and listing pages, and each click draws down their budget. The marketplace earns on every search, but only if the ledger is right.
That makes billing the part of an ads platform where mistakes cost the most. Count a click twice and a seller is overcharged. Miss one and Konga gives away inventory. Let a campaign run past its budget and someone has to explain the difference.
What I built
The billing and spend layer: the record of what each campaign has spent and what it has left.
Click-to-impression attribution, so every billed click traces back to the sponsored impression a shopper actually saw, and deduplication, so the same click is never charged twice.
Ad slot allocation, deciding which campaign fills each sponsored placement on search and listing pages.
Spend control across concurrent campaigns, so budgets hold while many campaigns compete for the same traffic at once.
How it fits together
The platform is built on NestJS and MySQL, with BullMQ, Redis and ElasticSearch.