Mamamia
More content, smarter monetisation.
Challenge
Editors manually tagged every article for topic & sentiment, then guessed which ad segment to book – time-consuming and inconsistent.
Solution
We trained a content-classification model paired with an LLM that suggests the optimal ad segment based on live inventory and reader profile.
Implementation highlights
- 1
Text Embedding
Articles embedded into a vector DB for similarity search.
- 2
Category Prediction
Classifier assigns multi-label tags with 99 % precision.
- 3
Ad Segment Suggestion
LLM cross-checks tags against inventory & outputs placement.
Results
40%
Editorial throughput boost
30%
Higher ad-segment CPM
99%
Tagging accuracy