
Nanga author
Louis Monier
CTO
Data engineering and artificial intelligence · 12 pieces of content
Background
An engineer by training who specialized in artificial intelligence, Louis designs Nanga’s technical architecture, from data collection pipelines to predictive models. He applies rigorous quantitative methods, such as Bayesian modeling, to marketing, and turns complex data problems into simple, reliable features.
Topics
Kily, Nanga’s AI agent, its analysis methods and the state of the art in artificial intelligence.
Articles 12
Businesses in ChatGPT: what an official profile changes for your brand, and what it does not
Since November 10, 2026, a brand can claim its profile in ChatGPT, much like a Google Business Profile. The profile makes its information accurate. It does not get the brand recommended, as OpenAI states itself. What this means for a marketing team, and how to prepare while the waitlist runs.
Proving GEO pays: the test design your CFO can accept
As of 7 November 2026, I could not find a single published study that measures GEO's effect on revenue against a control group. This guide gives you the test to run, the sums to do first, and what the test will not be able to prove.
On Black Friday, advertising costs almost twice as much. It also returns far more.
For four days, the price of advertising soars and it still pays off. By Tuesday, only the price is left. What three years of data say, from Black Friday to Christmas.
At 100 prompts, your AI visibility score still carries ±8 points
At 100 prompts and a 20% presence rate, one week's score carries ±8 points, before you count how alike prompts in a topic are. Here are the margins recomputed from 20 to 500 prompts, and the reading rule that follows from them.
New model, new curve: who moved, your brand or the engine?
Swapping a model is enough to move an AI visibility score, and nothing flags it. The October 2026 GPT-6 rollout is the case study: five rules for telling what actually moved before you report it, reward it or fund it.
A city an AI engine accepts is not a city it honors
Across 31,200 API answers collected in October 2026, ChatGPT followed the declared city, while Gemini ignored it and followed the network's country. Before you fund city-level tracking, a two-city test shows what your numbers measure.
Some of the answers behind your AI visibility score were never answers
An AI engine that repeats your prompt can produce a mention of your brand without answering anything. The Nanga team found this case and four others in its own data. Here is how to spot them, and what to ask any GEO tool.
Kily Competitors: Comparative Analysis and Competitive Intelligence
Compare your creative groups against each other and benchmark your strategy against competitors with Kily Competitors.
Kily: Ask Questions to Your Advertising Data
Kily turns advertising analysis into a conversation. Ask a question, get a contextualized answer, with access to all your reports and data.
Kily Weekly Recap: Your AI-Generated Weekly Performance Brief
Every week, Kily Weekly Recap analyzes your advertising performance, flags standout campaigns, and delivers a structured brief, with zero manual configuration.
AI: Current Dynamics and What to Expect in 2026
Notes and reflections on the state of AI in 2026.
Kily Creative: How AI Identifies Your Best-Performing Ads
How Kily Creative analyzes your ads, surfaces top performers, and detects creative fatigue using personalized benchmarks and AI trained on your own data.