TP Infinity Lab
A product studio inside TP Infinity.
The lab
A customer moves through a brand in one line: the review they read, the campaign they saw, the store visit, the call afterward. Inside the brand, that line is cut into departments. We work the whole line, on the brand's own data and on the public record, and orchestrate one response across it.
Solutions
Each chapter is organised around who signs the invoice, not around how we built the tool. Everything below has a live door or a real engagement behind it.
What the market, the category, and your customers are saying, read at scale.
聖人無常心 以百姓心為心
The sage holds no settled mind of his own; he takes the people's mind as his mind.
Public consumer voice across Google Maps, Instagram and Xiaohongshu, read and scored at scale. The beauty demo covers 13 brands over 5 markets and 3 platforms, with leaderboards, benchmarks and complaint patterns for the whole category rather than one brand.
How a brand, institution, or destination is talked about across a market.
A brand and market intelligence report compiled from public signals in days.
Case
Part of the Social Hub suite. Emerging narratives, sentiment shifts, and crisis precursors, with severity tiers and escalation logic built in. Runs on the shared Social Lens dashboard.
Who to reach, what they are worth, and what to say when you reach them.
Case
Part of the Social Hub suite. Voice-of-customer signal distilled into the cues that inform high-touch engagement, so every conversation begins where the customer already is.
Case
Part of the Model Gallery. Segmentation, CLTV, TA Scoring, and Recommendation, each running live in the browser on synthetic data with known structure planted in it.
AI put to work on the campaign floor.
治大國 若烹小鮮
Govern a great state as you would cook a small fish.
Case
Part of the Social Hub suite. Cultural signal turned into campaign concepts and platform-ready formats, plus KOL and KOC discovery, channel mix evidence, and post-campaign read-out.
Case
Part of the Model Gallery. Response curves learned from a year with dark weeks, then the mix rebuilt on sliders.
Reading and raising the contact-center floor.
What the numbers will do, and what to do about it.
Case
Part of the Model Gallery. Pick target, cohort, window; the backtest report card comes first.
AI use of your IP happens inside your rules. One system with four faces: Protect, Create, Monetize, Foundation.
Which of your accounts to act on this week, and why.
The Journal
Essays and field notes on consumer behavior, AI, and the craft of CX. Open to read.
Essay
One customer spent €40,000 in her first three months and has not replied to a message. Another has spent about €3,000 a year for eleven years. Spend, stage and closeness, kept apart, and the test the third axis has to pass.
Read →Essay
A way to classify contact-center interactions on three independent axes — wind, wave, channel — so you can finally answer why a surge happened, not just what it was about.
Read →Scrollytelling
Every customer drifts down a river at their own speed, below a surface you can't see. Sixty years of models got good at clocking those speeds and stayed blind to the moment one of them turns.
Read →