HtPi HtPi CLASS 課
About these classes

Bring the problem. Leave with a clearer way forward.

Data1/3 · Online orders — the data these steps were written for2/3 · Customer survey3/3 · Card statement

QuestionAverage order, shipping included?

OrderAmountShipping
#1001$320$30Check
#1002$450—0no answerCheck
#1002$450—0no answerCheck
#1003$99,999$0Check
#1004$280$30Check

QuestionAverage score (1–5)?

ReplyScore
R015Check
R024Check
R03—0no answerCheck
R0444Check
R055Check
R055Check

QuestionAny suspicious charges?

TimeMerchantAmount
08:12Convenience store$32Check
12:40Café$58Check
12:40Café$58Check
19:05Supermarket$214Check
03:12Overseas web shop$38,000Check
Steps in the playbookOrdersSurveyCard
  1. Why: the same record may have been entered twice.This data: #1002 appears twice — one order, exported twice. Remove it.This data: R05 replied twice — one person. Remove it.This data: the same amount twice in one minute — maybe a double charge. Flag it; don't delete it.
  2. Why: a blank usually means “none”.This data: a blank shipping fee means free shipping, which is 0. Fill it.This data: a blank is no answer, not a score of 0. Don't fill it; leave it out of the average.This data: there are no blanks. Nothing to do.
  3. Why: an extreme value is usually a typo.This data: $99,999 is a system test order. Remove it.This data: 44 on a 1–5 scale can only be a typo. Remove it.This data: 3am, overseas, $38,000 — exactly what you're looking for. Flag it; don't delete it.
Following the steps$3703.5None
The real answerKnowing why$3704.72
A ready-made playbook: three standard cleaning steps — remove duplicates, fill blanks with 0, drop extreme values. Just follow them.On this data, following the steps gets the right answer — they were written for it.Change the context: new data, the same steps again. The same steps give 3.5; the real answer is 4.7. Every step was followed, and the answer is still wrong.The same steps find nothing suspicious; there are 2. Every step was followed, and the answer is still wrong. Know why each step is there, then check whether that holds for this data — if it does, do it; if it doesn't, change it.The same step: right for this data, wrong for that. Remember only the steps, and you can't adapt when the situation changes.

Finding an answer isn't hard any more. Articles, videos, AI — and plenty of free courses — can quickly hand you a clear next step.

But if all you take away is a ready-made playbook, a change in context can make it useless. And if you only need the immediate result, a consultant or contractor may be the more efficient choice.

HtPi classes are after something else: not just dealing with the problem in front of you, but cutting through the noise to take apart the structure, judgement and methods behind it. After class, even when the problem changes shape, you still know where to begin, what to look for, and how to adjust. And if one class isn't enough, the 1:1 Studio picks up where the class leaves off — or subscribe to HtPi and keep working through the interactive essays.

In class, you are not there to watch. Bring your own questions, experience, or even an idea that has not fully taken shape yet. Small groups, active participation, hands-on work.

What you leave with is no longer a trick or two, but methods you can keep using, connect with one another, and multiply with what you already know.

Teaching & public sharing

Some classrooms taught in, and discussions joined.

Over the years, Horris has spoken on data, technology, AI and digital transformation — at universities, professional bodies and industry events. A partial record follows.

Horris Tse speaking on a Women in Finance panel
Panel discussionWomen in Finance
Guest lecture at HKU SPACE
Guest lectureHKU SPACE
01

FORTUNE INSIGHT

命中注定的 AI 創業家

AI, entrepreneurship and real-world practice View the record ↗
02

THE STANDARD

Open up on data, says venturer

Open data and AI View the record ↗
03

CAMPAIGN ASIA

The next KOL battleground

Platforms, content and digital markets View the record ↗
04

SING TAO

把握機遇，AI 夢想起飛

Business, technology and AI View the record ↗
Who teaches

The class design is mine. The teaching doesn’t have to be.

Every class is designed by Horris Tse — the content, the pacing, and how the hands-on parts work. Most classes he teaches himself; some runs invite a guest instructor, or are co-taught.

Whoever leads the room, the curriculum, materials and way of teaching are stated plainly on the class page.

Four principles
  1. 小班

    Small groups

    Room enough that questions aren’t passed over.

  2. 參與

    Participation

    Not just listening — asking, trying, trading views.

  3. 實作

    Hands-on

    Real tools and real cases — limits and trade-offs included.

  4. 延續

    Continuity

    Materials stay after class; keep following the question — one sign-in across htpi.me.

The room

Small by design — the size where every question is heard, and every screen that needs seeing gets seen.

Seat counts, length and format are set per class, and each class page states its own. Venues vary too — Kwun Tong most often, sometimes Sheung Wan. Wherever it is: tables, power, and enough quiet to think.

Seats
small · set per class
Length
2–3.5 hrs typical
Format
participatory

The next class is forming.

See open classes