3.18.2005

田口法則

 
田口玄一的田口法則(Taguchi Method)在工業工程界是一個被廣泛應用的技術,幾十年前他在日本電信研究所為了提升通信品質所做的研究,提供了一個能夠大幅降低測試成本的方法.這個新的品管技術大大改變了日本的工業界,到了八零年代也被引進美國,成為改變美國汽車工業的神奇學說.

簡單的說,當我們試著最佳化一個生產程序的時候,比如說是摺紙飛機吧,傳統的統計學者會先列出所有可能影響紙飛機飛行距離的變數,像是紙的重量,重心的位置,機翼的長短,飛機的名字等等,然後,在一次只能改變一個變數的前提下,進行重複的實驗.就算以上四個變數都只有兩個可能性,也會有二的四次方也就是十六種不同的組合.這十六種組合需要個別經過十幾次或是幾百次重複的測試,才能夠消除其他未知變數的雜音,一千六百次的紙飛機測試之後,雖然有可能可以找出最佳的組合,不過負責折飛機的倒楣鬼也去了半條命.

田口法則提供的是一個測試的矩陣.在這個矩陣裡,並不是所有的變數都需要獨立的測試,不同的變數可以被結合在一起,然後因為矩陣的設計,還是得到效果相近的測試結果,也可以說是懶人的測試法.雖然他的方法廣泛地被統計學者質疑,可是在生產線上,這個矩陣的威力是十分巨大的.當甲公司可以用比較少的成本去實驗而找出生產程序的最佳化的時候,乙公司還在煩惱為了折飛機而終身殘障的員工勞工保險的問題.

在行銷上,其實我在做的事情也是不停的實驗而已,一個產品在市場的成功與否,有不同的贈品,價格,廣告郵件的大小,顏色,字體,許許多多不同的變數需要被測試,可是往往一個測試兩三個變數的實驗,就可以花掉幾萬塊美金的預算.所以真正被挑戰的變數少之又少,往往最後產品推出的時候,資深經理人的常識跟直覺還是佔了百分之九十的決定.

The biggest thing that we learned is that your natural instinct isn't always right --Offermatica

在網路上找田口法則的時候看到這家叫做Offermatica的行銷公司的廣告辭,"我們學到最重要的一件事,是你的直覺並不是永遠正確的",跟我討厭常識是一樣的邏輯.可是由於資源有限,有的時候討厭歸討厭,什麼辦法也沒有.

INC.跟廣告世代雜誌上介紹的Dr.Kowalick,倒是為同樣受到統計學限制的廣告主提供了一線希望.不知道為了什麼田口法則這個已經在工業界普遍被應用的技術需要十幾年才進入行銷界.反正不管怎樣,Dr.Kowalick,原本在大學裡教田口法則的學者,跨越了這個鴻溝,成為第一個把田口法則跟廣告行銷結合的人.

今天讀著他的資料,然後在電話上跟他公司裡的另外一個顧問聊了將近一個小時.整家公司只有六個人(其中只有Dr.Kowalick跟他是顧問),可是這幾年來專業雜誌的報導已經讓他們變成一個十分受到矚目的公司,他們每一個案子的服務費是在四萬美金到十五萬美金之間."下個月我們的專訪會出現在兩百萬訂戶的富比士雜誌(Forbes) 上,到那個時候我應該沒有空跟你講電話了吧",那個傢伙說.對於我依照傳統統計學的訓練所提出來的問題,他的輕蔑是毫不隱瞞的,"我們用的是田口方法,統計學者是抓破頭也想不通的",他說.

不知道為了什麼,我一點也沒有不舒服的感覺,我想,除了在工作上體認統計學的限制之外,在Krannert第二年的上學期,為了證明田口法則跟傳統統計方法的不同,跟同學在系上大樓的長廊折幾百次紙飛機手差點斷掉的回憶,突然跑回大腦裡,也有直接的關係吧.

4 則留言:

方祖涵 提到...

三月中Dr.Kowalick的講座已經過去了,下一次應該是在九月,如果能夠參加應該是不錯的事情.明年要加入棒球統計學會的計畫到目前也沒有改變,這麼多跟數學有關係的事情真是令人頭痛呢.

匿名 提到...

http://www.olsonzaltman.com/
我永遠弄不懂數字的東西,這是Zaltman發明,用人的潛意識找出的終極行銷方法~~

匿名 提到...

You remind me of last fall when I tried to talk the marketing manager into replacing "A-B split" with "fractional factorial experimental design" (which serves the same purpose as Taguchi Methods). He couldn't get it and turned down my suggestion due to, I believe, fear of not being able to control something he doesn't understand.

So, I can understand how frustrated you are fighting others' common senses.

That said, I don't dislike or hate common sense. Although common sense (especially an experience-based one) may not be always right, it is right in most cases. It gets you into the ballpark right there --- cheap! That's what I like about it.

Besides, there's hardly any marketing analysis without common sense.

Take the example you mentioned, were it not for common sense, how do we determine which attributes (paper quality, font, etc) we should spend $$ testing, let alone which levels (font size 7,9, or 10, etc) to be tested?

As you said, we cannot trust common sense, but I believe common sense makes our lives easier.

Speaking of the Taguchi Method, I know very little about it, does it invole less "black box" math/statistics?

If so, the Taguchi Method is probably an easier sell to my boss who doesn't have advanced knowledge of stats.

方祖涵 提到...

Of course...what we should hate, or dislike, are those common senses that cannot be challenged.

Here is some information for the next Taguchi seminar. Good luck!

DR. KOWALICK'S TRIZ-TAGUCHI SEMINAR
JUNE 16-17, 2005
SAN FRANCISCO, CALIFORNIA

This is Dr. James Kowalick, and I was simply overwhelmed at the response to what occurred just a few weeks ago – on March 10 and 11, when we hosted the First TRIZ-Taguchi Seminar at Half Moon Bay, California, just outside of San Francisco.

We literally packed the beautiful Strawberry Ranch with exceptional people who are now moving toward brown-belt or black-belt status. That means they can really "do" it!

I said there would be only one more seminar like this, and that we’d have it in beautiful San Francisco. I know that everyone from Half Moon Bay will be there, and that leaves some space for a few dozen more. So here is your chance:

MY SECOND TRIZ-TAGUCHI SEMINAR IN SAN FRANCISCO, CALIFORNIA ON JUNE 16-17, 2005.

Those who come will receive everything he or she must know to actually do TRIZ-Taguchi Ad Optimization – plus more. It’s what we did at Half Moon Bay, PLUS three new top secrets. Here's what the chosen few will receive:

Instructions on how to use my Kowalick-Taguchi Methodology that multiplies responses by hundreds - sometimes thousands - of percent.

Discussions of the ins and outs of TRIZ-Taguchi Ad Optimization for each advertising medium.

My NEW Pay-Per-Click optimization system.

And the three new top secrets for new attendants as well as Half Moon Bay veterans.

Seminar attendees will fly to the San Francisco International Airport.

We'll begin Thursday June 16th at 9:00 AM and finish Friday June 17th at 5:00 PM. The seminar cost is $2,495 but...we are offering an Early Bird discount of $500, good only until April 25th. The fee includes lunches for both seminar days.

See you at the seminar! I can't wait. You may not know it, but this will change your life in advertising.

Ciao!

Jim Kowalick
P.S. If you need more information about the seminar, call me at (877) 820-4164 or email me at jim@kowalick.com

費曼與陳偉殷

得過諾貝爾獎的理論物理學家費曼是曼哈坦計畫,奥本海默團隊裡最年輕的成員。這幾天運動時候聽Freakonomics Podcast說他的故事,像是在原爆讓日本投降後,如何面對內心掙扎,甚至到日本學日文跟佛法找尋救贖等等;另外,他對所有事情保持好奇心與探索精神,包括愛情,也很引人入勝...