2018年12月，美國(guó)第十大精釀啤酒廠德舒特釀酒公司宣布計(jì)劃裁員10%。公司將其歸咎于銷售業(yè)績(jī)和銷量的下滑，而這也是精釀啤酒市場(chǎng)達(dá)到飽和后的常態(tài)。

對(duì)于大多數(shù)釀酒商來(lái)說(shuō)，如此大規(guī)模的裁員還意味著銷售和產(chǎn)量的削減。但德舒特并沒(méi)有這個(gè)打算，因?yàn)榫驮谒哪昵埃緵Q定在生產(chǎn)流程中整合接入互聯(lián)網(wǎng)的傳感器。

當(dāng)然，啤酒釀造依然是一個(gè)以人力為中心的行業(yè)。傳統(tǒng)上，工作人員在生產(chǎn)過(guò)程中會(huì)對(duì)啤酒進(jìn)行手工選樣和分析，來(lái)確定啤酒是否應(yīng)該從一個(gè)釀造階段轉(zhuǎn)至下一個(gè)階段，這一流程稱之為階段轉(zhuǎn)移，而這樣的階段一共有9個(gè)。過(guò)早或過(guò)晚的轉(zhuǎn)移都會(huì)影響成品啤酒的品質(zhì)。

與眾多釀酒商一樣，德舒特總部位于俄勒岡州本特市，它一直保留著這些樣本和分析記錄。隨后，公司決定發(fā)揮這些記錄的效用，并利用微軟和OSISoft，在云端對(duì)數(shù)據(jù)進(jìn)行解析，從而預(yù)測(cè)生產(chǎn)過(guò)程中的轉(zhuǎn)移次數(shù)。由此得出的結(jié)論幫助簡(jiǎn)化了釀造流程，為公司帶來(lái)了事半功倍的效果。

德舒特公司的釀酒師布萊恩·法芙芮說(shuō)：“當(dāng)出現(xiàn)產(chǎn)量問(wèn)題或裁員時(shí)，公司不大愿意投資其他的資產(chǎn)。我們已經(jīng)不再為全天候運(yùn)營(yíng)而招聘新人。通常，我們不得不做出犧牲，而且以往都是以質(zhì)量、產(chǎn)能或員工的幸福為代價(jià)……如今我們可以說(shuō)的是，我們對(duì)這一模式充滿信心。”

預(yù)測(cè)分析框架已經(jīng)被植入到公司的所有約50個(gè)釀酒罐，其容量從100桶到1000桶（3150加侖至3.15萬(wàn)加侖）不等。當(dāng)前，公司在釀酒師確認(rèn)某一釀造階段完成之后才會(huì)通過(guò)人力進(jìn)行階段轉(zhuǎn)移，但法芙芮稱，公司正在探索實(shí)現(xiàn)這一流程的自動(dòng)化。

釀造過(guò)程數(shù)據(jù)分析都會(huì)帶來(lái)什么凈效應(yīng)？德舒特將每批啤酒的發(fā)酵流程時(shí)長(zhǎng)降至48個(gè)小時(shí)，較之前減少了24個(gè)小時(shí)。這也讓公司能夠在不購(gòu)買額外設(shè)備的情況下提升其年產(chǎn)量。

到目前為止，德舒特是唯一一家使用傳感器和數(shù)據(jù)分析來(lái)協(xié)助釀造的公司，但法芙芮表示，像Sierra Nevada這樣的一些釀酒商也慕名前來(lái)了解這一模式。

該項(xiàng)目還催生了一個(gè)精釀啤酒釀造商開源數(shù)據(jù)搜集項(xiàng)目，這些啤酒商會(huì)分享其啤酒釀造期間各個(gè)階段轉(zhuǎn)移時(shí)長(zhǎng)的歷史記錄。

法芙芮說(shuō)：“大多數(shù)公司可能都有傳感器，但他們將數(shù)據(jù)記錄在紙上或電子表格中。我們的這一舉措可以幫助他們搜集數(shù)據(jù)并構(gòu)建一個(gè)數(shù)據(jù)庫(kù)，然后為其提供一個(gè)空間，供它們打造上述數(shù)據(jù)的數(shù)據(jù)庫(kù)，這樣，它們便可以在未來(lái)改進(jìn)其生產(chǎn)。”

他繼續(xù)說(shuō)道：“在精釀行業(yè)，做出這種調(diào)整是困難的。如今，人們逐漸適應(yīng)了這種做法。他們將其看作是一種工具，而不是搶奪其飯碗的事物。”

隨著工作量的減少，德舒特如今正在尋找新的方式來(lái)利用其數(shù)據(jù)分析工具。在探索將其運(yùn)用到日常操作的同時(shí)，例如在設(shè)備即將破損時(shí)用于預(yù)警的預(yù)測(cè)分析，公司還在思考更具行業(yè)針對(duì)性的運(yùn)用方式。

其中的一個(gè)便是使用光譜儀來(lái)測(cè)量啤酒風(fēng)味。

法芙芮說(shuō)：“我們將所有配方放在數(shù)據(jù)庫(kù)中。如今，我們?yōu)檫@些配方找到了匹配的數(shù)據(jù)，因此我們將進(jìn)行試驗(yàn)分析，以便對(duì)啤酒中的各類化合物進(jìn)行測(cè)量。這便是我下一步要做的事情，也就是利用這些數(shù)據(jù)來(lái)努力嘗試是否能夠找到讓啤酒出現(xiàn)某種特質(zhì)的化合物搭配比例，而正是因?yàn)檫@些特征，消費(fèi)者才會(huì)對(duì)我們的啤酒感興趣，我們的啤酒才能與眾不同。”

這種分析并非意味著取消啤酒釀造中的人力因素，它更像是對(duì)釀造流程的加速。有時(shí)候，在釀酒公司發(fā)現(xiàn)一個(gè)成功的配方時(shí)，它們已經(jīng)反復(fù)進(jìn)行了100多次或更多次數(shù)的釀造，為的是尋找它們所追求的特定風(fēng)味。然而借助科技，測(cè)試次數(shù)可以降至10次。

不可否認(rèn)，對(duì)于一些釀酒公司來(lái)說(shuō)，此舉會(huì)讓釀酒工作失去一些樂(lè)趣。最好的啤酒釀造商有三分之一歸功于其瘋狂的科學(xué)實(shí)驗(yàn)，它們會(huì)嘗試用各種超乎人們想象的事物與啤酒花和麥芽進(jìn)行混合，而且通常會(huì)催生出可口的新風(fēng)味。

為了確保這個(gè)傳統(tǒng)一直延續(xù)下去，德舒特設(shè)立了一家測(cè)試工廠，每一次僅釀造一桶啤酒，并不斷地進(jìn)行試驗(yàn)，然后通過(guò)其品嘗室獲得反饋，從而了解啤酒擁躉對(duì)新風(fēng)味的反響。（這些小批量啤酒數(shù)據(jù)的記錄方式與量產(chǎn)啤酒無(wú)異，說(shuō)不定某一個(gè)試驗(yàn)就會(huì)大獲成功。）

法芙芮表示：“精釀啤酒需要人們傾注大量的心血和精力。因此，將這些內(nèi)容從啤酒中剝離開來(lái)是一個(gè)敏感的話題。在這一方面，公司必須創(chuàng)建信任，并進(jìn)行對(duì)話。這一舉措是一個(gè)工具。我們并不打算將啤酒釀造交給機(jī)器和工程師來(lái)做，它只是提高釀酒效率的一個(gè)手段罷了。”（財(cái)富中文網(wǎng)）

譯者：馮豐

審校：夏林

In December 2018, Deschutes Brewery, the nation’s tenth largest craft brewer, announced plans to lay off 10% of its workforce. Declining sales and volume were cited as the reason, a familiar refrain as the craft beer market hits a saturation point.

For most brewers, a layoff that significant would also mean cut distribution and production. But Deschutes has no such plans, thanks to a decision made just under four years ago to incorporate Internet-connected sensors into the brewing process.

Beer making, of course, remains a very human-centric industry. Traditionally, workers have manually sampled and analyzed beers during the production to determine when the beer should be moved from one of the nine brewing phases to another, a process called phase shift. Transferring a beer from one step to another too early or too late impacts the quality of the final product.

Like many breweries, Deschutes, based in Bend, Ore. kept the records of those samples and analysis. Then it decided to put it to work by tapping Microsoft and OSISoft to use data crunching in the cloud to predict transition times during production. The results helped streamline the brewing process and helped the company do more with less.

“When you’re struggling with capacity or have a layoff, you don’t want to invest in another asset,” says Brian Faivre, brewmaster at Deschutes Brewery. “We’re no longer staffing the personnel to be able to operate 24/7. We’d normally have to make a sacrifice and historically that would be either quality or lost capacity or the happiness of our employees. … This is an opportunity to say we have strong confidence in this model.”

The framework for the predictive analysis has been built into all of the brewery’s 50 or so tanks, which range in capacity from 100 to 1,000 barrels—3,150 to 31,500 gallons. Currently, the actual shift in stages is done manually after brewers confirm the readiness of the brew, but Faivre says they company is looking to automate that.

The net effect of tapping data in the brewing process? Deschutes is able to reduce the fermentation process by 24 to 48 hours per batch. That gives the brewery a chance to increase its annual production without buying additional equipment.

So far, Deschutes stands alone in using sensors and data crunching to assist with the brewing, but Faivre says some other brewers, including Sierra Nevada, have visited to learn more.

The project has also spawned an open source data collection project among craft brewers, who share their historical records of how long it takes to move beer in production between the various stages.

“The majority of these folks, they might have sensors, but they write [the data] down on a piece of paper or in a spreadsheet,” says Faivre. “This is a way to structure and collect that data for them in a database and get them to a space where they’re building a database of that data so they can do things better in the future.”

He continued: “In the craft industry, to make that sort of adjustment is hard. Now, people are becoming comfortable with it. They see it as a tool, rather than something trying to take their job.”

With the reduction in work, Deschutes is now looking for new ways to leverage its data crunching tools. The usual practices, like predictive analysis that alerts when equipment is about to break, are being explored, but the company is also thinking about more industry-specific possibilities.

One of those could lie in using a mass spectrometer to measure flavor.

“We have all these recipes in the database,” says Faivre. “Right now, we have match data to these recipes, so we’re getting lab analysis so we’re able to get measurements of the various compounds in these. That’s where I want to go next, to take all the data, and really try to see if we can crack the nut of what combinations lead to these characteristics that consumers are so interested in that are so polarizing. There’s not an exact formula for figuring that out as a brewer.”

That analysis isn’t meant to remove the human element from developing beers. Rather, it’s meant to speed up the process. Sometimes before a brewer hits upon a successful formula, they’ll go through 100 or more iterations of a beer, looking for the exact flavor they want. Technology could help cut the number to 10.

Admittedly, for some brewers, that could take some of the fun out of the job. The best beer makers are about one-third mad scientist, experimenting with things you may never expect to be mixed with hops and malts—and often coming up with delicious new styles.

To ensure that tradition stays alive, Deschutes has a test plant that brews one barrel at a time, constantly experimenting and putting the results out in its tasting room to see how fans react to it. (Data on those small batch beers is recorded the same way it is in larger beers, in case the experiment is a resounding success.)

“Craft beer is something where so much of your soul and heart goes into it,” says Faivre. “So the concept of taking anything out of that can be a touchy subject. It’s one of those things where you have to build trust. You have to have the conversations. This is a tool. We’re not turning the way we make beer over to machines and engineers, but it’s a way to be more efficient.”