当大数据遇到啤酒行业，会如何带来创新？

2018年12月，美国第十大精酿啤酒厂德舒特酿酒公司宣布计划裁员10%。公司将其归咎于销售业绩和销量的下滑，而这也是精酿啤酒市场达到饱和后的常态。

对于大多数酿酒商来说，如此大规模的裁员还意味着销售和产量的削减。但德舒特并没有这个打算，因为就在四年前，公司决定在生产流程中整合接入互联网的传感器。

当然，啤酒酿造依然是一个以人力为中心的行业。传统上，工作人员在生产过程中会对啤酒进行手工选样和分析，来确定啤酒是否应该从一个酿造阶段转至下一个阶段，这一流程称之为阶段转移，而这样的阶段一共有9个。过早或过晚的转移都会影响成品啤酒的品质。

与众多酿酒商一样，德舒特总部位于俄勒冈州本特市，它一直保留着这些样本和分析记录。随后，公司决定发挥这些记录的效用，并利用微软和OSISoft，在云端对数据进行解析，从而预测生产过程中的转移次数。由此得出的结论帮助简化了酿造流程，为公司带来了事半功倍的效果。

德舒特公司的酿酒师布莱恩·法芙芮说：“当出现产量问题或裁员时，公司不大愿意投资其他的资产。我们已经不再为全天候运营而招聘新人。通常，我们不得不做出牺牲，而且以往都是以质量、产能或员工的幸福为代价……如今我们可以说的是，我们对这一模式充满信心。”

预测分析框架已经被植入到公司的所有约50个酿酒罐，其容量从100桶到1000桶（3150加仑至3.15万加仑）不等。当前，公司在酿酒师确认某一酿造阶段完成之后才会通过人力进行阶段转移，但法芙芮称，公司正在探索实现这一流程的自动化。

酿造过程数据分析都会带来什么净效应？德舒特将每批啤酒的发酵流程时长降至48个小时，较之前减少了24个小时。这也让公司能够在不购买额外设备的情况下提升其年产量。

到目前为止，德舒特是唯一一家使用传感器和数据分析来协助酿造的公司，但法芙芮表示，像Sierra Nevada这样的一些酿酒商也慕名前来了解这一模式。

该项目还催生了一个精酿啤酒酿造商开源数据搜集项目，这些啤酒商会分享其啤酒酿造期间各个阶段转移时长的历史记录。

法芙芮说：“大多数公司可能都有传感器，但他们将数据记录在纸上或电子表格中。我们的这一举措可以帮助他们搜集数据并构建一个数据库，然后为其提供一个空间，供它们打造上述数据的数据库，这样，它们便可以在未来改进其生产。”

他继续说道：“在精酿行业，做出这种调整是困难的。如今，人们逐渐适应了这种做法。他们将其看作是一种工具，而不是抢夺其饭碗的事物。”

随着工作量的减少，德舒特如今正在寻找新的方式来利用其数据分析工具。在探索将其运用到日常操作的同时，例如在设备即将破损时用于预警的预测分析，公司还在思考更具行业针对性的运用方式。

其中的一个便是使用光谱仪来测量啤酒风味。

法芙芮说：“我们将所有配方放在数据库中。如今，我们为这些配方找到了匹配的数据，因此我们将进行试验分析，以便对啤酒中的各类化合物进行测量。这便是我下一步要做的事情，也就是利用这些数据来努力尝试是否能够找到让啤酒出现某种特质的化合物搭配比例，而正是因为这些特征，消费者才会对我们的啤酒感兴趣，我们的啤酒才能与众不同。”

这种分析并非意味着取消啤酒酿造中的人力因素，它更像是对酿造流程的加速。有时候，在酿酒公司发现一个成功的配方时，它们已经反复进行了100多次或更多次数的酿造，为的是寻找它们所追求的特定风味。然而借助科技，测试次数可以降至10次。

不可否认，对于一些酿酒公司来说，此举会让酿酒工作失去一些乐趣。最好的啤酒酿造商有三分之一归功于其疯狂的科学实验，它们会尝试用各种超乎人们想象的事物与啤酒花和麦芽进行混合，而且通常会催生出可口的新风味。

为了确保这个传统一直延续下去，德舒特设立了一家测试工厂，每一次仅酿造一桶啤酒，并不断地进行试验，然后通过其品尝室获得反馈，从而了解啤酒拥趸对新风味的反响。（这些小批量啤酒数据的记录方式与量产啤酒无异，说不定某一个试验就会大获成功。）

法芙芮表示：“精酿啤酒需要人们倾注大量的心血和精力。因此，将这些内容从啤酒中剥离开来是一个敏感的话题。在这一方面，公司必须创建信任，并进行对话。这一举措是一个工具。我们并不打算将啤酒酿造交给机器和工程师来做，它只是提高酿酒效率的一个手段罢了。”（财富中文网）

译者：冯丰

审校：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.”