Marketers sometimes have to be creative to offer customers something new without the luxury of that new item being a brand-new product or built-from-scratch service. In fact, incrementally introducing features is familiar to marketers of consumer goods. But what features are consistently being selected by customers? What should marketers look for to tell development teams, production teams, and partners to plan upon? Applying a regression may help.

营销人员有时必须具有创造力，才能为客户提供一些新东西，而不能将新商品的奢侈性视为全新产品或从头开始的服务。 实际上，消费品的营销人员已经逐渐引入功能。 但是客户一直在选择哪些功能？ 营销人员应该寻找什么来告诉开发团队，生产团队和合作伙伴进行计划？ 应用回归可能会有所帮助。

It’s not an easy question to answer, even among startups that usually have a unique take on a market. Entrepreneurs other confuse features with products and services, leading to poorly crafted business models. Products and services solve problems for people. Features are characteristics that add value to the products.

即使在通常在市场上占有独特地位的初创企业中，这也不是一个容易回答的问题。 企业家将产品和服务的其他功能混为一谈，从而导致制作不当的商业模式。 产品和服务为人们解决问题。 功能是增加产品价值的特征。

The major challenge for managers is determining what features would add value in a sustainable way. It’s easy to brainstorm ideas about what features a product or service should contain. The real challenge however is selecting that features that will draw customers consistently. This is the essence of building a business model.

管理人员面临的主要挑战是确定哪些功能将以可持续的方式增加价值。 就产品或服务应包含的功能集思广益的想法很容易。 然而，真正的挑战是选择能够持续吸引客户的功能。 这是建立业务模型的本质。

The Chevrolet Corvette Grand Sport is an example of selecting features to produce a different product offering. The Grand Sport trim level was introduced in 2017 for the C7 Corvette. C7 is a nickname for the seventh generation of Corvette. The latest version, a mid-engine is a C8, was introduced in 2019.

雪佛兰Corvette Grand Sport是选择功能来生产其他产品的示例。 Grand Sport内饰级别于2017年为C7 Corvette引入。 C7是第七代Corvette的昵称。 最新版本是中置引擎C8，于2019年推出。

The Grand Sport is a hybrid of exterior and chassis features from a more expensive model, the Corvette Z06, with the engine and core interior features from the entry-level Corvette. The Grand Sport reached a 30% share of overall Corvette vehicle sales in its first year, so from an automotive industry perspective, it successfully contributed significant sales for the model.

Grand Sport混合了更昂贵的Corvette Z06的外观和底盘功能，以及入门级Corvette的发动机和核心内饰功能。 Grand Sport在第一年就占据了Corvette整车销售的30％，因此，从汽车行业的角度来看，它成功地为该车型贡献了可观的销售。

Combining features are not too complicated, a good brainstorming session with insightful customer comments can yield some ideas as to what features may be worth investigating. But you may want to see if those choices seem really correlated and sustainable. The last part is the most important. The last decision you want to make is one where features indeed seem correlated but are more of a blip rather than a predicted trend.

组合功能并不太复杂，一个很好的头脑风暴会议以及有见地的客户评论可以就可能值得研究的功能产生一些想法。 但是您可能想看看这些选择是否看起来确实相关且可持续。 最后一部分是最重要的。 您要做出的最后一个决定是，其中的功能看上去确实相关，但更多的是暂时的，而不是预测的趋势。

Regressions can back up a professional intuition by asking what sort of features are sustainable enough to support sales. A regression determines if data associated with the features reflect a relationship that represents sustainability. Regressions can add a logical influence on a business decision, especially on products that receive a ton of “fanboy” social media commentary but remain unclear if that interest turns into real sales. Fanboy comments on iconic sports cars like the Corvette are no exception.

回归可以通过询问什么样的功能足以支持销售来支持专业直觉。 回归确定与要素关联的数据是否反映了代表可持续性的关系。 回归可以对业务决策产生逻辑上的影响，尤其是对收到大量“粉丝”社交媒体评论但对这种兴趣是否转化为实际销售尚不清楚的产品。 Fanboy对像Corvette这样的标志性跑车的评论也不例外。

A useful first step in planning data is to consider a feature and how its observations are being measured. For example, if you want to compare engine size for a powered object like a lawnmower or motorcycle, then you would have a column with engine sizes for observations.

计划数据的有用的第一步是考虑一个特征以及如何测量其观测值。 例如，如果要比较割草机或摩托车等电动对象的发动机尺寸，则将有一个列，其中包含用于观察的发动机尺寸。

Determining a correlation is the next step. Gaining an idea of how the data roughly takes shape can be done in different ways. One quick-and-dirty way to create a scatter plot to see how the data takes a shape. A correlation is a more specific way to see if there is a linear correlation. The range of correlations is from -1 to 1. How close each value is calculated demonstrates the strength and direction of the correlation. You can then begin to examine a regression method that will likely provide the best fit that represents the correlation calculated.

确定相关性是下一步。 了解数据如何大致成形的想法可以通过不同的方式来完成。 一种创建散点图以查看数据如何成形的快捷方法。 相关性是查看是否存在线性相关性的更具体方法。 相关范围是-1至1。计算每个值的接近程度证明了相关的强度和方向。 然后，您可以开始研究一种回归方法，该方法可能会提供代表所计算相关性的最佳拟合。

A regression treats each feature as independent variables, then determines which one correlates to a dependent variable: the output we want. That output is usually sales, but the regression can also be created for a variation of sales, such as purchase orders or market share. Linear regressions are the simplest regression models. They can be made with just a few columns of data. They also assume a linear relationship between the outcome and the predictor variables.

回归将每个特征视为独立变量，然后确定哪个特征与因变量相关：我们想要的输出。 该输出通常是销售额，但是也可以为销售额的变化(例如采购订单或市场份额)创建回归。 线性回归是最简单的回归模型。 它们可以仅由几列数据组成。 他们还假设结果与预测变量之间存在线性关系。

For our Corvette example, a logistic regression can be useful. In a logistic regression, the dependent variable represents a binary choice (yes or no, approve or rejected). In this instance, it is easy to imagine columns of features and the dependent variable represents a binary choice, a sale or no sale. Dependent variables can also be represented as a lift in sales or sales decrease or an increase or decrease in market share.

对于我们的Corvette示例，逻辑回归可能会有用。 在逻辑回归中，因变量表示二元选择(是或否，批准或拒绝)。 在这种情况下，可以轻松想象特征列，并且因变量表示二进制选择，出售还是不出售。 因变量也可以表示为销售额上升或销售额下降或市场份额的上升或下降。

The method of regression can be influenced by the normality of the observations. The shape of the scatter plot can be a simple reveal. Does the data imply that a line best represents a regression, or is there a general curve that suggests a logistic regression? In most instances where the shape is not immediately obvious, you would apply a test such as Q-Q test.

回归方法可能会受到观测值的正态性影响。 散点图的形状可以很简单地显示出来。 数据是否暗示一条线最能代表回归，还是有一条总体曲线表明对数回归？ 在大多数情况下，形状不是立即显而易见的，您可以应用诸如QQ测试之类的测试。

No matter how you set up a regression, a good model reveals a strong correlation among the features represented as the influential independent variables, which means finding the right combination of features that would likely entice a sale. A regression model may not be needed in every instance of combining features. But they are valuable in instances of uncertainty, particularly as a service adds an offering.

无论您如何建立回归，一个好的模型都会揭示以有影响力的自变量表示的特征之间的强相关性，这意味着找到可能会吸引销售的正确特征组合。 在每个组合特征的实例中可能都不需要回归模型。 但是它们在不确定的情况下非常有价值，尤其是当服务添加产品时。

Overall regressions can provide a robust means to reinforce intuition on feature combinations or enlighten opportunities for new services. Either case can lead to better business choices for products and services that can incrementally increase sales while minimizing investment.

整体回归可以提供一种强有力的手段来增强功能组合的直觉性或启发新服务的机会。 不论哪种情况，都可以为产品和服务带来更好的业务选择，从而可以在不增加投资的情况下逐步增加销售额。