4分钟阅读

需求预测的重新发明:当你不能回顾过去时,为未来做计划

埃文Quasney

全球供应链解决方案副总裁

Volatility, uncertainty, complexity and ambiguity aren’t new dynamics for supply chains. The alarm bells have been ringing steadily in recent years, as evolving consumer behavior, new competition, and channel proliferation have made demand management more challenging than ever. Today, the alarm bells are deafening and impossible to ignore, as the outbreak of the novel coronavirus has upended life and business as we know it. Here’s the good news: most companies now have an undeniable call to action to transform their supply chains and move past the traditional demand planning and forecasting approach. Focusing on integrated demand management, which blends sales and finance functions with improvements in supply chain, creates a path forward for companies to achieve new levels of transparency, collaboration, and ultimately business performance, enabling them to face market challenges with new levels of agility and resilience.

\n

The best predictor of future events isn’t necessarily past events

\n

Historic performance has long been the cornerstone of traditional demand planning and forecasting. Looking at year-over-year or quarter-over-quarter sales volumes and assuming some growth in statistical models serves as the primary forecasting approach for many companies.

\n

The COVID-19 outbreak has rendered that approach useless, sending shockwaves throughout worldwide supply chains. The efficacy of traditional demand planning in our modern, fast-changing business environment had been waning for some time as advances in collaboration, analytics, and process management have given life to broader forecast inputs. But the pandemic lays bare the limitations of traditional demand planning based on historical statistical modeling and patterns and has left many users of traditional models struggling to make sense of their path forward.

\n

The diminishing value of traditional demand planning and history-based statistical modeling underscores the need for a new approach for predicting the future, which requires planning that integrates market intelligence, advanced algorithms, and internal and external collaboration across multiple processes.

\n

Bringing together external and internal signals

\n

As we look at where we’re going now, it’s critical for companies to assess not only their forecasting process, but also the external market forces that drive demand. Combining the process, and internal and external data, in a modeling platform gives the insights necessary to better understand where the market is going and how to respond with maximum profitability and service. While historical signals will always maintain some level of utility, market-based forecasting minimizes historical data as the overriding signal and instead incorporates many input signals to create a robust, multi-faceted picture to sense demand over multiple horizons—short-term through long-term.

\n

So, what are the input signals that companies should focus on? Answering this question starts with understanding both the independent and dependent drivers of demand. External data like weather, seasonal travel patterns, foot traffic, or pollen count can be incorporated into a forecasting platform to understand how external factors impact consumer behavior, which, in turn, affects demand. Trends from social media platforms like Facebook and Instagram deliver a real-time pulse on the market based on topics discussed and trending hashtags, creating an opportunity to drive sentiment data into a forecast. In the current environment, public health data, overlaid with these inputs, becomes a critical driver for organizations of all types to best assess the product mix and volume that should be offered.

\n

Additional “owned” inputs from all key stakeholders in the supply chain, such as manufacturer incentives, balance of trade, competitor products and pricing, or end-distributor promotions, serve as additional valuable data points to incorporate into an integrated revenue management solution to understand the impact on product quantity, as well as net price paid. Combining these internal and external data sets in a common demand signal repository and then applying advanced algorithms—such as artificial intelligence or machine learning—to evaluate which demand drivers will influence volume and price adds further richness to the forecast picture and enables more prescriptive decision making.

\n

Layer forecast process atop demand signals

\n

More often than not, external factors and indicators become powerful signals—but how are those signals consumed by the organization to arrive at a final plan? This is where the forecast process itself becomes critical. Oftentimes, forecast process is misconstrued as the “statistics” part before consensus demand and adjustment. Integrated demand management takes a different approach. It starts with looking at the performance and impact of each substep in forecasting (e.g. segmentation, naïve forecast, unconstrained algorithmic forecast, over-ride, promotion lift, etc.) and methods like forecast value added (FVA), to measure incremental performance of specific steps of the forecast process for that product and location combination over time.

\n

Using insights from FVA, integrated demand management then assesses performance of each step, identifies areas of waste, and re-orders some of these activities to increase productivity of commercial and volume planning teams, giving confidence to business decision makers based on using a single source of truth. As a result, planners spend less time processing data and can pivot to become more focused on external drivers, insights, and market intelligence. In this role, they can drive strategic, commercial direction into decisions and demand in the volume and business forecasting process.

\n

The future of demand forecasting starts now

\n

Successful demand forecasting requires a platform that enables companies to act with agility, recalibrate rapidly and most importantly, incorporate different scenarios as needed to ensure increased forecast accuracy delivers the ultimate business outcomes. This relates back to the business’ financial and revenue plan because improving forecast accuracy consistently over time drives the kind of profitability that instills confidence in board members and business leaders. Leveraging a platform that supports strategic decisions gives financial leaders the predictability and visibility they need to navigate uncertainty and position their business to accelerate out of the curve in the next phase.

\n

\n"}}" id="text-676387d8a9" class="cmp-text">

波动性、不确定性、复杂性和模糊性并不是供应链的新动态。近年来,随着消费者行为的演变、新的竞争和渠道的扩散,需求管理比以往任何时候都更具挑战性,警钟一直在敲响。今天,随着新型冠状病毒的爆发颠覆了我们所知的生活和商业,警钟震耳欲聋,不容忽视。这里有一个好消息:大多数公司现在都有一个不可否认的呼吁,要改变他们的供应链,摆脱传统的需求规划和预测方法。专注于集成的需求管理,将销售和财务功能与供应链的改进相结合,为公司实现新的透明度、协作和最终的业务绩效创造了一条前进的道路,使他们能够以新的敏捷性和弹性水平面对市场挑战。

未来事件的最佳预测者不一定是过去的事件

长期以来,历史业绩一直是传统需求规划和预测的基石。对许多公司来说,主要的预测方法是查看同比或季度同比的销售额,并假设统计模型有一定的增长。

COVID-19的爆发使这种方法变得无用,在全球供应链中引发了冲击波。随着协作、分析和流程管理方面的进步赋予了更广泛的预测输入以生命,传统需求计划在我们现代、快速变化的业务环境中的有效性已经减弱了一段时间。但这场大流行暴露了基于历史统计模型和模式的传统需求规划的局限性,并使许多传统模型的用户难以理解他们的前进道路。

传统的需求规划和基于历史的统计建模的价值日益减少,这强调了对预测未来的新方法的需求,这需要将市场情报、高级算法以及跨多个过程的内部和外部协作集成在一起的规划。

将外部和内部信号结合在一起

当我们审视我们现在的发展方向时,公司不仅要评估他们的预测过程,还要评估驱动需求的外部市场力量,这一点至关重要。在建模平台中结合流程、内部和外部数据,可以提供必要的见解,以更好地了解市场走向,以及如何以最大的盈利能力和服务做出响应。虽然历史信号将始终保持一定程度的效用,但基于市场的预测将历史数据作为压倒一切的信号,并结合许多输入信号,创建一个强大的、多方面的画面,以感知多个水平(短期到长期)的需求。

那么,企业应该关注哪些输入信号呢?要回答这个问题,首先要了解需求的独立驱动因素和依赖驱动因素。天气、季节性旅行模式、人流量或花粉计数等外部数据可以纳入预测平台,以了解外部因素如何影响消费者行为,进而影响需求。Facebook和Instagram等社交媒体平台的趋势根据讨论的话题和趋势标签提供了实时的市场脉搏,创造了将情绪数据转化为预测的机会。在当前环境下,与这些投入相结合的公共卫生数据成为所有类型的组织最好地评估应提供的产品组合和数量的关键驱动因素。

来自供应链中所有关键利益相关者的额外“自有”输入,如制造商激励、贸易平衡、竞争对手产品和定价,或终端分销商促销,作为附加的有价值的数据点,纳入综合收益管理解决方案,以了解对产品数量的影响,以及支付的净价格。将这些内部和外部数据集整合到一个共同的需求信号库中,然后应用先进的算法(如人工智能或机器学习)来评估哪些需求驱动因素将影响数量和价格,从而进一步丰富预测图景,并使决策更具规范性。

需求信号之上的层预测过程

通常情况下,外部因素和指标会成为强有力的信号——但组织如何利用这些信号来达成最终计划呢?这就是预测过程本身变得至关重要的地方。通常,预测过程被误解为共识需求和调整之前的“统计”部分。集成需求管理采用不同的方法。它首先着眼于预测中每个子步骤的性能和影响(例如细分,naïve预测,无约束算法预测,覆盖,促销提升等)和预测附加值(FVA)等方法,以衡量该产品和位置组合的预测过程中特定步骤的增量性能。

使用来自FVA的见解,集成需求管理然后评估每个步骤的性能,确定浪费的区域,并重新安排这些活动中的一些,以提高商业和批量计划团队的生产力,基于使用单一的事实来源,给业务决策者信心。因此,规划人员在处理数据上花费的时间更少,可以转向更关注外部驱动因素、洞察力和市场情报。在这个角色中,他们可以在数量和业务预测过程中推动战略和商业方向的决策和需求。

需求预测的未来从现在开始

成功的需求预测需要一个平台,使公司能够敏捷地行动,快速地重新校准,最重要的是,根据需要结合不同的场景,以确保提高预测的准确性,从而实现最终的业务成果。这与企业的财务和收入计划有关,因为随着时间的推移,不断提高预测的准确性会推动盈利能力,从而给董事会成员和企业领导者灌输信心。利用一个支持战略决策的平台,为财务领导者提供了他们需要的可预测性和可见性,以导航不确定性,并定位他们的业务,以便在下一阶段加速走出曲线。

探索商业的未来,装备你的组织走向成功