In Clear Focus: Demystifying Marketing Mix Modeling

接下来是本期《清晰聚焦》的内容。

Coming up in this episode of In Clear Focus.

我们做的和传统意义上的营销组合咨询有很大不同。

There's a big difference between what was traditionally marketing mix consulting and what we do.

无论是营销团队还是代理合作伙伴，都把这套工具当成一个能看懂、能模拟结果、能执行计划，并通过系统验证最初方案的规划工具。

The use case for both marketing teams and agency partners is that they're using this as a planning tool that they can understand and simulate outcomes and then execute those plans and be able to validate the original plans through the system.

您正在收听《清晰聚焦》，每周由Big Eye为您带来广告行业的全新视角。Big Eye是一家全方位创意机构，致力于为全球客户打造强势品牌。

You're listening to In Clear Focus, fresh perspectives on the business of advertising, produced weekly by Big Eye, a strategy alert full service creative agency growing brands for clients globally.

你好。

Hello.

我是主持人阿德里安·坦南特，首席战略官。

I'm your host, Adrian Tennant, chief strategy officer.

感谢您的收听。

Thank you for joining us.

如今的媒体环境越来越复杂，营销人员可选择的线上和线下渠道多得数不清。

Today's media environment is increasingly complex with a myriad of online and offline channels available to marketers.

理解混合型消费者旅程，以及零售媒体等新兴渠道的广告效果，是一个巨大的挑战。

Understanding hybrid consumer journeys and the effects of advertising in newer channels like retail media is a significant challenge.

经济不确定性、更严格的隐私法规，还有第三方cookie的消亡，都让一种分析工具重新受到关注，也就是营销组合模型，简称MMM。

Economic uncertainty, tighter privacy regulations, and the death of the cookie have all renewed interest in one analytical tool in particular, marketing mix modeling or for short.

它能整合来自多个渠道的数据，提供历史分析和未来营销规划的情景模拟，帮助营销人员从整体上规划和优化媒体投放。

It promises to provide marketers with a holistic approach to planning and media optimization by combining data from multiple sources to generate historical analysis and future marketing planning scenarios.

今天我们邀请到格雷格·多兰，他是Keen Decision Systems的联合创始人兼首席执行官。这是一家入选《财富》美国最快增长企业5000强的SaaS公司，其统一的测量与优化平台，帮助品牌方做出数据驱动的决策。

To discuss marketing mix modeling, we're joined today by Greg Dolan, the co founder and CEO of Keen Decision Systems, an Inc 5,000 software as a service company whose unified measurement and optimization platform enables brand marketers to make data driven decisions.

在联合创立Keen之前，格雷格曾在多家快消品企业担任品牌和企业战略相关工作，包括卡夫食品、坎贝尔汤公司、纳贝斯克和亿滋国际。

Prior to co founding Keen, Greg had brand and corporate strategy experience at CPG companies including Kraft Foods, Campbell Soup Company, Nabisco, and Mondelez.

他经常为行业研究报告撰稿，同时也是北卡罗来纳大学凯南-弗拉格勒商学院的客座讲师，给MBA学生讲授高级数据分析课程。

He's a frequent contributor to research publications and a guest lecturer at UNC Keenan Flagler Business School teaching MBA students advanced analytics.

今天，我们请格雷格来聊聊，营销组合模型如何帮助品牌评估营销投入的效果与效率。他现在正从位于北卡罗来纳州达勒姆的Keen Decision Systems办公室连线加入我们。

To discuss how marketing mix modeling can help brands understand the effectiveness and efficiency of their marketing investments, Greg is joining us today from Keen Decision Systems offices in Durham, North Carolina.

格雷格，欢迎来到《清晰聚焦》。

Greg, welcome to In Clear Focus.

谢谢，阿德里安。

Thanks Adrian.

很高兴能和你们交流。

It's great to be with you.

我在开场时简单提了一下，你能再多讲讲你创立Keen之前的经历吗？

I gave a brief description in the intro, but can you tell us more about your background before co founding Keen Decision Systems?

当然可以。

Absolutely.

我的职业背景其实直接促成了Keen的诞生。

My background really led to the creation of Keen.

我最初是做快消品品牌营销的。

So I started my career as a consumer packaged goods brand marketer.

后来在纳贝斯克、卡夫食品，还有金宝汤公司这些食品企业工作过。

So I've worked in the food industry at Nabisco and then Kraft Foods and the Campbell Soup Company.

我主要负责整体管理，包括品牌盈亏和预算，最大的增长杠杆就是营销。

And I was in primarily general management focused, p and l management of brands and I had responsibility for large marketing budgets and my biggest lever in generating profitable demand was marketing.

可惜的是，我那时用的还是传统营销组合模型，每年才做一次分析，而且数据交给咨询公司六个月后才拿到报告，内容只是回顾过去两年发生了什么，根本没法告诉我接下来该怎么做。

And unfortunately, I had old fashioned marketing mix where I had a marketing mix study that was delivered once a year about six months after the data was delivered to the consultant and they gave me a report card of what happened in the previous two years, but it didn't give me a good understanding of what I do next.

作为负责盈亏的管理者，我必须搞清楚，怎么根据手头的信息做出更好的决策。

And as a P and L owner, as someone that was managing a business, I really had to understand how do I make better decisions based on the information I have.

幸运的是，我遇到了我的技术联合创始人约翰·布斯比，他之前一直在为制药行业做营销组合分析，但他早就厌倦了花几个月做一份两百页的PPT，最后却没人拿它来做实际决策。

So I was lucky enough to meet my technical co founder John Busbis, who had been doing marketing mix analytics for the pharmaceutical industry and he was tired of putting together the marketing mix presentation, 200 page PowerPoint deck and delivering it to marketers for it not to be used for decision making.

所以我们觉得，这里头有颠覆市场的机会——开发新一代的软件解决方案，不只关注效果衡量，更重要的是，能在信息一出来就引导决策，直接跟品牌所承担的财务结果挂钩。

So we felt together there was an opportunity to disrupt the market, develop a next generation solution that was delivered through software and not only really focused on just measurement, but more importantly started to guide decision making as information became available and really tie that directly to the financial outcomes that the brand was being held accountable for.

在当今经济前景依然不明朗的背景下，你最近在Quirks上写了一篇文章，批评企业盲目削减营销和广告预算的做法。

Well, in the context of today's still uncertain economic outlook, an article you wrote for Quirks recently discussed the shortsightedness of companies slashing their marketing and advertising spending.

格雷格，无论经济环境如何变化，为什么坚持长期品牌建设这么重要？

Greg, why is it important to maintain long term branding efforts regardless of the prevailing economic conditions?

作为营销人，我们知道，营销带来的价值是随着时间积累的。

Well, as marketers, we know that marketing drives value and value over time.

但我觉得，很多传统的营销分析恰恰忽略了这一点。

And I think that this is missed in a lot of traditional marketing analytics.

所以我们这套系统、这套软件能做到的，就是把营销投放的时间差影响纳入计算。

So what we've been able to do as a system and as a software application is account for the timing impact of marketing.

我们在系统里内置了长期效果模型，能清晰追踪每一笔营销投入分别对短期营收和长期价值产生了什么影响。

So in our system, we were able to build in long term effects and really understand how each dollar of marketing is impacting both short term revenue and long term value.

我们研究发现，专注研究营销连续性的亚伦·伯巴斯等学者也都认同这个结论：如果砍掉营销预算、停止投放，甚至改成断断续续的脉冲式投放，品牌就会彻底淡出消费者的视线。

And what we found was that, and this is supported by Aaron Burbas and others who really focus on continuity of marketing, is that when you cut marketing and you go dark and even flight the business in intervals, then you're not in front of the consumer.

之前投到市场里的所有营销积累都会开始贬值，之后你得花更多钱、花更长时间，才能把原来的业务规模重新做回来。

And that marketing that you have put into the marketplace and deployed starts to decay and it takes that much more spend and time to get the same amount of volume back.

要是你彻底停掉投入、削减预算、不再为消费者投入资源，之后再想重新做起来，长期来看成本只会高得多。

So it's more expensive over time to build back after you've gone dark and cut budgets and not invested in the consumer.

我刚才开头的描述可能太简单了点。

I probably gave an overly simplistic description in the intro.

那能不能麻烦你再详细解释一下，什么是营销组合模型？它和品牌经理现在已经从内部营销团队或者合作agency拿到的各类研究报告，有什么不一样？

So could you explain in a bit more detail what marketing mix modeling is and how it differs from the kinds of research and reporting that brand managers are likely already receiving from their marketing teams or agency partners?

好的。

Yes.

传统上的营销组合咨询和我们做的东西有很大区别。

So there's a big difference between what was traditionally marketing mix consulting and what we do.

过去我也用过传统的营销组合方法。

In the past and again, I was a consumer of traditional marketing mix.

每年你都会和顾问一起做一次分析，把你的财务数据、所有营销活动数据——包括电视、广播、数字渠道，还有快消品行业的渠道投入——全交给他们。

Every single year you would go through an exercise with a consultant where you would deliver to them your financial data, all of your marketing activity data across your marketing mix, your TV data, radio, all your digital, your trade for the CPG world.

这些顾问会为你搭建一个定制的分析模型，主要是时间序列回归。

And those consultants would build a custom analytics model, primarily time series regression.

他们分析的是你的所有营销投入和销售业绩之间的相关性。

So they're looking at the correlation between all of your activity and spend and your sales.

然后他们会用一份PPT给你出报告，告诉你过去两年品牌或业务的表现如何。

And they would deliver that analysis in a PowerPoint presentation that basically gave you a report card for, you know, how the brand had or business had done over the past two years.

这种分析每年都要重新做一遍，每次都从零开始。

And that is an exercise that's recreated and basically done over every single year created from scratch.

它确实能帮你了解过去哪些做法有效，但就像我说的，对决定下一步该怎么做，帮助就有限了。

So it was helpful in terms of, hey, I I have an understanding of what worked in the past, but it wasn't, as I mentioned, as helpful when it came to making the next decision, understanding what you do next.

这正是我们想要改变的地方。

And that's really what we are looking to disrupt.

过去，品牌经理用这份报告来验证过去发生了什么。

So the brand manager in the past would use that as validation for what happened in the past.

我们想做的，是把重点从回顾决策转向指导决策。

What we're trying to do is really pivot that from measurements of decision making.

在这个环境更复杂、渠道更多、组织内部更加割裂的时代，做营销规划时，拥有全渠道的视角至关重要。

So we know in this world where it's a lot more complex, there's a lot more channels, a lot of silos within organizations that it's important to have an omnichannel perspective when it comes to planning your marketing.

所以我们致力于整合所有能获取到的信息。

So we are focused on bringing in all the information that we can bring to bear.

这包括传统营销组合分析中常用的那些时间序列数据。

So that's a combination of the traditional time series data that you would give to a traditional marketing mix.

但我们会把这些数据，和我们专有的营销弹性引擎所积累的行业知识结合起来——这套引擎是模型的基础，同时也会整合公司内部其他所有分析成果，比如多触点归因、广告文案效果研究、A/B测试等等，只要是能帮上忙的信息，我们都用上。

But combine that with industry knowledge through our proprietary marketing elasticity engine, which serves as information base for our models, but also any other analytics that's being done across the organization whether that's MTA or copy effectiveness studies or AB tests or any of that information that can also inform the model.

我们把所有相关的信息都整合进来，不仅是为了搞清楚过去发生了什么，更重要的是，能实时指导下一步行动。

So we're bringing the full knowledge of state to the table in order to inform not only what happened in the past, but more importantly on a real time basis.

我该怎么做出下一个决策？

How do I make the next decision?

下一美元投下去，哪儿能带来最大回报？

Where am I gonna get the biggest return on the next dollar invested?

我该怎么持续提升品牌未来的表现？

And how do I show continuous improvement for the brand looking forward?

所以对营销团队和代理合作伙伴来说，这个工具能帮助他们制定计划、模拟结果、执行计划，并通过系统验证最初的方案。

So the use case there for both marketing teams and agency partners is that they're using this as a planning tool that they can understand and simulate outcomes and then execute those plans and be able to validate the original plans through the system.

这其实是一个持续规划、执行与验证的系统。

So it's a continuous planning execution and validation system.

你认为品牌营销人员通常是什么原因开始寻求营销组合建模方案的？

What typically prompts brand marketers to seek out a marketing mix modeling solution do you think?

现在要搞清楚营销的各个部分，以及如何平衡漏斗顶部、中部和底部，变得越来越复杂了。

Well, it's becoming a lot more complex to understand all the different pieces of marketing and how do you balance top of the funnel, mid funnel and bottom of the funnel.

我们发现，要衡量影响效果，最成功的还是在漏斗底部，也就是那些直接带动收入的交易环节。

I think what we've seen is that where we've had the best luck understanding impact is at the bottom of the funnel and those revenue driving transactions.

所以我们知道，整个营销漏斗中存在大量复杂性和碎片化问题。

So what we know is that there's a lot of complexity, a lot of fragmentation across the funnel.

在漏斗顶部保持平衡，和在底部大力投入一样重要。

It's as important to have balance at the top of the funnel as it is to have a heavy investment at the bottom of the funnel.

因此，品牌经理和营销人员普遍希望了解营销组合对财务表现的全面影响，而这一点长期以来一直是营销人员难以证明的。

So brand managers and marketers in general are looking to understand the full impact of the marketing mix on financial performance and that's been largely something that marketers haven't been able to demonstrate.

这也是为什么首席营销官在高管层的任期往往很短的原因。

And that's one of the reasons why CMOs have a short tenure in the c suite.

他们无法以可问责的方式，清晰展示自己在营销组合上的各项决策，对实现财务目标所产生的实际影响。

They're not able to demonstrate with accountability the impact of the decisions that they're making across the marketing mix in service of those financial objectives they're looking to achieve.

所以关键是要填补这个营销效果的证明缺口——弄清楚我把每一美元投在各个渠道上，到底为公司创造了多少价值，以及在收入层面的短期效果如何。

So really closing that marketing proof gap, being able to understand, I'm putting a dollar in all these channels, what does it mean for the value that I'm driving for the organization, as well as the short term impact from a revenue perspective.

MMM最常见的使用场景有哪些？

What are some of the most common use cases for MMM?

最主要的应用场景是，不过我们的做法有点不一样。我们从决策的角度出发，和客户及代理机构一起，构建一个围绕他们需要做哪些决策的决策框架。

The primary use case and it's a little bit different for the way we approach it, so we approach it from a decision making perspective and we're working with clients and agencies to develop a decision frame around what decisions they're looking to make.

最好的理解方式是把它看作一个层级结构。

And the best way to think about that is in terms of a hierarchy.

如果我是这家公司的首席营销官，旗下有多个品牌，我首先要决定的是，该给品牌A和品牌B各自分配多少预算。

So if I'm the CMO of an organization I have a lot of different brands across my portfolio, the first decision I want to make is understanding how much do I put behind brand a versus brand b.

然后我们再往下细化到更具体的层面。

And then we can get to lower levels of granularity.

比如针对品牌A，我可能想弄清楚子品牌A和子品牌B之间的投入分配，再进一步看每个品牌内部各个渠道的投放情况。

So for brand a, maybe I want to understand sub brand a versus sub brand b and then across all the channels within each one of those brands.

这样就能逐步深入理解，所有营销手段和渠道是如何共同推动财务成果的。

So it becomes a deeper and deeper understanding of how all of the marketing pieces, all the marketing channels are contributing to the financial outcomes.

对我们来说，最主要的应用场景就是从品牌层面出发，以全渠道视角来分析营销组合的整体影响。

And that's really the primary use case for us is that in a brand based macro down perspective around marketing mix from an omnichannel perspective.

当然，我们也能深入到更细致的场景中。

But then we can get to more granular use cases.

比如现在快消品行业一个非常热门的应用，就是研究零售媒体——它如何在不同零售商之间发挥作用，与消费者营销和渠道促销产生怎样的互动，以及在沃尔玛投入零售媒体广告，会对塔吉特或其他零售商产生什么影响。

So an example would be a huge use case in CPG right now is understanding retail media and how that plays across retailers, interaction effects with shopper marketing and trade, and how potential retail media investment in say a Walmart impacts Target and other retailers.

所以我们能在决策发生的层面建立模型，提供独特的视角来优化这些资源。

So we're able to build models at the level where decisions are being made and give unique perspective on how to optimize those resources.

传统上，营销组合建模是由一些大型知名营销和消费者行为测量公司提供的，客户往往要等上好几个月才能看到模型。

Marketing mix modeling has traditionally been offered by large well known marketing and shopper measurement companies, and it can often be several months before clients are presented with a model.

这类项目通常还需要投入六位数的资金。

These engagements also typically require a 6 figure investment.

格雷格，你的创新方法大大缩短了传统流程的时间，同时通过一套能提供实时洞察的系统降低了成本。

Greg, your disruptive approach considerably compresses the traditional timeline and reduces the cost with a system designed to provide real time insights.

你是怎么做到这一点的？

How do you achieve that?

秘密就在于我们的软件。

Well, the secret software.

我们很早就意识到，市场变化的速度远超人类能跟上的节奏。

So we realized early on that the market was moving a lot faster than a human could actually inform decisions.

我九十年代末刚入行做营销的时候，那时候主要就是选电视、广播还是印刷媒体。

Back when I started my marketing career in the late nineties, you know, had decided between TV, radio, and print.

对吧？

Right?

当时就那么几个渠道。

So there were a couple channels.

我们可以花时间慢慢来。

We could take the time.

那时候采购都要提前很久，对吧？

There were long lead time buys, right?

很多都是预付的。

A lot of upfronts.

我们那时候就是这么做的。

And we were able to do that.

一年出一份报告完全没问题。

And it was perfectly fine to have a report that was provided once a year.

后来数字媒体越来越普及，渠道也越来越分散，我们才开始做快消品领域的购物者营销，再后来就是零售媒体了。

And as digital became more prevalent and that fragmentation started happening, then we got into shopper marketing on the CPG side, and then there's retail media.

碎片化越来越严重，购买周期也越来越短。

There's more and more fragmentation and the buying cycles are shorter.

如果没有技术，没有足够的数据支持，你根本没法做决策，也搞不清效果。

So you can't make decisions or understand impact without technology and without the right amount of data to be able to support that.

如果你只靠纯咨询模式，成本根本扛不住，就像你提到的，每个模型费用都超过六位数。

And if you're looking at a pure consulting model, it's cost prohibited because each model is very expensive to your point, north of 6 figures.

所以技术必须得提升效率。

So technology had to drive efficiency there.

我们打造了一个模型，能实时抓取、整合和映射信息，点一下按钮就能生成模型，还能基于最新数据建立历史视角，为规划优化提供基础，进而向前预测，构建预测性计划——整个过程都嵌入了机器学习和人工智能，让决策更智能。

So we were able to build a model that leverages information, adjusts and maps information in real time, builds models with the click of a button, and then you have a historical perspective that's based on the latest information and a foundation then to run planning optimizations and look forward and start to build prescriptive plans based on the latest information based on a learning model that embeds machine learning and AI into that decision making process.

这一切都靠技术实现，能把随时出现的数据快速消化、转化成可执行的方案，让营销人员能直接落地。

So really all enabled by technology and the ability to quickly ingest the map data when it's available into something that's actionable that the marketer can though go execute.

基恩最近发布了一项研究，揭示了传统电视广告最有效的投放策略。

Keen recently published details of a study that reveals the most effective flighting strategy for linear TV advertising.

格雷格，能跟我们分享一下你们的发现吗？

Greg, could you share with us what you learned?

没错。

Absolutely.

这在我们系统的数学逻辑里是基础。

It's fundamental to the math in our system.

就像我之前说的，我们的模型已经考虑了长期影响。

So as I was talking about before, know, we've accounted for long term effects in our models.

在我们的营销弹性引擎中，每个营销渠道都有各自不同的衰减率。

So in our marketing elasticity engine, we have different decay rates for every single marketing channel you can think of.

比如电视、广播、印刷广告、优惠券，还有那些短期促进销售的手段。

So for TV versus radio print versus couponing and some of the shorter transaction driving tactics as well.

基于这些，我们构建了每周持续向前的响应曲线。

So we've built based on that a response curve weekly into the future.

我们能精确判断，无论是内部还是外部因素，每个渠道——包括电视——的回报递减点在哪里。

So we know exactly based on both internal and external factors, what the point of diminishing returns are for a given channel, including TV.

因此，我们能清楚看到，如果广告投放突然停止，营销人员会损失多少收入和利润。

So we can see what the marketer is giving up from a revenue and profitability standpoint by going dark.

对吧？

Right?

如果你是隔一段时间投放广告，比如播两周停两周，或者播三周停三周，那停播的那三周其实会错失掉本可以带来的盈利需求。

So if you're flighting and you're two weeks on, two weeks off, or three weeks on, three weeks off, those three weeks that you're off, actually, there's an opportunity cost that's created where you're giving up profitable demand.

我们已经证明了持续投放的力量，以及它对长期价值积累的影响。

So we've been able to show the power of continuity and the impact that has on building value over time.

就像我之前说的，一旦停播，重新启动广告、再次触达消费者，并让投入在短期内产生盈利效果，成本会高得多。

And as I was talking about before, when you are dark, it's more costly to then get back on, get in front of consumers and have that spend drive profitable demand in the short term.

所以关键就在于保持活动的连续性，这样才能在广告效果自然衰减的过程中持续积累价值。

So really, it is all about driving continuity of activity as a means to continue to build as marketing decays over time.

我们先休息一下。

Let's take a short break.

广告结束后马上回来。

We'll be right back after this message.

你好。

Hi there.

我是莉迪亚·迈克尔，《品牌之爱：构建强大的消费者品牌连接》的作者。

I'm Lydia Michael, the author of Brand Love, Building Strong Consumer Brand Connections.

基于我作为多元文化市场营销与品牌策略师的经验，这本书适合所有从事营销、品牌工作的专业人士、创业者，以及负责品牌传播和消费者策略的人士。

Reflecting my experience as a multicultural marketing and brand strategist, Brand Love is for any marketing and brand professionals, entrepreneurs, and those who oversee the brand messaging, communications, and other consumer facing strategies.

无论你为大型品牌还是小型品牌工作，这本书都能为你提供切实可行的策略，帮助你打造和壮大任何成功品牌。

Whether you work for a big or small brand, the book is designed to provide you with actionable strategies to grow and build any successful brand.

作为《In Clair Focus》的听众，你可以使用专属优惠码 BIG I 25，享受《品牌之爱》纸质版或电子版25%的折扣。

As an In Clair Focus listener, you can save 25% on a print or electronic version of Brand Love by using the exclusive promo code BIG I 25.

这个优惠码适用于所有产品和预售订单，并可叠加Kogan Page的免费平装书与电子书组合优惠。

This code is valid for all products and preorders and applies to Kogan Page's free paperback and ebook bundle offer.

直接从Kogan Page下单，美国和英国地区一律免运费。

When you order directly from Kogan Page, shipping is always free to The US and The UK.

要购买《品牌之爱》，请访问 koganpage.com。

So to order your copy of Brand Love, go to koganpage.com.

谢谢大家。

And thank you.

欢迎回来。

Welcome back.

我正在和格雷格·多兰交谈，他是Keen Decision Systems的联合创始人兼首席执行官。

I'm talking with Greg Dolan, the cofounder and CEO of Keen Decision Systems.

现在人工智能话题很热，但在Keen Decision Systems，你们已经用了好几年了。

Artificial intelligence is currently very much in the news, but at Keen Decision Systems, you've been using it for several years.

格雷格，机器学习的加入对你们平台的预测和建议性规划有什么影响？

Greg, how does the inclusion of machine learning impact your platform's predictive and prescriptive planning outputs?

我们把自己看作是一个AI决策支持助手。

So we think about ourselves as an AI decision support agent.

对吧？

Right?

系统的核心是，模型会随着新数据的输入不断学习。

So the system is based on the fact that the model will learn from new information as it's ingested in the model.

对吧？

Right?

就像我们之前说的，传统那种老派的营销组合分析，其实只是对某个时间点的数据做一次性的评估，每年都要重新做一遍。

As we were talking about before, to old fashioned, more traditional marketing mix is a point in time analytics exercise that's recreated every single year.

所以它根本不会从新数据中学习。

There's So no learning from the new information.

每次都是全新的分析，跟以前没关系。

It's actually a new analysis.

而我们的系统是随着时间不断累积、逐步优化的。

Ours is actually built an additive over time.

我们的模型基于成千上万的时间序列数据点，这些数据来自我们营销弹性引擎，每次新模型发布时都会进一步完善。

So our model is based on thousands of data points over time from our marketing elasticity engine that's built on with every new model that's published in the application.

使用我们系统的营销人员，不仅能从行业其他人的数据中学习，还能把他们自己新增的数据融入定制模型里。

So marketers that are using our system are actually learning from the rest of the industry as they go, but also new data that they ingest that's their own as part of their custom models.

每一个进入模型的数据点，都会让模型从中学习、不断进化。

So every single data point that comes into the model teaches the model based on that data point.

所以营销人员每次做决策，都是基于最新、最实时的信息。

So the marketer is constantly making decisions based on the newest most recent information.

你提过好几次了，你们平台有个正在申请专利的功能，叫营销弹性引擎。

Well, you've mentioned it a couple of times, a patent pending feature of your platform is the marketing elasticity engine.

格雷格，能解释一下这是什么吗？它怎么帮助优化不同渠道的营销投入？

Greg, can you explain what that is and how it supports the optimization of marketing investments across different channels?

这是目前业内最全面的信息。

So it's the best industry information possible.

我们的模型基于贝叶斯回归，也就是说，所有模型都以一些既有信息为起点。

So we built our models based on Bayesian regression, which means that there's prior information that becomes a starting point for all of our models.

我们建立了一个数据库，最初填充了大约四十年的学术研究成果。

So we built this database that's initially populated with about forty years of academic research.

不同行业普通品牌和企业的GRP每单位投放平均能带来多少回报？

How much does gRpFTV deliver on average for typical brands and businesses across industries?

我们还内置了一个自动化机制，每个弹性值都会匿名反馈到弹性引擎中。

We've then built in this automated mechanism where every single elasticity is blinded and informs the elasticity engine.

每次发布新模型时，我们都在从各行各业的数据模式和信号变化中学习。

So when we publish a model, we're learning from the changes in the data patterns and signals for businesses across industries.

它能为营销组合模型注入更多、更相关的数据，这些数据来自各行各业的企业。

It helps to infuse the marketing mix model with more information, more relevant information we're able to source from businesses across industries.

所以随着时间推移，模型的预测能力和建议能力都会越来越强。

So it actually makes the model more predictive and more prescriptive over time.

另外我想补充一点，我们可以用这些数据来模拟过去从未尝试过的营销渠道。

The other aspect I'll just add is that we can use that information to simulate marketing channels that have not been executed in the past.

比如，如果一个品牌考虑投资一个新的社交媒体平台，我们的引擎里已经有相应的弹性系数，可以估算出初始投入的效果。

So if the brand is considering as an example, investing in a new social media platform, we will have an elasticity in our engine that can form that initial investment.

这样一来，如果你想要优化一个固定预算，就能模拟新渠道的资源分配，清楚地看到该从哪些地方抽调资金来支持这个新渠道。

So you can actually simulate if you're looking to optimize a fixed budget, you can simulate the allocation for a new channel and we could see exactly where we'd pull those dollars to be able to fund that new channel.

这确实非常强大。

So that's very powerful.

如果客户想在营销组合里新增一个渠道，但自己没有历史数据，Keen Decision Systems 的模型也能预测它的表现。

If a client is considering adding a new channel to the media mix but has no prior data of their own, Keen Decision Systems model can predict their performance.

没错。

Correct.

等我们在这个新渠道积累了实际经验——也就是真正投入了预算，再把这些新数据录入模型后，模型就会根据这些新的数据信号，还有市场上的真实反馈自动调整。

And then as we get experience in that channel, so we execute the spend and we then incorporate that new data in the model, the model will adjust based on that data signal and the actuals coming out of the market.

那你的模型支持纳入各类影响因素吗？比如价格变动、新增的分销合作伙伴，还有大的经济环境，像通胀压力、季节性因素这些？

Does your model typically allow for factors such as pricing changes, new distribution partners, the broader economic environment such as inflationary pressures and seasonality?

这些都可以纳入。

All of the above.

要是不先把所有影响业务的外部因素都梳理清楚，根本没法制定出合理的投资方案，也没法准确预测营收和利润情况。

So you can't really forecast the right investment plan or forecast the top and bottom line without a good understanding of the external factors that are impacting your business.

我们会把这些要素全部纳入模型，这些核心因素能帮我们算出各渠道的合理投入规模，以及这笔投入能带来多大效果。

So we incorporate all of those into the model and their key factors that contribute to what the right amount of spend is and what the impact of that spend is across channels.

举个最典型的例子，就是新冠疫情那段时间。

You know, a good example obviously is COVID.

疫情刚暴发的时候，整个市场充满不确定性，我们立刻和客户紧密配合，把这种百年一遇的大疫情的影响全部整合进了模型里。

I mean, as we went in, we really leaned in with our clients with all that uncertainty incorporated one hundred year pandemic into the models.

这样一来，我们就能在疫情发展的全过程中实时观测到消费者行为的变化，以及各个渠道的表现出现了哪些改变。

So we are able to see in real time as the pandemic unfolded, how behavior was changing, how channel performance changed.

在这段时期，我们能够灵活调整，帮助品牌在如此严峻的环境下依然实现增长。

And we're able to make changes with a lot of agility through that time period that helped the brands grow even in the face of all that adversity.

你们的平台支持哪些类型的数据集成？

What types of data integrations does your platform support?

过去，营销人员在做营销组合分析时，最大的难题就是收集数据太费劲了。

So this was always historically the biggest challenge for marketers with marketing mixes, the heavy lift associated with collecting data.

过去十多年，我们一直专注于解决数据问题，尽可能提升数据接入的灵活性。

So we've spent the last ten plus years really focused on the data problem and have as much flexibility in ingesting data as possible.

我们提供了API接口，可以无缝对接各种数据系统。

So we have API's, so we're able to connect into data systems.

我们还预置了多种数据连接器，主流平台都能直接接入我们的系统。

We have data connectors that are connected in, so all of your major platforms can connect directly into our system.

如果无法直接对接，我们也准备了CSV模板，支持自动上传、映射，我们的系统还有智能映射功能。

And in the absence of that, we've developed some CSV templates that can be automatically uploaded and mapped and we have smart mapping within our application.

根据我们的建模方式和颗粒度，我们会把最合适的数据量导入系统，快速建模并制定计划。

Based on how we're modeling and at what level, we're bringing the right amount of data into the system to quickly model and build plans.

Keen Decision Systems 的模型多久更新一次？

How frequently are Keen Decision Systems models updated?

这取决于客户的决策周期，我们这么叫。

So it depends on the client's decision cycles as we call them.

他们多久查看一次营销情况，评估营销效果，然后做出调整？

How often are they looking at marketing and understanding marketing performance and then making any changes to marketing?

系统设计为每次有新数据时都会更新，但平均来看，客户通常是每季度更新一次。

So the system is set up to update every single time there's new data But on average, we see clients that are updating quarterly.

有些客户在越来越熟悉这个模型，并更多地用它来做决策后，已经改成了每月更新，这也能帮助他们制定每月的业务预测。

Some of our clients are as they get more accustomed to the model and leverage it more in their decision making have moved to monthly, which also then can inform their monthly forecasts for the business as well.

只要有新信息，就可以随时更新，但也要看这些信息什么时候会被用在决策上。

So as often as new information is available, they can update them, but it also depends on when the information will be utilized for decision making.

你们系统生成的收入预测，通常误差范围是多少？

Typically, what's the margin of error for revenue forecasts generated by your system?

一般来说，一开始的预测就挺准的。

So generally, it starts pretty accurately.

正如我之前提到的，我们的模型用到了非常多的数据，远不止时间序列数据这一种。

As I was mentioning before, we use a lot of data, a lot more data than just time series data to inform our models.

把营销弹性、先验知识、其他分析数据，还有时间序列数据和财务数据结合起来，最终打造出的模型既可以做预测，也能给出指导性建议。

So combination of the marketing elasticity, priors, as well as other analytics, as well as the time series and financial data actually leads to more prescriptive and predictive models.

平均下来，在初始阶段，我们的误差率不到4%。

So on average, to start, we're less than 4% error.

随着我们和品牌方、企业的合作不断推进，模型积累的信息越来越多，这个误差率还会进一步缩小。

And as the model learns more over time through the course of our relationship with brands and businesses, that will tighten even further.

而且随着我们对这家企业的了解不断加深，再纳入合适的业务数据，误差还会进一步缩小。

And the more we get to know the business and can incorporate the right information, will tighten as well.

格雷格，你能不能给我们举几个实际案例，讲讲你们平台的功能都有哪些实际应用？

Greg, could you walk us through some case studies that illustrate your platform's capabilities?

当然可以。

Sure.

我们合作的客户类型其实非常多元。

I mean, we work with a pretty diverse group of clients.

我们设计的系统对行业和数据类型都不敏感。

So we've built the system to be industry and data agnostic.

我们合作过的客户涵盖从日用清洁品到高尔夫球杆握把，再到外卖配送系统等各种领域。

So, you know, we've worked with clients anywhere from direct to consumer cleaning products to golf grips to food delivery systems, right?

我们跟很多不同品牌都有合作。

So there's a lot of different brands we work with.

最主要的应用场景是财务优化。

The primary use case really is around financial optimization.

比如在日用清洁品这个品类上，我们帮他们实现了每月一次的优化循环，通过平台更新数据、优化渠道和时机。对这个品牌来说，关键在于——我待会儿讲的每个案例都会提到这一点——我们都能发现相当可观的财务机会。

So on the d to c cleaning products, so we really wanted to get to a monthly optimization cycle with them using the platform to refresh data, channel optimization, timing optimization and for that brand in particular, and this is the key that I'll mention across all the cases I'm gonna talk about, we're able to identify pretty sizable financial opportunities.

在这个案例里，我们识别出了高达3300万美元的营收增长空间。

And in this case, we're able to identify a $33,000,000 top line opportunity.

一旦你搞清楚了哪里存在浪费，就能重新分配资源、优化配置，实实在在地省钱赚钱。

So there's real dollars and cents once you understand where the waste is to be able to reallocate and optimize.

我之前提到过那个冷冻食品配送公司，他们曾因疫情遇到一些困难。

I mentioned the frozen food delivery company that we work with, they had some challenges from COVID.

疫情期间，很少有人去餐厅吃饭，越来越多的人选择点外卖送到家。

During COVID, a lot less people were going to restaurants, a lot more people were getting food delivered to them.

所以他们的业务在2020年达到了顶峰，到了2021年就进入了下行周期。

So their business really peaked as they went through 2020, but then had to cycle it in 2021.

我们的角色也在这两年里发生了变化：2020年我们帮他们推动业务增长，到了2021年就得帮他们规划缩减开支，合理调整投入规模，确保投资回报率仍然维持在最优水平，哪怕外部环境不利，也能继续实现一定的增长。

So our role changed from '20 to '21 where we're helping them to fuel the growth in 2020, but then had to identify how to cut back in 2021, right size the investment to make sure that we're still optimizing the ROI and able to drive some growth even though there were external headwinds.

再说说那个做高尔夫握把的客户，他们当时的目标，首先就是在预算不变的前提下，把投入的效果最大化。

And then the example with golf grips, right, is they were looking to, one, get a lot more impact with the same budget.

所以我们帮他们找到了优化方案，能调整各渠道的投放时机和分配比例，实现业务增长。

So we're able to show them a path to optimize both their timing and their mix of spend across channels to drive.

这话听着可能很夸张，但他们的投资回报率足足提升了344%，营收还因此增加了1400万美元。

And I know it's gonna sound crazy, but a 344% increase in ROI that increased revenue by $14,000,000.

这些都是实打实的收益和价值，只要能实时掌握资金在哪能发挥最大效用，知道哪块可以加大投入，就能拿到这样的结果。

So real dollars and cents, real value being driven, just knowing in real time where the dollars are most effective and where you can step on the gas and invest more.

在过去几周的节目里，我们一直在探讨品牌如何实现增长，也聊到了埃伦伯格-巴斯研究所研究总结出的若干通用规律。

Over the past several weeks on this podcast, we've been examining how brands grow, discussing some of the universal laws established by the work of the Ehrenberg Bass Institute.

格雷格，基恩的分析是如何体现或支持艾伦伯格-巴斯原则的？

Greg, how does Keen's analysis reflect or support Ehrenberg Bass' principles?

我们之前提到过广告投放的持续性，我认为伯尼·夏普关于品牌增长的核心观点就是确保你始终出现在消费者面前，覆盖尽可能广的受众。

So we talked a little bit about continuity of spend and I think paramount to Byron Sharp and how brands grow is making sure that you're constantly in front of consumers and as broad reach as possible.

这一点我们也通过科学数据验证过了。

And I think that's what we've scientifically proven out as well.

你不能断掉曝光，必须持续出现在消费者眼前，保持投放的连贯性和稳定性，才能推动品牌持续增长。

So you can't be dark, you have to be in front of consumers, you have to maintain continuity and consistency in your spend to be able to continue to drive growth in the brands.

由于我们从长期效应的角度看待企业增长，我们能清晰展示出这些积累的层次，我们的系统里有一张图表，能清楚呈现营销如何年复一年地累积价值。

And we given that we have that long term effects viewpoint of how businesses grow, we're actually able to show the layers of rock and we have a chart in our system that show layers of rock and how marketing year after year is building value over time.

这和艾伦伯格-巴斯所倡导的观点完全一致。

And that's very consistent with what Ehrenberg Bass would promote as well.

如果InClear Focus的听众想进一步了解如何用基恩决策系统做营销组合建模，最好的联系方式是什么？

If InClear Focus listeners are interested in learning more about marketing mix modeling with Keen Decision Systems, what's the best way to get in touch with your team?

欢迎访问我们的网站 www.keands.com，申请一次演示或初步交流，我们很乐意分享更多细节。

Please reach out via our website, www.keands.com, and you can request a demo or an introductory conversation, and we'd love to share more.

如果想了解这个平台的具体功能，可以点击本集节目文稿中附上的短视频链接查看。

And if you'd like to see some of the platform's capabilities, follow the link to the short video included in the transcript for this episode.

格雷格，非常感谢你本周做客《清晰聚焦》。

Greg, thank you very much for being our guest this week on In Clear Focus.

谢谢，阿德里安。

Thank you, Adrian.

这次交流真的很愉快。

Really enjoyed it.

如果有人想更多了解Keen的背景，我们很乐意和你聊聊。

And, we'd love to have a conversation with anyone that's interested in more background on Keen.

再次感谢本周的嘉宾格雷格·多兰，Keen决策系统的联合创始人兼首席执行官。

Thanks again to my guest this week, Greg Dolan, the cofounder and CEO of Keen Decision Systems.

和往常一样，我们这次对话的完整文稿以及所有提到的资源链接，都可以在Big Eye官网 bigeyeagency.com 上找到。

As always, you'll find a full transcript of our conversation along with links to the resources we discussed on the Big Eye website at bigeyeagency.com.

在网站菜单里选择‘播客’即可。

Just select podcast from the menu.

感谢收听由Big Eye制作的《In Clear Focus》。

Thank you for listening to In Clear Focus produced by Big Eye.

我是您的主持人阿德里安·坦南特。

I've been your host, Adrian Tennant.

我们下周再见。

Until next week.

再见。

Goodbye.