Demand forecasting, i.e. product forecasting has always been there. But predicting customer demand has always been problematic. In B2B markets, forecasting based on direct customer sales has been possible through the sales force’s forecast. Though not 100% accurate, at least there was dialogue with a prospect. In contrast, consumer demand in B2B markets has been most challenging: trying to forecast people’s behavior — not products.
But now through social and mobile, and some retail forecastable business models such as home delivery that have ‘subscriber elements’ (i.e. weekly home delivery of staple food items, for example) we should have the making of a customer forecast.
Customer vs. Product Demand Planning…
I’m offering an opinion piece here, since a clear path forward is not yet there. So this will be my prediction — our supply chain planning and forecasting methods are about to get a makeover — using new analytics based on entirely new sources of information. But first, let me ruminate a bit here and then walk a potential roadmap to the future. Those who know me know that I can do a couple of streams of ideas, so I request a bit of patience.
Customer Demand?
When we read articles, reports, and academic books about demand forecasting or ‘customer demand,’ they say, “To understand demand, start with a product forecast.” But isn’t that product forecasting? To explain further: A product forecast is a projection by the ‘sell-side’ — manufacturers or retailers — based on their existing products. And wouldn’t a customer forecast be one that forecasts customers’ behavior? It is not just a matter of semantics.
From decades of experience we know we are working with a faulty premise, since the world is littered with excess inventory — on the shelf, in the home, or as throwaways. (In my personal life, I will admit to buying items I don’t really need and making trips to thrift charity stores or Goodwill to dispose of these items.) Fundamentally, there is nothing wrong with those items. We know that in all industries the disposal of unsold or obsolete goods is a huge problem, incurring losses in the form of logistics costs, and product design and creation costs. And the truth be told, the total cost/loss for these items is not completely understood, especially if we factor in environmental impact.
Thought: New product introduction is an inexact process due to poor insight into the future.
If we look at what is embedded in the numbers for product demand, the information contains several aspects. Product forecasting is the result of predicting demand by channel, locations, or segments, and so forth. The rate at which products might be refilled or replaced (replenishment) can be adjusted by the terrain — time and expense to get the item to a specific location. Terrain also has other attributes such as environmental factors — disease, weather, local events, etc. All these factors can impact demand on product categories. In addition, a location has other attributes such as the demographics that describe the consistency or stability of buyers in that locale plus their past history of buying your brands or products.
Thought: Thus forecast is relevant to, and often influenced by, the location.
In the age of the customer, I think we need to amend these concepts and look anew. I won’t go as far as saying that technologies such as social, search, and mobile will destroy the past approaches; but surely those approaches need to be amended, since location should no longer be considered a fixed coordinate of the terrain, but rather, data attached to the customer who may also be mobile and not part of that neighborhood at all. Now that they have access to the world, customers have shown their ability to shop globally, right from their chair. And conversely, with mobile, they have shown their ability to ‘pop in’ anywhere.
So location is both a fixed coordinate and a dynamic data point. And demographics may point to a concentration of a certain wallet size, but they don’t tell us much about individual customers. With mobile, the consumer has a way to be self-declaring in a very personal way. I write about this topic a lot because each year more options to understand customers and access data enable us to become more analytical and incisive about what customers will actually want.
Thought: So mobile and dynamic location — location-based data1 — are important elements in thinking about new modes for demand management.
Search — Questions to Ponder
Is search buy-side (customer) driven or is it demand shaping (sell-side)? When universal search first made its debut, it appeared that the consumers were steering the queries. But after many years of search-engine-analytic-driven selections, I wonder. If you are a determined type, you can get beneath the front page and the ads to find something you are looking for. But otherwise not.
So are search results a product of buyers’ interests or are they circular sell-side projections? Probably a little of both. And judging by the main sources of revenue for the search community, i.e. marketing professionals, search results are probably more a projection (i.e. marketing) than understanding customers.
Thought: To understand customers, we need their input — not ours.
Big Data and Analytic Engines
We are getting pretty good at observing data patterns and correlating them. For ten years at least, there have been advanced analytics that look at various patterns and glean some important insights. But in reality, we are still new at this. Sifting through web data is still in its early stages in terms of learning new things about customers in an empirically meaningful way — for example, connecting web data to a predictable forecast of events or product purchasing. The challenge for the sell-side is to not jump to conclusions too fast — applying a subjective filter to what they are seeing. It is all too easy to do this; hence models have to be developed that test the patterns over time. And we have some experience with this in the demand-planning community. Forecasting systems that model new product introductions, for example, may propose a forecast, then rapidly test activities and sales to see if that is actually happening; then amend plans, if need be, as real data comes in.
All forecasting systems need to be broad enough to look at the supporting or environmental events surrounding product sales — not just the sales — to determine if those factors are the cause of any unexpected results, and if it is possible (or desirable) to change those events to create the desired outcomes.
Thought: We have the ability to analyze and correlate many aspects of data.
Social — the Next Frontier
My recent experience purchasing a rainproof jacket highlighted a few things to me. After looking at more than 35 different products, I narrowed the selection down to the ones with the most favorable reviews. I am sure you often do this. Reading reviews has become a habit for many of us — from selecting a cab company (my driver was late!) to consumers’ complaints about hotels, pricing, quality and true-to-advertised statements. A company that deals with consumers is smart to let these comments flourish (and of course address them). But they should also remember that these comments do correlate to potential or actual product sales. Recent attempts at social analytics are a bit crude at this point, but the early developers and users are learning a great deal.
One nuance of social is the gamification approach. Here, the product company engages in ‘what ifs,’ pricing, design, among others, to elicit consumer feedback before a product launch.
Thought: There is a value to social data in predicting low demand — and why!
I Get Sentimental over … Good Numbers
Social has its issues though: gauging the accuracy of sentiment and how sentiment translates into hard forecast numbers. Today, sentiment is also a hotly debated/contested field. If consumers don’t like or actually voice displeasure in some way, you have a lot of challenges in getting around that. If you try to ‘shape’ the dialogue, then you gain no real insight. On the other hand, social phenomena have allowed savvy marketers to gauge the public’s reaction to certain ideas and impressions before they create marketing campaigns.
Most social analytics sit within a company’s own website. So they can obviously monitor their own traffic, but they may not reach the wider world. Tools that go across browsers allow companies to see topics on Twitter, Facebook, and Google, for example, to see what is happening across the societal landscape. These views of consumers’ online habits can provide very useful insights, but also point to major challenges with the data. Consumer analytics (tracking cookies and other web analytics) are somewhat controversial; many consumers over 40 years of age object to them. However, they appear to be here to stay. The downside of the information is that it can be extremely inaccurate. Consumers can tailor their personas to avoid being targeted by certain types of advertisements, lie about their age, gender, and so forth. However, in all that data there still is co-relatable data that can be used for creating patterns and understanding consumers.
It sounds ‘mashed up,’ and we are still pretty far off from the precision required to use this data for financials and supply chain planning, but insights are there. Today, it is high-value information for early-stage category or brand-launching ideas. Unmet demand is pretty obvious already if you use the right search and analytics.
Not much has been done — yet — to correlate the success of the subsequent marketing campaign and its translation into sales and the creation of a forecast model. But that is coming and soon. We have been asking many of the demand players for a while about this, and their answers about the future are getting more articulate. You will see more correlative results as these programs move from pilots in end-user companies to really useful products in the market.
Thought: Social data that indicate popularity of brand, category, and reputation can serve as a signpost of early-stage demand for marketing and product planning.
Conclusion: An Opportunity within Easy Reach
Consider this: lifetime value or follow-on sales. Here is an opportunity to harvest a low-hanging-fruit, if we can get it right. We can leverage the data from existing customer relationships to educate the market and gauge results. Many companies already have decent data about their lifetime relationships with their customers. Computer and automobile purchases are markets in which the continued customer relationships are known, through services, add-ons, upgrades and eventual replacements. Some smart loyalty-program databases also contain information about customers’ brand preferences and interests and will market to them. (When my hotel loyalty program stopped sending me Caribbean get-away package promotions and switched to the London packages, I knew their analytic engine must have kicked-in!)
Now there are better ways to engage consumers, and if used intelligently (i.e. not daily email offers and useless marketing), can allow companies to design and sell products that result in additional sales.
It’s an easy opportunity. If you have a customer who spends $500 a year buying slacks for ten years and then suddenly stops, your analytical system — social or not — should be flagging this issue. And if it is statistically relevant — many customers’ behaviors are similar — you know you have a BIG issue to resolve. Using social networks allows you to dialogue with the customer and understand if you are taking your products and services in the right direction. I bring this up because, unfortunately, we don’t see much of this. One or two companies send out “we miss you” emails. But most of these emails are projections: here is a coupon — come spend. They don’t ask, “Why not?” There is no analysis of the root cause of the absence, no opportunity to “tell us how/why we are not making the most of our relationship.” Herein lies the power of social networking — it is a scalable way to engage the consumer.
Thought: Social can be input into loyalty programs, shape future demand, and forecast long-term demand.
These are early ideas of where we need to go to actually forecast customer demand and some ways in which to begin to leverage our new world to do so. Skeptics have told me that they already have customer demand since they have sales history and web analytics. But, again, these look at current or past behavior — they are not forecasts.
Many successful companies or product launches are attributable to deep and painstaking customer analytics. How refreshing when consumers are delighted with a product and how disappointing when they see the same old thing or poorly thought out products. And sadly, most companies have very sketchy insights into why they are failing or succeeding with a certain product. Creating customer relationships and analytics is the new way forward. It’s time to move beyond Facebook and create a new face for the company.
In my next article Rethinking the Customer at the Point of Experience, I will talk about concepts and technology that connect at the Point of Experience — the customer — to help provide customer services and also enhance the relationship between the merchant/brand and their customers. In the final article I will talk about transforming Point of Sale.
More to come — — — — —
References:
_____________________________________________________
1Often when we discuss location-based data, we refer to an object’s location — destination. Here, we are talking about the ‘subject,’ the person in motion and their relationship to the location: swarming, for example, as a crowd leaves a concert or sporting event.
— Return to article text above
To view other articles from this issue of the brief, click here.