How to Forecast When History is No Longer Relevant

Discussion
Jun 03, 2009

By Janet Dorenkott, VP and Co-owner, Relational Solutions, Inc.

During the first day of IE Group’s Consumer Packaged Goods Forecasting and
Planning Summit last week, almost every speaker reiterated that historical
trends and year-over-year comparisons do not carry the same weight that they
have in the past. And although these metrics are still valuable for forecasting,
many other factors now need to be considered due to the rapidly changing
economy and buying behaviors of consumers.

One of the main thoughts behind this reconsideration of traditional methods
is that consumers are saving again and, thereby, taking money out of the
market. Pat Conroy of Deloitte Consulting estimated that if consumers save
just seven percent more than they have in the past – not an unlikely scenario,
according to his research – they will be removing $500 billion from the economy.
He also made an excellent point that, as consumer spending goes down, a greater
percentage of the money that they do have to spend may be allocated toward
services vs. products – his theory being that their expendable income will
go toward fixing products they already own to stretch their investments.

We also heard in the presentations that there is a new attitude – "Frugal
is cool." People are buying fewer luxury items. Even those consumers
who can afford them are leaning toward purchasing less conspicuous and more
"everyday" products. Another factor affecting purchasing is the
general fear in the market coming about due to the government’s explosive
spending, the housing crisis, the banking and auto crisis and more.

Tim Weidenhaft of General Mills echoed the sentiment that shifts in demand
patterns due to economics also makes purchase history less relevant. He pointed
out that this is causing higher demand volatility and a resurgence of coupon
redemption.

A key sentiment in many presentations was that automating the integration
of POS scan and forecast data and integrating it with things like promotions,
shipments, forecasts and orders through applications like POSmart, is a critical
factor. Also, leveraging panel data, weather trend data, syndicated data
and even economic data that can be purchased through third party sources
are several more ways to improve forecasts.

Discussion Questions: Given the current trends in consumer spending (and
saving), what factors in addition to historical trends should be used to
help determine your forecasts? What other impacts from the economy should
be considered?

Join the Discussion!

19 Comments on "How to Forecast When History is No Longer Relevant"

Notify of

Sort by:   newest | oldest | most voted
Nikki Baird
Guest
8 years 6 months ago

Consumer data is the most interesting to me in terms of adjusting demand forecasts, especially some of the interesting things that panel companies have been doing around purchase intent, and tying demographics to saving and spending behavior changes to get a better profile of who’s buying and who isn’t.

Whether that data comes from a big consumer data company, from a retailer’s direct research into its customer base, from a manufacturer’s direct research into the market, or from somewhere else, it comes down to this: you can’t be customer-centric if you’re not using customer data, especially in how you plan.

Alison Chaltas
Guest
Alison Chaltas
8 years 6 months ago

Projections based on forward looking shopper insights are the way to go. This recession has shocked shoppers into a totally new mindset that will take years (if not decades) to evolve. Historic models need to be re-evaluated. Historic analysis has its place, but can no longer be counted on to predict sales in the short or medium term.

The smart retailers and marketers will use this time to invest in getting to know their customers and testing new propositions to win their loyalty with new, more cautious, savings and environmentally oriented American shoppers.

Liz Crawford
Guest
8 years 6 months ago

The virtual store research methods mentioned in this space over the last few months could prove to be a faster, more accurate forecasting tool than history.

The rate of change is accelerating across society. Using historical trending works when there is long term stability in demand and conditions–making consumer response predictable. Those times are behind us.

Michael Tesler
Guest
Michael Tesler
8 years 6 months ago

The stores with positive comparisons to last year (and there are some in existence) have no problem at all with the continued relevance and importance of this measurement.

David Morse
Guest
David Morse
8 years 6 months ago
It may not be completely necessary to throw historically based forecasts out with the proverbial bath water, even in these strange economic times. Historical, or time-series based forecasts, are a function of three factors: level (how high your sales are), trend and seasonality. In some cases, it might just be a question of adjusting one of those factors. For instance if you’ve been trending upward at 20%, you might need to adjust your model to affect zero growth. If you’ve got a growing product, but you expect to be hit by hard times, maybe what you need to adjust is your sales level. The problem with these models is when history is no longer a predictor. In that case, a time series model can be about as effective as driving while looking in the rear view mirror (as a friend used to say). In that case, your forecasting methodology might need an overhaul. Maybe your sales are tied to gas prices or the Consumer Spending Index, or a combination of factors. In that case, you… Read more »
Jane Sarasohn-Kahn
Guest
8 years 6 months ago

In times like this, we need a large dose of scenario planning. In scenarios, we consider what we know we know (e.g., that consumer spending is down), what we know we don’t know (e.g., that consumers are saving, but how long will this persist and at what rate?), and what we don’t know we don’t know.

This last category we refer to as “wild cards.” In my work in health forecasting, we often refer to acts of bioterrorism or pandemics (think: swine flu). Wild cards are the low probability events that blow our ‘reasonable’ forecasts to smithereens.

When it comes to saving, who would have thought? New data points and qualitative factors will be instrumental in helping us better forecast and respond to consumer demands.

Li McClelland
Guest
Li McClelland
8 years 6 months ago
I have come to believe that the increased “savings” rate for Americans that is frequently mentioned in studies and articles is very misleading and very misunderstood. In fact, the savings rate does not necessarily mean that people right now are voraciously socking money away in emergency funds, CDs, their 401ks, or stuffing cash in mattresses. Rather, if they are employed, many are busy paying down debt (HELOC, credit card, student loans) with that “extra” money they have each month because they are currently NOT spending as much with retailers. After their outstanding debt is more comfortable for them, then what? Consumers will start buying again at some point. But, perhaps not right away. Perhaps, now more financially savvy, they will start to actually put some money in the bank to “save” for the future. Retailers need to recognize this and make plans on the basis that recovery is not just around the corner–and that there is going to be a new normal that history doesn’t clearly lay out for us.
Ben Sprecher
Guest
Ben Sprecher
8 years 6 months ago

One challenge comes from trying to tease apart a number of overlapping and interacting trends: 1) shoppers spending less overall; 2) shoppers diversifying channels to include discount stores, dollar store, club stores, mass drug, etc; 3) shoppers cherry-picking deals; 4) shoppers trading down within categories (from national brand to discount brand) and across categories (from prepared foods to basic ingredients); 5) shoppers spending more in stores as they eat out less.

Retailers can look to panel data and industry studies and surveys, but that only gets you so far. Each chain is different, each region is different, each store is different. Retailers with loyalty data have a great opportunity to dig into the data and observe where the trends are taking *their* shoppers and where the demand is heading in *their* stores. It’s hard work, and many retailers don’t have the tools to make it easier, but the payoff can be huge.

Herb Sorensen
Guest
8 years 6 months ago
Foretelling the future is always a fraught business, much given to touting successes and sweeping failures under the rug as quickly as possible. Having said that, it is an absolute necessity for anything beyond animalistic existence. Does the future simply happen to us? (B.F. Skinner, Beyond Freedom and Dignity) Or do we consciously create it? (Free will.) It’s hard to argue it on a personal level, where we see those with strong personal visions of the THEIR own futures creating them day by day. It’s the interaction of those billions of personal achievements, crossed with “the weather” that led Bobbie Burns to declaim, “The best laid plans of mice and men often go awry.” With these words of caution we note that the long scope of known human history has shown the positive impact of progressive human creativity, not only on those humans but on the halo surrounding them. At the same time, we have been plagued with an advanced form of doomsterism, particularly amongst the intellectual elite who virtually align with Anthony Newly’s, “Stop… Read more »
M. Jericho Banks PhD
Guest
M. Jericho Banks PhD
8 years 6 months ago
There’s shopping (research, aka “looking”), and then there’s shopping (buying). During my days creating advertising for the Sears Catalog while with Campbell-Mithun Advertising in Chicago, the #1 wish among the Sears “wishbook” staffers was to know which and how many pages shoppers dog-eared in their catalogs. This form of shopping – in which customers display interest in a product but are not yet buying – intrigued Sears because they saw an impossible-to-access way to reach out to customers with special offers on the products that interested them most. It was frustrating. The catalog is long-gone but internet shopping is here in a big way. As we all know, consumers frequently shop (research) online before shopping (ordering online or purchasing in a store). And therein lies another tool for forecasting production and sales. Our team is developing ways for brands and retailers to harvest all internet clicks on their products and services regardless of the site. In addition to producing valuable metrics regarding interest in various goods, the data harvesting will be backed up with the… Read more »
Frank Beurskens
Guest
8 years 6 months ago

Some insight may come from the exchange-traded commodity markets where uncertainty and chaos induced shocks frequently occur. Historical data can be comforting but as the fine print always reminds us, past events have little if any bearing on the future. One tool traders utilize is implied volatility which quantifies daily, the degree of market uncertainty. (Livestock, grains, energy, interest rates, and stock indices’s are all exchange traded).

While many forecasts are seldom more accurate than a coin flip, adjusting trend based estimates based on some aggregate indication of volatility may result in a slight reduction of risk.

Emmett Cox
Guest
Emmett Cox
8 years 6 months ago

Pattern Recognition during good “Stable” times is difficult to do correctly. Forecasting from these patterns are even more difficult.

If you are forecasting what the next 5-6 months will entail (Demand, Trends, Pricing…) the current trend data is appropriate. You would not build a 5 year strategic plan based on current (previous years) data alone.

The fundamentals still hold true: Incorporate a blending of data types/sources and provide appropriate weighting to each variable; POS scanning market basket data, Credit card data, Loyalty/membership data, Geo-Demographics data, Syndicated data, GDP etc.

Each of these provides particular insight into consumer behavioral changes. Targeting BEST customers and keeping them, does not change. Converting “Near” best customers to best, does not change.

Still like steering a big ship, frequent corrections are required. Not infrequent wild changes. More frequent analytics measurements and near term tactics are more prudent than 4-5 year strategic plans during economic turmoil.

But (always that) it will depend on the industry.

John Crossman
Guest
John Crossman
8 years 6 months ago

I still believe that many historic methods still work. It’s just that many of the fundamentals have been ignored the last several years.

Ralph Jacobson
Guest
8 years 6 months ago
Lora Cecere of AMR said it best a few years ago, actually. She was talking about Demand Signal Repositories (DSR). A DSR is a robust centralized database that stores, harmonizes, and normalizes large volumes of demand data including point of sale (POS), wholesale distribution information, inventory movement, promotional demographics, market demographics, third-party market content, and customer loyalty data to support better decisions in the areas of category management, joint value creation, vendor-managed inventory (VMI), trade promotion management, supply chain management, and promotion management. However, the challenge is HOW to use that data. More and more CPG companies say that POS data is gaining importance. Predictive Modeling is now taking shape after years of just talk. Predictive models are constructed using historical data and are then applied to current data to generate forecasts. Demand forecasting is challenging because multiple complex phenomena must be accounted for simultaneously: Promotional Lift, Non-promoted items can experience lift from promoted items, Estimate of what Base Sales would have been had there been No Promotions and Actual Sales, just to name a… Read more »
W. Frank Dell II
Guest
8 years 6 months ago

Whoever said history repeats itself never worked in retail. The problem with forecasting with consumer products and retailer is no metrology works long term or consistently.

The first step is to move off weekly data to daily data. Add in a quick adjustment component and hope for the best. Note the greatest inefficiency in retail is created by forecasting. The better approach is to quit forecasting. Use POS sales to drive store replenishment. Most shelf space is designed to support a full case and additional facings for faster moving items. That means every item has some days supply. When the POS equals a delivery, have the distribution system kick into action. Operating with real numbers, not a forecast, will greatly improve the in-stock position while reducing inventory. Using this approach, whatever the consumer does, the store is in sync.

Ted Hurlbut
Guest
Ted Hurlbut
8 years 6 months ago

History may be somewhat less relevant to forecasting in this environment, but I don’t think significantly so. History, after all, has always required interpretation and adjustment, and the relationships between items and categories has always been as important as discreet sales trends of those items or categories.

Having said that, it’s important to note that forecasts are rarely spot-on correct. They are, after all, the statistically most-likely level of demand, and other levels of demand may be only slightly less likely. What we are obviously seeing in the current moment is greater variances between forecast and actual.

Which means that how the plan is executed is more important than exactly what the forecast is calling for. Retailers are holding more dollars in their pockets longer, so they can respond to actual data rather than forecasts, and we’re not likely to see a meaningful rebuilding of inventories until after the consumer clearly demonstrates that they are indeed ready to spend again.

Scott Knaul
Guest
Scott Knaul
8 years 6 months ago

I have the opportunity to work with a lot of retail clients and for the most part, history is still relevant. For the purposes they are using it for, the day-to-day and week-to-week builds are still relevant. The trick is to keep up with the current trend and apply that to the build factors. This is helpful for short-term planning for payroll, inventory, and a few other key drivers.

Mark Lilien
Guest
8 years 6 months ago

Smart forecasters use longer-term trends as well as short-term results. When overall spending is down, the basic shape of the demand curve is usually the same, regardless. And so much volume is promotionally-driven anyway, so the forecast accuracy is often based on the promotional tactics being repeated (or not). Furthermore, overall spending doesn’t matter if some of your worst competitors are closing stores.

Additionally, just because Americans save more, it doesn’t mean the funds are taken out of circulation. The money isn’t destroyed, it’s just transferred.

Gordon Arnold
Guest
8 years 6 months ago
Accurate forecasting of fiscal year monetary requirements is indeed an important part of a Corporation’s integrated list of business needs. Confusing a company’s forecast as a collection of market predictions for given itemizations and not as a yearly revenue requirement is dangerous. When market conditions radically shift in “any” direction (up, down, or evolving) the company’s life is at risk. Changes that fit the corporate dynamics and capabilities must be made allowing the company to latterly expand into verifiably stable or growing market segments and items. Failure to provide dependable positive revenue flow and predictable dollar turns at the forecast fiscal required pace is a sure means of forcing the market to terminate the corporation. This end result can be delayed as clearly seen with General Motors, Chrysler and others but it is the end when the company consistently fails to match market opportunities and corporate revenue investments per the fiscal year forecasts. In addition to exploiting inviting well-suited market opportunities and closely monitoring the company’s revenue, identifying lively and dead-end work efforts on a… Read more »
wpDiscuz

Take Our Instant Poll

Do you agree that current consumer spending behavior is rendering traditional forecasting models less accurate?

View Results

Loading ... Loading ...