Forecasting in a Challenging Business Environment: Lessons From Procter & Gamble

Dick Clark

Dick Clark

Developing accurate forecasts is never easy.  The current economic environment makes it even more challenging and many businesses have a bias to over forecast.   At the same time, businesses need the most accurate volume forecast possible to enable effective planning and decision making.  It is important to understand that this is a business issue, not a forecasting capability issue.  Demand Planning does not have a crystal ball.  There is no silver bullet.  There is not one work process, one training class, one software solution, or one method of collecting assumptions that will eliminate forecast bias.  We must approach the problem holistically and recognize that many basic elements must come together.  While we must be aware of the current volatility and uncertainty of demand, the areas of focus are unchanged:

  1. Within Demand Planning, we must stick to the basics.  Basic models, standard analysis techniques, and “normal” forecasting processes really work.  The same things that improve forecast accuracy and reduce error in good times work now as well.
  2. Within a business, we must recognize that macro events and processes are the largest source of forecast bias in most businesses.  Bias is not coming from our software solutions.  Rather, it is macro assumptions, management over-rides, and other “big” decisions that lead to forecast bias.  Addressing culture, behavior, and rewards are critical to eliminating bias.
  3. Within a market, we must better understand macro market trends.  In the current economic environment we are seeing unexpected changes in market size and dramatic changes in customer, shopper, and consumer behavior.  We need to better understand the relationship between market size, shipments, and share through collaboration with business partners.

Many businesses are dealing with persistent bias.  It is an on-going problem that not only impacts business results, but also erodes Demand Planning credibility with the customers of the forecast.  Eliminating forecast bias is difficult and no single approach works for all businesses.  I will be covering some of these ideas in my Luncheon Keynote presentation and Executive Forum at the IBF Conference in Orlando.  I hope they will spark your thinking and lead to dialog with others at the conference and with your business partners when you return home.  I look forward to having the chance to talk with and learn from many forecasting professionals during the conference.

Dick Clark
Demand Planning Global Process Leader
Procter & Gamble

See DICK CLARK Speak at The IBF’S:

$695 (USD) for 3 Full Days!

October 12-14, 2009
Orlando Florida USA

7 Responses to Forecasting in a Challenging Business Environment: Lessons From Procter & Gamble

  1. Forecasting accuracy is definitely a very important subject but can be applied practically to limited extent because sometimes overall business strategy takes over as the deciding factor

    By Manoj Sharma

  2. There are 2 dimensions to the forecast –
    1. The aggregate volume
    2. The mix within the categories

    Each is equally important, but the mix forecast is far more important to product availability and demand response. We have learned that even if the aggregate volume is off, if the mix forecast is correct, the turns and product availability are significantly better.

    By Jeff Fisher

  3. Developing accurate forecasts is even more difficult in a high growth market, where one just does not know the true potential; with repeated unidirectional forecast biases. Forecasting will always remain a challenge particlualry in the FMCG industry with large number of skus and variables. Having said that, I think forecasts are impacted more by internal factors than external eg trade & consumer promotions, lack of internal communications with lead-lag in information flow etc.

    By Tanmay A

  4. Nice article, you make some excellent points!

    Some starting points:
    Tracking the performance of the “management overides” against the baseline or system forecast?

    After the results are analyzed would

  5. Excellent “strategic” insight into the real issues for demand planning. However, What happened to all the ECR initiatives, I worked with Ralph Drayer and others while I was Principle, Integrated Logistics with IBM Consulting Group in early 90’s.??

    IBM had “purchased” the “Continuous Replenishment” software, including the “agreed to rules of opeations” that allowed actual POS data by sku, by store, to be uploaded everynight and determine total demand and forecast, build and ship deliveries, automatically – I.e., Proctor & Gamble placed orders on themselves for replenishing the likes of “Walmart per se”.

    PCA / Sara Lee / Dominick’s at the time – where also instrumental in leading edge “pull demand planning” vs “push”.

    Lastly, I know it worked well – because in 1996 – I did a major Supply Chain (ECR Readiness Assessment) consulting project for LG Chemicals in Korea and their largest customer! WalMart had already moved into Korea and P&G was coming. The LG chemical division, included the major competitive products that P&G sold. OH – by the way – did I tell you in Korea all inventory at Retail store is “vendor managed inventory” on consigment, with no cost borne nor paid by the retailer until (some time after) it is sold. That may throw a wrinkle into some of your standard modelling assumptions — Which gets back to the point of the macro-economic factors that effect all of this – and of course LG has 48 DC network with their own truck deliveries (kinda like the Coca Cola network distribution model, with DSD (Direct Store Delivery) and direct sales force.

  6. Nice article, you make some excellent points!

    Some starting points:
    Tracking the performance of the “management overides” against the baseline or system forecast?

    After the results are analyzed would both parties (management and Forecasting Team) evaluate the outcome and decide which approach is best for the company.

    Decisions should be based on data, sometimes a little compromise is neccesary, however you need to follow the model:

    Plan
    Execute
    Measure
    Adjust

    At the end of the day it should not me a finger-pointing exercise but a learning experience that will make for better processes.

  7. Great points. I’m particularly curious to get more insights on best practices and watch outs related to store level order forecasts that Don Madsen is refering to. I have seen a couple of attempts at different FMCG organization never getting closer than 50% at the store/sku/day level. Any thoughts on tools, statistical techniques, business processes or input used to get to an executable store/sku/day level order forecast to drive replenishment. There is a lot of discussion around POS data, but what role if any does in-store inventory levels play.
    Don can you share some learnings from the exercise is Korea.

Leave a Reply

Your email address will not be published. Required fields are marked *

WordPress Anti Spam by WP-SpamShield