Forecasting as an improvement project

[RosieA]

My company has been struggling with forecast accuracy as an improvement issue for the last two years. Despite a lot of focus and a bunch of actions related to it, we continue to struggle with an accuracy level that bounces all over the place.

One current thorn is the need for a 6 month purchasing window for PC boards when our forecast window is less than 90 days. Historical forecasting hasn't been a great help as our business model has changed radically over the past three years.

Has anyone else tackled this wonderous process, successfully or not? I need some new ideas about how to tackle this issue

[Chris May]

Rosie,

I have also been there with PCB procurement.

The conflict seems to be with your incoming 90 day forecast (Sales I presume ?) and the outgoing one to the supplier.

I assume that you are giving a 6 month forecast so that you can benefit from a pricing structure.

However, as you mentioned you have spent a lot of time (money) trying to get the forecasting working.

Try to get sales to expand their incoming forecast and also see what the cost implications are for lowering your 6 month forecast to your supplier, to, say, 3 months.....OK they may want to up the unit price, but what hidden costs would you save by getting it working....if that makes sense.

In the PCB assembly business and especially in subcontract, you are quite correct by saying that you cannot use historical data.

A way to maybe "wind-up" the incoming sales (I assume) forecast would be to point out that, as a business, it may not be very healthy to only know what you are going to be doing for the next 90 days.

Maybe contact another PCB supplier and see if thay can work with you on a smaller scale forecast.

Our big runners are actually made in China (subbed out by our supplier) and we benefit from good pricing.

Hope this helps,

Regards,

Chris May

[RosieA]

Thanks, Chris

The 6 month leadtime is being driven by one or two board components that are on allocation, vs price. We are paying premiums to expedite the components, but it isn't enought to offset the constant crisis mode we seem to be in.

We've been through quite a journey, first outsourcing our PCB manufacturing, then outsourcing our end product assembly. At the same time, our business model changed from an OEM model, with 90% of our business coming from major OEMs who gave us long term forecasts, to a shorter, quick turn model, with a lot of guesswork for the forecasts. The answer would seem to be an investment in finished goods, but management is reluctant to commit the necessary dollars and our product has a large number of optional variations which makes it hard to pick the right models to stock.

Forecasting is a gray process with lots of multiple inputs. It's like a monthly darts game vs a clear process. Im looking for inputs on how this process can be made more predictable. Our goal is 80% accuracy, but we don't often hit that.

[NYHawkeye]

RosieA said:
Im looking for inputs on how this process can be made more predictable.

Rosie -

Boy, if you can figure this one out I think you will be in a position to make a fortune!

Seriously - I have been working to solve this for quite a while with little success. Unless you have some fundamental flaw in the forecasting process the accuracy of your forecast is pretty well determined by the nature of your business. There usually isn't much you can do to change the nature of the business.

What is more important is to understand the accuracy of your forecast very well, share this knowledge with your suppliers and spend your time trying to develop a supply chain that is flexible enough to work with the realities of your forecast. Sometimes holding inventory is the answer, but there are often more innovative things that can be done working with suppliers and customers that entail less risk.

So, while I wouldn't give up trying to improve the forecast I would suggest that you concentrate on understanding the inherent accuracy of your forecast and trying to put processes in place based on the realities of the situation.

Hope this helps.

[RosieA]

Thank you for the insights, Hawkeye

I've been managing continuous improvement projects for 15 years and this is, by far, the grayest and most difficult process to define, much less improve.

What kinds of innovative vendor things do you mean? I'm seeking new ideas!

[NYHawkeye]

Rosie -

Well, innovation to me can mean two things. First, you can apply things already being done elsewhere to your situation for the first time. Second, you can be really creative and come up with totally new solutions.

Since I am not that creative I usually try to take good ideas being done elsewhere that have not been tried in my current situation for one reason or another.

I don't know your current situation and perhaps you are doing or have tried some of these:

1. Schedule Sharing
2. Procurement, Assembly, and Delivery Time Fences
3. Volume Pricing Agreements
4. Replenishment (Pull) Systems With Set Liabilities
5. Long-Term Agreements With Shared Liabilities
6. etc...

Three years ago our organization was not using the forecasts provided by the business at all with our suppliers because they were so volatile. Now the variability has not changed but we are using schedule sharing for 70% of the material on our new product lines and have achieved lower costs and better on-time performance.

There isn't necessarily one thing we always do with a supplier but we try and show them that even a volatile forecast is of some value to them and encourage them to work out some type of win-win agreement.