Forecasting for Projects within your Portfolio

It goes without saying that the successful plans are those that are coming from accurate forecasting. Right forecasting and informed decision making are variables of a successful competitive advantage for our company and healthy products in our portfolio.

Some managers spend a good amount of forecasting and some others don’t touch it. An example is the hardware PMs who spend more time in forecasting than software PMs. This is mostly because of the nature of the job as hardware PMs have to plan the supplies and logistics of their products. A software PM though has to forecast the sales and subscriptions of the software but this doesn’t entail as much risk as in the hardware industry. As we understand, the amount of forecasting is hugely dependent on the product lifecycle also. Let’s go and analyze how forecasting requirements changes from stage to stage in the PLF of our portfolio products.

  • Conceive phase: Initial market sizing – quick and dirty for three to five years in most cases
  • Plan phase: Busines case – Estimate quantities to justify investment and three to five years in most cases
  • Develop phase: Planning for launch – New product forecast
  • Launch and market phases: Operations – Monthly/quarterly/yearly
    • New product availability
    • Ongoing product production and resources allocation
  • Retire phase: End-of-life plan – Monthly to years, depending on program specifics, parts and spares needed

Some of the questions that arise related to forecasting as we go from phase to phase in PLF are:

  • Enough interested customers?
  • Enough profit from those customers?
  • How many do we expect to sell?
  • Really, how many do we expect to sell?
  • Oh darn, this is how much we are selling?
  • What are the sales trends?
  • Do we need more marketing?
  • What are the end-of-life plans?

What makes good forecasting

The one ingredient that creates successful forecasting is the right data. Before you choose what forecasting method to use better to understand what kind of information have you gathered about your products and the quantity of it. Based of that, there are 3 methods on how to define which forecasting method to use:

If you have a lot of historical data, numerical methods are easy to construct and use as you can see on the chart. The indirect historical data means if you have some data to use from other markets that can be used as proxies for our market. And on the left, you can see the methods when we don’t have enough data. This is when we have new products or new markets etc. it is good to have Market experts in this situation to help us with their knowledge.

The less information you have the most difficult is to predict as we can see on the diagram below:

And the estimate is always a range and not something fixed. Being wrong is not an issue here and it’s a reality that happens to each PM. This is the nature of our work. Actually, what the PM’s responsibility is to reduce the chances of being wrong as little as possible and work on a backup plan. Steps to do this are below much information do you have?

  1. What category of the forecast are you in?
  2. What information are you missing?
  3. What are your assumptions?
  4. How can you check any assumptions?

Write as many details as it is possible and ensure that the key stakeholders from other departments, marketing, sales, distribution etc agree.

How to use the dta that you can find?

There are two sources of reliable data to do a good forecast – External and Internal.

External are government agencies, trade associations, market research and analyst firms, investment recommendation firms, customers and customer advisory panels, social media, chat forums, blogs, and finally competitors.

Internal sources: sales and sales operations, distribution channel partner management and channels, finance, web and CRM analytics teams, “Big Data” analysis team.

If you are currently managing a product you can have access to current and historical data. How many units you sold last month and how many YTD? How many new users did you acquire? Are the market conditions when you created the product still applicable? Add seasonality to your analysis. Maybe last month there was no covid and you had 1000 users in your online ordering app and this month cases in covid raised. Check your environment and look out for quantifiable information.

But what happens when we have a new product with no historical data?

This definitely is more challenging and the tools to use are analogy to similar product strategies. For products that are already selling in the market you can easily find this data. Be careful to have a similar target audience, positioning, and pricing.

 The next method is called the input-output method where the sales of one product are directly related to the sales of another product, an example is the sales of pc cameras will be linked to more downloads and subscriptions for zoom software. The usage patterns of one product make you understand how to manage your new product.

Another method is the Top-down/ Bottoms-up approach where we look at the size and growth of the market and then we check our own capabilities to fulfill these needs. This is when we miss important data also. Top-down is when you start with market size, estimated growth, and assumptions and bottoms up when you build forecasts based on the current small data about sales, distribution etc. when you make these small assumptions you can find the bottlenecks in your products and start do some optimization. Let’s say if the market is 1 million people and you have the capacity to serve 1000 then you should work with your devs etc..

When using the analogy method be careful to use difuccion curves like below:

In this way, you can see what drove adoption and sales at a certain period of the product increase. As we can see for the first product it took 25 years to get full adoption, and for the second less. These are details that change from product to product. Apple watch was introduced in 2015 and fit bit in 2014, but the apple watch was adopted immediately. Comparing apples with apples in a similar chart will give you important information to forecast your new products.

One good tool for forecasting the manufactured products and the resources behind them is the waterfall forecasting table below:

Let’s say that we prepare for July’s launch. You should always forecast with the other departments to have a successful launch. We always focus on the number of products that will be delivered in July. we can see the quantity forecasted in Dec is 1000, then 1500 etc..but on the other side operations need to prepare beforehand to help the delivery of units up. the same as the other functions. any chances made after Dec are difficult to adjust the operations part. An example is the booking of factory capacity for March which is 1000 units. But we can see that in March already the products that are in demand are 2500. In this way, we will not comply with the demand at the end of June. The waterfall forecasting tool is very important for products that have dependencies with other departments.

Now, this was when we have little data, but what when we don’t have data at all, what are we doing to forecast?

Then we use qualitative estimates by experts. Some of them are: Salesforce, customer panels, focus groups, market surveys, outside experts etc. Some well known methods are:

  • Delphi method
  • Jury of experts opinion
  • Salesforce composite
  • Consumer market survey

For all these methods, we should ensure that participants are well qualified.

Wrap up and final thoughts

  1. More data is better for a successful forecast
  2. Share your in insights with experts in your company to get buy-in in your forecast
  3. Don’t wish to always make the right forecasts, don’t stress about this, just focus to minimize risks and have back up plans if your forecast is not correct.
  4. Create multiple scenarios and avoid bias when you forecast with limited data.
  5. One day a black swan event will happen, just be sure to be agile when it comes.

Leave a Comment

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