Forecasting Stories 5: The story of the launch
New products forecasting can be very difficult - there is no history to start with, and hence no base line. The number of assumptions can be huge. The best way to forecast then, is to try parallel approaches, build different views and triangulate on a common range.
By Rajneet Kaur, Data Scientist
Launch forecast accuracy can be 40%, or even 1%
New products forecasting can be very difficult - there is no history to start with, and hence no base line. The number of assumptions can be huge. A plethora of techniques exist, but none are even close to accurate. The best way to forecast then, is to try parallel approaches, build different views and triangulate on a common range.
Let me take you to 2017 - we have been forecasting system volumes for a US OEM
and I encounter a different forecasting request. We do have a few quarters of data for the Ella(name changed) gaming laptop series, but this case would still qualify as new product forecasting.
Let us think of launch forecasting as a running race. What factors will determine Ella's rank in the race?
Let's get to how it was done, and we will compare that with a race at every step:
- Order of Entry - Order of entry refers to the position at which a player enters a market. The number and type of competition already present is an important factor to determine uptake of a new entrant. This effect may be more prominent in certain industries like pharmaceutical compared to certain others. As a thumb-rule, in a mature market, the top 3 players take up 70% of market share and other or new players would have to cater to niche categories to survive. Read more on Order of Entry: Rule of 3. Gaming laptops is a moderately competitive market and not mature yet.
Think of order of entry as how many are ahead of Ella in the race
- Analog studies - Analog studies refer to studying examples from the past. Using secondary market research, we follow the following set of steps:
The curve types as seen in the image, can be a combination of various statistical curves - the bottom most is type 0 and top one is type 10. They can be consumed in several other areas of forecasting as well , Event impact studies for instance.
Curve types can become different types of running styles: Ella may run consistently(linear), start fast and slow down(log) or warm-in and then speed up(exponential)
- Market & Marketing Patters - The gaming market itself is driven by certain trends, hence accounting for them is essential. In addition, seasonality exists in buying behavior since consumers generally await holiday and high discount seasons. Also, impact of business controllables like marketing spends and pricing need to be assessed.
Let's think of marketing factors as inputs during and before the running race: rest, food, exercise etc
The Result: Using the above, the following different scenarios scenarios were created for Ella:
While there were several revisions to the range of forecasts, the above were used as the starting points by our clients, which in itself is a win in the launch forecast world.
Summarizing the crux of our journey on product launches in the below visual:
Hope you enjoyed my concluding story on forecasting - any feedback or suggestions are welcome. To start with the first story in the set, follow the order: stories 1, stories 2, stories 3 and stories 4. I have also written about Analytics Frameworks for problem solving in the data science world.
Bio: Rajneet Kaur is a passionate and result-driven marketing and data science executive with 6 years of business and technology experience.
Original. Reposted with permission.
- Stock Market Forecasting Using Time Series Analysis
- Forecasting Stories: Is it seasonality or not?
- Backcasting: Building an Accurate Forecasting Model for Your Business