This is exactly what you will never see in IPnest surveys (and written in Semiwiki):“Advanced Process Control Market worth $1,465.62 Million by 2020” …sorry for you if you like very precise figure (like $ 1, 465.62 million) but you will NEVER see such insane forecast in any of our surveys or blogs. Building a forecast is a difficult exercise, essentially based on assumptions (except if you have bought a crystal ball) and we will list the reasons why it’s far to be easy- and can’t be precise at 1/100, 000.
Before even thinking about building this forecast, you need to make a market size evaluation (as precise as possible, say at less than 1 %). In certain cases, you simply can’t access to these data, but let’s imagine that you can do it. Then, you propose a snapshot of the market status as of today (in fact, last year). If you have been rigorous, this market view will be useful for the various actors, suppliers, customers or acting in support functions to this market. You can find below a good example of a well-done job: ranking of the Semiconductor Top 20 in 2014 by IC Insights:
Even if such a ranking from independent analyst is very useful, customers will more likely ask for a forecast, usually 5 years long. At this point, you are leaving the comfortable, rational area to enter into “forward looking” type of analysis. I mean that you have to think in term of assumptions, not actual data. When you are doing a serious job, such assumptions are comforted by a deep knowledge of the specific market you are reviewing. Let’s add a few insights coming from key actors and private discussions held with some people (from your network). In fact, assumptions are only to be used after making an exhaustive review of the market under review, starting from the above table if you plan to forecast the semi market for example. One option can be to segment this market into high-end, mainstream and low-end segments or into analog, memory, processor and logic segments. Splitting to better evaluate each of these segments as the market dynamics will be different from one segment to another.
At this point of time you can start using forecast maker’s best friend: Excel spreadsheet. If your assumptions are based on rational, if you don’t make any mistake when turning it into formulas, pushing Excel button will give you results that you may organize into nice curves or table and print. By the way, don’t forget to check if the result at 1 year makes sense (and rethink the whole process if it doesn’t, identify why and correct). Once you think your forecast is solid and you come to a result you can share with customers or partner (or with your boss), PLEASE don’t print figures with 8 digit precision! This is simply NON-scientific; anybody having passed a degree in physics should know that the final precision when giving a result is always less than the weakest part of the equation. In this case, the precision linked with your assumptions…
Every time I see such results for a 5 years forecast (1, 465.62 million), my first reaction is to think that the person who has built this forecast is just… non-scientist, and I strongly doubt about the results accuracy.
PS: in 2009 IPnest has built the first 5 years forecast for the “Interface IP Market”, see above. In 2008, this market was weighting $240 million, and the forecast for 2013 was $425 million. If I am ashamed when looking at the syntax or the way this first survey is organized, I am very proud when realizing today that the 2013 actual market result is accurate by less than 5% with the forecasted market size in 2009…
PPS: For 2020, IPnest is predicting the Interface IP market to weight $820 million (not K$ 820, 150.17)… see you in 5 years.
From Eric Esteve from IPNESTShare this post via: