ShadowLink93 said:
freedquaker said:
Awesome table! I have applied a simple intuitive regression analysis based on the data above in order to estimate the quarterly software sales figures. The model simply assumes;
a) Quarterly NEW hardware sales (x1) are positively correlated with software sales;
b) Quarterly CUMULATIVE hardware sales (x2) are also positively correlated with software sales (thru a different channel).
In other words, both the quarterly hardware sales AND the Install base simultaneously affect SW sales. Analytically speaking, the model is defined as,
Y = Betha_0 + Betha_1 * X1 + Betha_2 * X2 + Error Term
where Betha_1 is the slope (impact factor) of quarterly sales, an Betha_2 is the slope of cumulative hardware sales.
Interestingly the estimate function is acquired as Y = -0.4328 + 3.7718 * X1 + 0.4133 * X2,
where R_square = 95% with P-values and F-value practically zero. In other words, the model can explain 95% of the variation in SW sales with 99.99% confidence. Long story short, the moral of the story,
1) We can easily and accurately estimate the software sales by knowing the quarterly and cumulative hardware sales,
2) Cumulative hardware sales (Install base) is incrementally getting more important according to the model, so as time goes on, it becomes incrementally easier to estimate SW sales even without quarterly sales.
Here is a comparison between the quarterly software sales and the model estimates from 2016-2017 q4 thru 2021-22 q2
| 5.46 | 11.03 | | 8.14 | 8.90 | | 13.88 | 13.77 | | 25.08 | 32.98 | | 16.41 | 17.97 | | 17.96 | 14.79 | | 24.17 | 21.05 | | 52.51 | 48.40 | | 23.91 | 23.24 | | 22.62 | 22.84 | | 35.87 | 34.89 | | 64.64 | 62.03 | | 45.59 | 35.02 | | 50.43 | 46.34 | | 49.82 | 53.67 | | 75.85 | 76.21 | | 54.78 | 52.33 | | 45.29 | 53.15 | | 48.60 | 52.39 |
Thanks for reading and please feel free to comment.
|
I added cumulative if it helps with your estimation, say if the Switch sells the same as last year (11.57m) in this Q3 how much software will it sell considering the higher install base and the install base is still actively buying software. 
|
Nice table, I love it. Well, I had already included the cumulative sales (calculated from your data in excel) but your illustration is more beautiful. Just to be clearer on the estimate, basically just multiply the monthly HW sales estimate by 3.77 and the cumulative HW sales by 0.4133 and subtract -0.4328 from the sum of the two as in
Y = -0.4328 + 3.7718 * X1 + 0.4133 * X2
I am pretty sure one can devise better but more complicated estimates but I am very surprised how deterministic (and highly correlated) the relationship already is between SW sales and the two measures of HW sales!
