0

What is the approach to solve such a question, given that I need to simply analyze a dataset and express findings I find?

Problem:

I am given a dataset of everyday revenue for three children companies of ABC for the period of January 1, 2014 to March 31, 2014.

I am also given revenue and profit for the three children companies for last five quarters of the above period, i.e Q4 2012, Q1 2013, Q2 2013, Q3 2013, and Q4 2013.

My aim is: Based on the everyday revenue data for the period of January 1, 2014 to March 31, 2014 (Q1 2014) and the historical data (revenue and profit) to come up with an estimate for the profit that ABC will report as their Q1 2014 profit.

How should I approach this problem?

Alpha Bing
  • 11
  • Well...you only have a handful of data points so whatever you do is going to require a lot of assumptions. Simplest thing is just to take the mean (working child by child). Lots of reasons why that may be a bad idea though...many businesses are very cyclical so no one quarter looks that much like the mean. – lulu Aug 15 '20 at 22:36
  • I'm sorry I don't understand what you mean by using mean? – Alpha Bing Aug 15 '20 at 22:39
  • You have profit for several quarters. You can view that as a percent of revenue (for example) and then take the mean. If profit is regularly $5%$ of revenue, say, then this method might look good. But in other contexts it would look terrible. – lulu Aug 15 '20 at 22:47
  • You should be aware that people model cash flows related to business. A lot of time and energy is spent on that. Just looking at a handful of numbers and extrapolating numerically is no substitute for that. But, if you are lucky, maybe some simple measure is good enough. – lulu Aug 15 '20 at 22:49
  • Ok so from your logic historical data of previous five quarter gives a mean profit percent of 1.8, so how can I use that to find profit for the current quarter that has passed, just say that it will be 1.8% of the revenue that I had? – Alpha Bing Aug 15 '20 at 23:06
  • That's one way to get a result, sure. But, to be clear, I'm not claiming anything about that result...it's just easy to produce. Is the profit $1.8%$ in all prior quarters? – lulu Aug 15 '20 at 23:07
  • I see. No, the 1.8% figure is the mean of all the quarters. Individually, I have this: 2.00, 1.844, 1.818, 1.8636, 1.633 (all in percentages) – Alpha Bing Aug 15 '20 at 23:09
  • Well, I'd call that pretty stable really. There are businesses for which you'd get much wilder behavior. – lulu Aug 15 '20 at 23:10
  • So this is it? Is there any other approach that I can use, what do you think? – Alpha Bing Aug 15 '20 at 23:10
  • Based on such little information, my advice would be to keep it simple. Of course, if you have more information about the cashflows you can embed that in your estimate, but if all you have are these numbers then I'd keep it simple and stick with this. – lulu Aug 15 '20 at 23:12
  • I see. Do you happen to know any resource or lesson that I can refer to that might be of some help in questions/problems like these? – Alpha Bing Aug 15 '20 at 23:12
  • This isn't really a math question...it's much more of a business question. I'd look at how active businesses model their cash flows. – lulu Aug 15 '20 at 23:15
  • If I were to take it more complex or legit way with this limited information, say some theorem or concept considering multiple assumptions, do you think you have some way? – Alpha Bing Aug 16 '20 at 00:04
  • No. Like I say, this is a business problem, not a math problem. – lulu Aug 16 '20 at 00:36
  • If you want to make it sound more mathematical, you could take a linear regression model of the data (here a data point would be a pair (Revenue, Profit)). That might pick up some correlations between the magnitude of the revenue and the profit margin. Again, with so few data points and no business modeling, this sort of study wouldn't mean much. But if the linear fit really seemed to match the data well, you might take some comfort. – lulu Aug 16 '20 at 00:42
  • But linear regression for just 5 data points doesnt mean much, even if I tried some correlation coefficient, in that case it would generally give a higher relation significance – Alpha Bing Aug 16 '20 at 00:49
  • Well, that's what I keep telling you. $5$ data points isn't worth much. That's why my (strong) preference is to keep it simple and just take the mean. Should say, though, that I wouldn't pay much attention to a linear regression model for business earnings anyway. Not unless it is accompanied by a thoughtful business analysis of the cash flows. – lulu Aug 16 '20 at 00:51
  • I see. Thank you so very much! That definitely helps! – Alpha Bing Aug 16 '20 at 01:11

0 Answers0