Data Analysis Techniques

Linear Regression:
Here one dependent variable is described as a combination of a list of independent variables
Assumptions:
There is one dependent variable and a number of independent variables
Process:
Excel will let you run linear regression given a bunch of columns of independent variables
The distribution of independent variables should be normal?
Output:
An Equation. which describes which variables are significant in describing the variability in data of dependent variable
The result of the equation is a numeric value which can be used for forecasting or predicting
Sales = 0.8 * # of people entering the store
where sales is the dependent variable and
# of people entering the store is independent variable


Linear Discriminant Analysis
In order to discriminate or seperate out 2 ore more groups of data and characterizing the difference by a vector, we use Linear Discriminant Analysis.
Assumptions:
the independent variables need to be normally distributed
Process:
You can obtain LDA equation using R given a set of data
Output:
# of equations equal to the number of groups/classes -1
The result is used to find if a particular set of data belongs to group#1 or group#2 or group#3 or group#4 & so on
Eg. For 3 classes of flowers, there are 4 characteristics which classifies one flower from another
the 4 characterizing features are Sepal width, sepal length, petal width, petal length.
LD1 = 0.39 * sepal length + 2.067 * sepal width -2.27 * petal length -2.28 * petal width
LD2 = 0.56 * sepal length -2.45 * sepal width +0.41 * petal length -2.12 * petal width

LD1 line goes parallel through the means and LD2 line gives the distance between the means

If for a random flower we know the sepal length, sepal width , petal width and petal length, we can evaluate equation 2 to find which flower it is based on the sepal length, sepal width, petal length, petal width

Disadvantage:
It needs variance within each group to be same.

Logistic Regression
In order to run linear regression on a set of data but have an outcome which is dichotomous, we use logistic regression.We use logistic regression when we want to know which variables the dependent variable is dependent on.
Assumptions:
The inputs need not be normally distributed.
Output:
Output is an equation which is a natural log of the likelihood of the occurance of a state of the dependent variable.
eg.
1 = female
0 = male
ln(P(1)/(1-P(1)) = 0.2 + 0.4*height
Disadvantage:
It needs large sample size

ANOVA
tests whether the groups are different enough to be classified
the output of anova too is a true or false whether teh data columns are statistically different. There is a way they can be seperated
F value =SSE(between)/ SSE(Within)
Process:
You can run it in excel
Output
Output is an F value and probability of F-value. if p < 0.05, then the null hypothesis is rejected and the data is proven to be statistically different

How do you define a good Product?

Whether a product is tangible or intangible, the criteria that defines a good product are quite the same.
The four Ps of Product Marketing are: Product, Price, Promotion, Place. It is just impossible to define a good product without considering the other 3 Ps. however, here is my attempt at what makes a good product.
The product has to
1) Provide value and enrich the customer in some form or another.(Product)
2) Provide ease of use (Product)
3) Be unique. There has to be differentiation (Product)
4) Have a well defined use of product (Product)
5) Be interesting and exciting (Product)

Product Management in a Startup

I work for a start-up and as with any job in a start-up you end up taking a lot more responsibility than your title demands.
When I finished my MBA, I went to ask for a product management position. They said we are short of developers. So I ended up taking on Product Management as well as development role.
Now as I am performing product management responsibilities, I realized that its a not as glamorous a job as I thought.
There are multiple new products you are working on. At the same time you need to make sure you are not building your stepping stone for new customers on the grave of your existing customer. Bug fixes need to take priority as well.
There there are inputs from development team that we need to upgrade the platform and need new tools as technology advances.
But there are only so many resources available.
I devised a means to assign all Development team requests lower weight, all bug fixes highest weight and new product development moderate weight. However, this made the development team not so happy as well as the upper management not so happy. The professional services guys were happy though.
How can you make everyone happy? Is it even possible?
Should I use some other mechanism to intersperse the weights so that some new development and some development team requests can be accommodated with higher weights & priorities?

Business Valuation

What is free cash flow?
It is the cash that is free to be distributed to share holders of the company to keep investors happy.
Hence we calculate NOPAT (Net Operating Profit After Tax) and subtract Net CAPEX (Capital Expenditure) from the NOPAT. CAPEX is subtracted because it is the investment already made by the company to grow its revenue. hence it is not available to the share holders.Change in Net Working Capital is also subtracted from NOPAT since it seems that next year they might need that changed working capital in addition to the already deducted working capital from NOPAT.
FCF = NOPAT - Net CAPEX - Delta NWC

Calculate NOPAT
NOPAT is calculated easily (EBIT * (1 - Tax Rate)). However, it is the adjustments that need to be made on NOPAT that makes it complicated. Without adjustments, NOPAT value in itself is not sufficient to use in any kind of analysis.

Adjustments that go on NOPAT are any expenses that are one time expenses like restructuring charges or interest portion of operating lease payments since interest expense should not be deducted from NOPAT. Interest expense is not related to operations of the company so it is wise to add it back to Net Income/EBIT.

Calculate Net CAPEX

Capital Expenditure is money invested back into the company for future growth. Hence this money is not available to share holders. Hence it is deducted from the NOPAT to arrive at Free Cash flow.

The Capital Expenditure is mentioned in the Cash Flow Statement of company's financials.

Adjustments are made to Net CAPEX for depreciation. Depreciation is subtracted from capital expenditure All the investing activities which bring in money like sale of assets are subtracted from the CAPEX. All investing activities that result in outflow of cash are added to CAPEX like acquisitions.

Calculate Change in Net Working Capital (NWC)

Net Working Capital can be calculated as

Accounts Receivables + Inventory - Accounts Payable

Free Cash Flow (FCF)

FCF = NOPAT - CAPEX - change in NWC

Equity Valuation

To come to the Equity value of the company, you have to estimate the revenue for the next 5 years, estimate NOPAT as a % of Revenue, estimate CAPEX as % of Revenue, estimate change in NWC and calculate FCF for next 5 years. For years thereafter, assume a growth rate generally assumed to be 3%, and calculate Terminal Value. Calculate present Value for all the FCF values and sum them up to get the Equity Value of the Enterprise.

Debt Value

Debt value can be calculated by calculating present value for all future debt obligations and summing them up.

Enterprise Value

Enterprise Value is the sum of Debt Value and Equity Value.