Two Methods for Modeling LTV with a Spreadsheet

  • View
    5.843

  • Download
    14

  • Category

    Business

Preview:

DESCRIPTION

NB! The bitly link in the deck is hard to read, please use this one: http://bit.ly/2LTVXLS This is the presentation I gave at Slush 2013 in Helsinki, Finland. It describes two methods for modeling Lifetime Customer Value (LTV) in Excel. Linked within the presentation is a spreadsheet exemplifying both methods against 100k rows of fake user data that I generated with a Python script to "look" real (although they probably don't).

Citation preview

TWO METHODS FOR MODELING LTV WITH A SPREADSHEET

Eric Benjamin SeufertSlush 2013

WHO AM I?

WHO AM I?Head of Marketing at Wooga

WHO AM I?Head of Marketing at Wooga

MobileDevMemo.com

WHO AM I?Head of Marketing at Wooga

MobileDevMemo.com

Freemium Economics

DOWNLOAD THE SPREADSHEET100k rows of fake data

DOWNLOAD THE SPREADSHEET100k rows of fake data

• Probably doesn’t look like real data!

DOWNLOAD THE SPREADSHEET100k rows of fake data

• Probably doesn’t look like real data!

http://bit.ly/183lFwp

WHAT IS LTV?

FUNDAMENTALLY A PROJECTION

PRIMARILY USEFUL FOR MARKETING

DIFFICULT TO CALCULATE FOR FREEMIUM GAMES

USUALLY CALCULATED PROGRAMMATICALLY

GOOD REASONS TO CALCULATE LTV IN A SPREADSHEET

Prototyping using comparable data

GOOD REASONS TO CALCULATE LTV IN A SPREADSHEET

Prototyping using comparable dataProject investment decisions

GOOD REASONS TO CALCULATE LTV IN A SPREADSHEET

Prototyping using comparable dataProject investment decisionsSprint prioritizations

GOOD REASONS TO CALCULATE LTV IN A SPREADSHEET

Prototyping using comparable dataProject investment decisionsSprint prioritizations

General strategic decision making

GOOD REASONS TO CALCULATE LTV IN A SPREADSHEET

WHAT ARE THE INPUTS TO LTV?

WHAT ARE THE INPUTS TO LTV?Monetization (Value)

WHAT ARE THE INPUTS TO LTV?Monetization (Value)Retention (Lifetime)

WHICH DIMENSIONS?

WHICH DIMENSIONS?

WHICH DIMENSIONS?

LTV TIMELINE180 days or 365 days?Spreadsheet shows both

RETENTION APPROACH

RETENTION APPROACH

Real Retention Rates

ESTIMATING LIFETIME

Power function

ESTIMATING LIFETIMEPower Function: y = a * x ^ b

a -> =EXP(INDEX( LINEST( LN(Known Ys), LN(Known Xs)), 2))

b -> =INDEX(LINEST(LN(Known Ys), LN(Known Xs)), 1)

ESTIMATING LIFETIMEArea under the curve

ESTIMATING LIFETIMECalculating lifetime:

• Retention curve essentially a survival function

More info: http://data.princeton.edu/wws509/notes/c7.pdf

Source: Survival Analysis, a Self-Learning Text (Klein 2005)http://books.google.de/books?id=GNhzxRkFnJ0C&lpg=PA262&ots=Z3foUkFg-4

Example survival curves for two groups ofpatients given different Leukemia treatments

ESTIMATING LIFETIMECalculating lifetime:

• Retention curve essential a survival function • Mean is value of integral

More info: http://data.princeton.edu/wws509/notes/c7.pdf

CALCULATING LTVBuild retention function:

CALCULATING LTVBuild retention function:

Retention curve power function model( y = a x ^ b )

CALCULATING LIFETIMEBuild retention function: Lifetime (180):

Integral (1 -> 180):Y = 0.61x ^ -0.37

= 25.3 Days

Lifetime (365): Integral (1 - > 365):y = 0.61x ^ -0.37

= 40.1 Days

CALCULATING LTVFor the trailing 7 days of data…

CALCULATING LTVFor the trailing 7 days of data……for users in the UK…

CALCULATING LTVFor the trailing 7 days of data……for users in the UK……that downloaded the game organically…

CALCULATING LTV

LTV(180):25.3 * $0.13 = $3.30

LTV(365):40.1 * $0.13 = $5.21

HONING THE MODEL

More real data

HONING THE MODEL

More real data

MONETIZATION APPROACH

MONETIZATION APPROACH

© 2014 Freemium Economics

MONETIZATION-BASED APPROACH

Real Average Daily Cumulative Spend Data

MONETIZATION-BASED APPROACH

Logarithmic Function

PROJECTING THE CURVELogarithmic Function: y = (c * LN( x )) + b

c -> =INDEX(LINEST(Known Ys, LN(Known Xs)), 1)

b -> =INDEX(LINEST(Known Ys, LN(Known Xs)), 1,2)

MONETIZATION-BASED APPROACH

LTV

CALCULATING LTVBuild monetization function:

CALCULATING LTVBuild monetization function:

Monetization curve logarithmic model( y = ( c * ln( x )) + b)

CALCULATING LTV

CALCULATING LTVUsing trailing 3 days of retention data…

CALCULATING LTVUsing trailing 3 days of retention data……for users in the US…

CALCULATING LTVUsing trailing 3 days of retention data……for users in the US……acquired through Flurry Video Ads…

CALCULATING LTVUsing trailing 3 days of retention data……for users in the US……acquired through Flurry Video Ads…

y(180) = ( .45 * ln (180)) + .41 = $2.75y(365) = ( .45 * ln( 365 )) + .41 = $3.06

HONING THE MODEL

HONING THE MODEL

THANKS!

eric@ufert.se

Recommended