Lyft – User Based Valuation

Image Credits: grist.org, lyft

Edit (9th March 2019) – Using the user based statistics and financial information made public in Lyft’s S-1 prospectus, I have revalued Lyft at ~$17 billion. My valuation is presented in the link alongside https://investandrise.com/lyft-ipo-2019-valuation/

Ride sharing companies have revolutionized the way we commute. Both Uber and Lyft would be going public next year. While Uber has gone global and continues to expand to other businesses, Lyft has shown a much smaller narrative and is present only in US and Canada.

The most important parameter for companies born in the gig economy is the number of users/subscribers. VCs typically value (read price) such companies by “pricing” users/subscribers. The intrinsic value of an asset or business is the present value of its future cash flows (DCF – Discounted Cash Flow Valuation). So, an ideal way to value such companies would be to value users

In this blog post, I have attempted to value Lyft by valuing users using a framework taught by the renowned NYU Stern Prof. Aswath Damodaran. In his paper, Prof. Damodaran has explained how to incorporate user economics in a DCF Valuation or rather how would you use DCF to value user based companies born in the gig economy

The fundamental equation to value such companies that Prof. Damodaran gives is simple and intuitive:

Value of a user based company = Value of existing users + Value added by new users – Value eroded by corporate drag

Lyft’s being priced at $15.1 bn (as of this writing). Using a user based valuation,  I have valued Lyft at just under $5 bn by taking certain base case assumptions. The assumption values could be high or low. I do not claim any certitude to these numbers. To accommodate for different cases (or different values of assumption variables), I go on to use monte carlo to get a distribution of Lyft’s valuation across simulation trials. I find that ~80% of the distribution falls below Lyft’s current pricing of ~15 bn

In Lyft’s valuation below, I estimate values for Lyft’s existing users, new users and the value eroded by corporate drag

User Based Lyft’s Valuation

Since Lyft is yet to go public, so financials are hard to come by. I have used the 2018 Q1, Q2 & Q3 financials reported in the information to make estimates for Q4 and then used the base estimates for 2018 to forecast cashflows into the future

Exhibit 1: Estimates based on 2018 Q1, Q2 & Q3 as reported by The Information

a. Value of Existing Users

Simply put, the value of existing users is arrived at by estimating after tax operating profit per existing user for the base year and then forecasting it into the future, followed by taking the present value (PV) of the future cash flows. The PV per user is scaled by the number of existing users to arrive at the PV of after tax operating profit of all existing users. This PV is then slashed using the assumed probability of user lifetime

Base Year 2018 Estimates: (Data from Exhibit 1)

  • Base Year Operating Profit = 44.8% of Base Year Net Revenue per Existing User (this operating profit has been reported after deducting the cost of revenue only from the net revenue. Cost of revenue consists of insurance, credit card fees and technical infrastructure. It does not include sales & marketing, R&D and employee expenses)
  • Base Year Net Revenue per Existing User ($67.4 mn)= Estimated Net Revenue ($2129 mn) / Number of Users (Estimated *32 mn)

Assumption Variables for making Forecasts : 

  1. User Lifetime: For the base case, User Lifetime is assumed to be 15 years (This is an assumption. I do not have data backing user lifetime. I go on simulate this parameter to take on different values ranging from 4 to 20 years using a discrete triangular distribution)
  2. Probability of User Full Life: Annual Renewal Probability assumed at 95% ^ User Lifetime. Considering that ride sharing businesses have disrupted the market with many users preferring it over their own cars (I certainly do!), I reckon that a major chunk of the existing ride sharing users would stick, although their loyalty to one company is uncertain. A subscription business model would have more user stickiness as opposed to a transaction based. I have attempted to accommodate the uncertainty/variability in user stickiness, by inducing the probability of user full life take on different values in my simulation. (Since user lifetime is simulated to take on different values, hence probability of user full life becomes a variable as well. )
  3. Growth Rate (of Net Revenue): For the base case, Net Revenue per User is assumed to grow at 15% for the first 5 years, at 10% for the next 5 years and then at the risk free rate.  Again, these values could be high or low! I reckon that the company would mature after 10 years and hence, grow at the risk free rate. (To accommodate for the uncertainty on growth rate,  I simulate this parameter over a range of 8% to 20% using an asymmetric positively skewed continuous distribution)
  4. Growth Rate (of Cost of Servicing Existing Users): [x% * growth rate (of net revenue) ] + [(1-x%) * inflation rate], where x is assumed to be 80% for the base case (x is simulated to take on different values ranging from 70% to 100% using an asymmetric negatively skewed continuous distribution)
  5. Discounting Factors – Cost of Capital reflecting CashFlow uncertainty or User Risk: For the base case, cost of capital is taken as 10%, which is the 75th percentile for US companies. I just need to ensure that the cost of capital for existing users is lower than that of acquiring new users since the cash flows from new users would be more uncertain/risky. (Cost of capital is made to take on values ranging from 8% to 12% on a normal distribution. No skewness assumed since existing users would have relatively low risk vis-a-vis new users)
  6. Number of Users: As per Forbes, the number of users in 2017 were 23 mn. I couldn’t find the number of users in 2018, so I had to estimate it. Here is how –
  • Lyft achieved 1 billion rides in Sept 2018. It was at 500 mn rides around the same time frame last year. This translates into a compounded monthly growth of just under 6%
  • Extrapolating this growth till December of this base year 2018, I get an additional ~200 mn rides i.e. 1200 mn rides totally.
  • Taking out the number of rides till December 2017 (i.e 1200mn – (500+1.059^4)), I get ~600 mn rides in 2018
  • One of the other statistic that I found is that on an average each user took 19 rides. So, the *number of users in 2018 could be estimated as 600mn rides/19 rides per user ~ 32 mn users.  

*This is a crude method to estimate users and I would want to replace this estimate with the actual number of users as when that statistic becomes public*

Operating Profit (for each year in the future) = Net Revenue forecast  – Cost of Servicing Existing Users forecast

The future operating profit is discounted using cost of capital and then slashed using the probability of user full life

Exhibit 2: Value of Existing Lyft Users $6.1 bn (base case)

b. Value Added by New Users 

New users in the base year is the increase in users in base year 2018 over 2017. Base year value added by each new user is the amount by which value per existing user exceeds cost of adding a new user

Base Year 2018 Estimates:

  • Cost of Adding New Users is the amount spent over and above the spend on servicing Existing Users (and also excluding Corporate Expenses)
  • Amount spent over and above the spend on servicing Existing Users  = Operating Profit (on servicing existing users) + Net Loss Amount – Corporate Expenses
  • Therefore, Cost of Adding a New User = (Net Revenue + Net Loss – Cost of Servicing all Existing Users – Corporate Expenses) / (Users in 2018 – Users in 2017)
  • Base Year Value Added by New User = Value per Existing User – Cost of Adding a New User

Assumption Variables for making Forecasts :  

  1. Growth Rate (in # Users): Assumed 25% for the first 5 years and 10% for the next 5 years
  2. Annual Renewal Probability: Assumed at 95%.  I have attempted to accommodate the uncertainty/variability in user stickiness, by inducing the probability of user full life take on different values in my simulation.
  3. Discounting Factors – Cost of Capital reflecting CashFlow uncertainty or User Risk: For the base case, cost of capital is taken as 12%, which is higher than the cost of capital for existing users. A cost of capital of 12% occurs at the 90th percentile for US companies. (Cost of capital is made to take on values ranging from 9% to 18% using an asymmetric negatively skewed continuous distribution. I have assumed a negatively or left skewed distribution since existing new users would have relatively high risk vis-a-vis existing users)

New Users are estimated to increase at an assumed growth rate and decrease with the assumed annual renewal probability each year. Value per New User is forecast to increase each year with the inflation rate

Exhibit 3: Value added by New User $3.1 bn (base case)

c. Corporate Drag 

Corporate Drag of $400 mn in the base year is an assumed base case value. (It has been simulated to take different values ranging from $200 mn to $600 mn using a normal distribution)

Moreover, Corporate Drag is assumed to grow at 4% every year

Exhibit 4: Corporate Drag estimated at $4.4 bn (base case)

Exhibit 5: (base case) Value of Lyft $4.8 bn = Value of existing users $6.1 bn + Value added by new users $3.1 bn – Value eroded by corporate drag $4.4 bn

On running a monte carlo simulation, I get the following distribution for Lyft’s valuation (Exhibit 6)

Exhibit 6:  Distribution of Lyft’s valuation across simulation trials

Key statistics/observations from the distribution above:

  1. My base case valuation of $4.8 bn falls at the 44th percentile i. e. 44% of the distribution values are below $4.8 bn
  2. Median i.e. the 50th percentile occurs at a valuation of ~$6 bn
  3. Lyft’s “pricing” of $15.1 bn occurs near the 80th percentile, which means that ~80% of the values that I get are lower than this price
  4. Another interesting observation is that ~26% of values of the distribution are negative. This could be interpreted as Lyft’s probability of default

Valuation is very sensitive to the number of users. The above is just a snapshot in time valuation based on a lot of assumptions. Closer to its IPO in 2019, when Lyft makes its financials public, then I would replace the assumptions with actuals to arrive at its fair valuation