Download to read offline
Demystifying Value Presentationppt2.pptx
- 1. Proprietary + Confidential SarahN / November 5 Demystifying value The importance of lifetime value
- 2. Proprietary + Confidential Whatis Customer Lifetime Value? A long-term prediction of the future value of your customers’ interactions ● It is not a historical average ● It is a long-term oriented prediction ● It is valued at the individual level ● It is impactful across the entire value chain
- 3. Proprietary + Confidential OneYear Expected Value Total Future Customer Value Percent of Total Equity 1 $140.00 $80,000,000 80% 2 $30.25 $10,000,000 10% 3 $18.00 $5,000,000 5% 4 $14.75 $3,000,000 3% 5 $10.00 $2,000,000 2% Total $81.86 $100,000,000 100% LTV Level 1: This is the most basic output of LTV The most basic output is a prediction of the future revenue you will get from your current customers.
- 4. Proprietary + Confidential OneYear Expected Value Total Future Customer Value Percent of Total Equity Chance of Churn 1 $140.00 $80,000,000 80% 30% 2 $30.25 $10,000,000 10% 20% 3 $18.00 $5,000,000 5% 50% 4 $14.75 $3,000,000 3% 50% 5 $10.00 $2,000,000 2% 80% Total $81.86 $100,000,000 100% 50% LTV Level 1: This is the most basic output of LTV The most basic output is a prediction of the future revenue you will get from your current customers.
- 5. Proprietary + Confidential AcquisitionChannel One Year Expected Value Total Future Customer Value Percent of Total Equity UAC $150.00 $80,000,000 80% Social $130.25 $10,000,000 10% Organic $18.00 $5,000,000 5% Channel 4 $24.75 $3,000,000 3% Channel 5 $12.00 $2,000,000 2% Total $81.86 $100,000,000 100% LTV Level 2: Looking at LTV by Acquisition Channel Most people stop here, but you can go a lot further.
- 6. Proprietary + Confidential DeviceOne Year Expected Value Total Future Customer Value Percent of Total Equity App $150.00 $40,000,000 40% Mobile $130.25 $30,000,000 30% Desktop $100.00 $2,000,000 20% Total $81.86 $100,000,000 100% LTV Level 2: Looking at LTV by Acquisition Channel Learn your most valuable devices
- 7. Proprietary + Confidential FirstProduct Category Purchased One Year Expected Value Total Future Customer Value Percent of Total Equity Romantic Comedy Lovers $150.00 $51,000,000 64% Horror Film Lovers $77.99 $18,000,000 22.5% Action Lovers $52.10 $6,000,000 7.5% Drama Lovers $19.72 $4,000,000 5% Other $12.00 $4,000,000 5% Total $140.00 $80,000,000 100% LTV Level 3: Segmenting customers differently, looking for additional insights This chart below looks at what brought in the top 20% of customers.
- 8. Proprietary + Confidential FirstProduct Category Purchased One Year Expected Value Total Future Customer Value Percent of Total Equity Chance of Churn Romantic Comedy Lovers $150.00 $51,000,000 64% 40% Horror Film Lovers $77.99 $18,000,000 22.5% 10% Action Lovers $52.10 $6,000,000 7.5% 10% Drama Lovers $19.72 $4,000,000 5% 10% Other $12.00 $4,000,000 5% 10% Total $140.00 $80,000,000 100% 20% LTV Level 3: Segmenting customers differently, looking for additional insights This chart below looks at what brought in the top 20% of customers.
- 9. Proprietary + Confidential Mostoften category purchased One Year Expected Value Total Future Customer Value Percent of Total Equity Romantic Comedy Lovers $250.00 $31,000,000 39% Action Lovers $65.73 $29,000,000 36% Drama Lovers $51.10 $11,000,000 14% Other $18.40 $5,000,000 6% Horror Film Lovers $10.00 $4,000,000 5% Total $140.00 $80,000,000 100% LTV Level 3: Segmenting customers differently, looking for additional insights This answers what product types brought your best customers coming back.
- 10. Proprietary + Confidential UnderstandingLTV allows you to segment your customer base, and tailor your approach. Likely to churn, incentivise Good traction, engage & upsell Activate inactive customers Acquire more who act like them
- 11. Proprietary + Confidential Proprietary+ Confidential Tools for LTV Modeling BTYD Survival Analysis Feature Selection Use recency/frequency/value models to extrapolate lifetime value in non- contractual setting. lifetimes Use predictors to determine probability at time t of user’s subscription being “alive”. lifelines Determine the most important in-app actions that correlate with lifetime value. scikit-learn
- 12. Proprietary + Confidential Whatdata is needed for pLTV Unique identifier Transaction value Date of transaction Install Date
- 13. Proprietary + Confidential Whyis the work worth it? + it’s not as hard as it used to be ● Better decide who to target and who to exclude from targeting ● Refine product/service offering to highest value ● Determine most efficient way to drive customer loyalty ● Waste fewer marketing dollars
- 14. Proprietary + Confidential Thereare many inputs to a good CLV model First purchase App Install Subscription Sign Up Lead Submitted Application 1st Repeat Purchase 1st Paid Action Upgrade Subscription Lead Closed Approval Long term spend & Churn High Value Customer Acquisition Development Retention
- 15. Proprietary + Confidential Thereare many inputs to a good CLV model First purchase App Install Subscription Sign Up Lead Submitted Application 1st Repeat Purchase 1st Paid Action Upgrade Subscription Lead Closed Approval Long term spend & Churn High Value Customer Acquisition Development Retention ad solutions predict these inputs
- 16. Proprietary + Confidential Thereare many inputs to a good CLV model First purchase App Install Subscription Sign Up Lead Submitted Application 1st Repeat Purchase 1st Paid Action Upgrade Subscription Lead Closed Approval Long term spend & Churn High Value Customer Acquisition Development Retention digital behavior & pre-fab ML analysis helps predict this transactional behavior helps predict this
- 17. Proprietary + Confidential Thereare many inputs to a good CLV model First purchase App Install Subscription Sign Up Lead Submitted Application High Value Customer Acquisition Development Retention Machine Learning Solutions: - Clustering - Regression - Random Forest - Deep Neural Net Probabilistic Models: - Pareto/NBD - BG/NBD - BG/BB - Survival Curves
- 18. Proprietary + Confidential Wheredo you go from here? Today Tomorrow Next Year Identify your objectives, aligned to business strategy Investigate how customer data is stored, labeled, and formatted Begin exploring predictions to customer behavior using pre-fab models (building from scratch only if needed) Utilize tools and partnerships to push toward automation and new insights Evaluate customer response to strategy
- 19. Proprietary + Confidential Topways to action off of LTV 1. Bidding (UAC for Value) 2. Re-engagement 3. Feature Selection 4. Acquire customers similar to your best customers, raising the average LTV of your whole entire customer base!
- 20.
Editor's Notes
- #3 Comment here: -looking into the future -- can say oh hey burger king your goal was 150 million? With 95% accuracy, i can see that you are going to be $50 mm short of that if you keep things the same now. Points here -basic output: blue box. Doing this with really high accuracy. -segments, top 20% -really high accuracy- about 90% accurate or higher. -looking into the future - help with customer goal, use movies anywhere as an example
- #4 I'm calling out that you can use LTV to figure out the chance of churn. This will leave you room to segway into reegnagement Point here: alot of ppeople think that LTV is just an acquisition play, but its also a retention play
- #5 Maybe also, only focusing on the top 20% of customers.
- #6 will comment here: how many of you know if an app or a web user is more valuable to your client? Now you know.
- #7 Movies anywhere is the example here call out here is that we could also look at just app users, in addition to just looking at top 20% of customers. Can see which types of movies bring in your best customers
- #8 Chance of churn= 1-p-alve P-alive is their chance of coming back the call out here is that LTV can predict the chance of someone churning. Therefore we know who to retarget Chance of churn
- #9 Now let’s look at most often purchased
- #10 Key take away: segment!! A lot of people think that LTV is an acquisition play, but it also a retention play. Marketing! What can you actually do when you know the LTV of customers? Here’s an example: Let’s say you looked at your customer base, the value it drives for your business, you measured CLV of all your users, and you identified a few (4) ‘types’ of people: buckets or segments. It could look something like this + you could adjust how you treat these users through marketing efforts or inside the game (game play / monetization).
- #11 call out here is FS. Bc often if LTV isn't actionable it's bc what we need is FS! The stats and machine learning are built in! if LTV isn't actionable, it's usually bc what they actually need is feature selection Will comment that we do not own these models