Successfully reported this slideshow.
Your SlideShare is downloading. ×

Data Driven Growth at TransferWise

Ad
Ad
Ad
Ad
Ad
Ad
Ad
Ad
Ad
Ad

Check these out next

1 of 155 Ad

Data Driven Growth at TransferWise

Download to read offline

These are the slides of the first talk of the second Tech Talk@TransferWise Singapore, which happened on the 1st of March 2018.

In this talk, we reveal how TransferWise uses data science to drive it's rapid growth. We share what is mission driven startup and explain TransferWise product pillars. Take a glimpse on the 'speed' pillar and how we reinforce it in product-engineering.

These are the slides of the first talk of the second Tech Talk@TransferWise Singapore, which happened on the 1st of March 2018.

In this talk, we reveal how TransferWise uses data science to drive it's rapid growth. We share what is mission driven startup and explain TransferWise product pillars. Take a glimpse on the 'speed' pillar and how we reinforce it in product-engineering.

Advertisement
Advertisement

More Related Content

Slideshows for you (20)

Similar to Data Driven Growth at TransferWise (19)

Advertisement

Recently uploaded (20)

Data Driven Growth at TransferWise

  1. 1. Data Driven Growth. Using data to power the growth of a Mission Driven Startup
  2. 2. @nilanp @balazsbarn I am Vadym Product Engineer at TransferWise
  3. 3. @nilanp How does money moves around the world ?
  4. 4. @nilanp We keep it in a bank account
  5. 5. @nilanp
  6. 6. @nilanp
  7. 7. @nilanp
  8. 8. @nilanp
  9. 9. 13
  10. 10. @nilanp 1. Progress 2. Culture 3. Marketing
  11. 11. @nilanp @balazsbarn TransferWise… a Mission Driven Startup ?
  12. 12. @nilanp TransferWise was started to solve a problem @nilanp@nilanp
  13. 13. @nilanp In our case a personal pain point helped to see an epidemic dysfunction in the entire industry, which lacked any natural force of improvement or move towards resolution.  — Kristo Käärmann & Taavet Hinrikus@nilanp
  14. 14. @nilanp @cataflu A Mission Driven Startup = A catalyst for achieving a change in the world @nilanp
  15. 15. @nilanp @balazsbarn 1. Progress
  16. 16. @nilanp @cataflu £0.1 Million ➤ £1 Billion ➤ £1.5 Billion
  17. 17. @nilanp @cataflu 60 ➤ 940 TransferWisers we are more than doubling in size every year
  18. 18. @nilanp @cataflu Evangelism is the biggest driver of growth
  19. 19. @nilanp @balazsbarn Why do people recommend our product ?
  20. 20. @nilanp Would you recommend us to your friends? 1 2 3 4 5 6 7 8 9 10 Detractors Neutral Promoter s
  21. 21. @nilanp Would you recommend us to your friends? 1 2 3 4 5 6 7 8 9 10 Detractors Neutral Promoter s Tota l Promoters - Detractors = NPS
  22. 22. @nilanp The NPS Richter scale
  23. 23. @nilanp 1 2 3 4 5 6 7 8 9 10
  24. 24. @nilanp
  25. 25. @nilanp
  26. 26. @nilanp
  27. 27. @nilanp
  28. 28. @nilanp
  29. 29. @nilanp
  30. 30. @nilanp NPS IS the north star KPI for growth
  31. 31. @nilanp If promoters > detractors you WILL grow
  32. 32. @nilanp N.B. This is hard.
  33. 33. It works ! (But barely)
  34. 34. It works ! (But barely) It works with no significant issues !
  35. 35. It works ! (But barely) It works with no significant issues ! It works. Wow that was slick
  36. 36. It works ! (But barely) It works with no significant issues ! It works. Wow that was slick You’ve just blown my socks off that was incredible
  37. 37. @nilanp @balazsbarn RATIONalRational Emotional
  38. 38. @nilanp @balazsbarn RATIONalProduct Cause
  39. 39. @nilanp @balazsbarn RATIONal Speed Price Convenience Coverage Cause
  40. 40. @nilanp @balazsbarn To achieve advocacy your product needs to be an order of magnitude better than the alternative
  41. 41. @nilanp @balazsbarn But don’t do this as “growth hack” do it because the world deserves it
  42. 42. @nilanp @balazsbarn
  43. 43. @nilanp @balazsbarn
  44. 44. @nilanp @balazsbarn
  45. 45. @nilanp @balazsbarn
  46. 46. @nilanp @balazsbarn RATIONal Speed Price Convenience Coverage Cause
  47. 47. @nilanp @balazsbarn
  48. 48. @nilanp @balazsbarn Money without borders. Instant, convenient, transparent and eventually free We’re powering money for people and businesses: to pay, to get paid, to spend, in any currency, wherever you are, whatever you’re doing
  49. 49. You can’t copy that !
  50. 50. @nilanp @balazsbarn So… 10x product + Cause driven marketing = Explosive Word of Mouth growth @nilanp @cataflu
  51. 51. @nilanp @balazsbarn ONE question:
  52. 52. @nilanp @balazsbarn So how you do build a 10x product ?
  53. 53. @nilanp Autonomous independent teams that focus on KPIs that make a difference to customers and drive growth @nilanp
  54. 54. @nilanp Autonomous independent teams that focus on KPIs that make a difference to customers and drive growth @nilanp Currencies OperationsVerificationConversion Experience ViralityMarketing Regulated and available Instantly and almost zero cost IdentityOnboarding Worry Evangelism Problem awareness
  55. 55. Where actually regional team comes?
  56. 56. How to know what actually matters to the customer?
  57. 57. Customer Interviews
  58. 58. Customer Surveys
  59. 59. Customer Feedback
  60. 60. @nilanp@nilanp Data scientist
  61. 61. P O R D U C T E I N G N E E R
  62. 62. Enginee rs Product Manage rs Data Scientis ts Complia nce + Banking Design Support + Operatio ns Work within teams: To help teams understand the impact of their actions • Build underlying data infra • Eval impact of AB tests • Analyse, identify and prioritise opportunities • Build product X Tell the team what to do next
  63. 63. @nilanp @balazsbarn
  64. 64. @nilanp @balazsbarn
  65. 65. @nilanp @balazsbarn
  66. 66. Build something that does not exist today
  67. 67. @nilanp
  68. 68. @balazsbarn@nilanp I am Balazs Barna Product Engineer at TransferWise
  69. 69. Building speed as a product.
  70. 70. @balazsbarn@nilanp Evolution of the delivery estimator
  71. 71. @balazsbarn@nilanp Statistical model
  72. 72. @balazsbarn@nilanp Statistical model Rule based engine
  73. 73. @balazsbarn@nilanp Statistical model Rule based engine Controlling speed
  74. 74. @balazsbarn@nilanp 10x cheaper product to move money across borders.
  75. 75. @balazsbarn@nilanp Oh, and one more thing!
  76. 76. @balazsbarn@nilanp We are also faster. WE ARE ALSO FASTER!
  77. 77. @balazsbarn@nilanp GBP to EUR SPEED
  78. 78. @balazsbarn@nilanp
  79. 79. @balazsbarn@nilanp What is the TransferWise way?
  80. 80. @balazsbarn@nilanp SIMPLE.
  81. 81. @balazsbarn@nilanp SIMPLE. EASY TO IMPLEMENT.
  82. 82. @balazsbarn@nilanp SIMPLE. EASY TO IMPLEMENT. BRINGS CUSTOMER VALUE.
  83. 83. @balazsbarn@nilanp We don't need a perfect solution.
  84. 84. @balazsbarn@nilanp What is a transfer?
  85. 85. @balazsbarn@nilanp
  86. 86. @balazsbarn@nilanp Let's find similar ones in the past, and see how fast we were.
  87. 87. @balazsbarn@nilanp Let's find similar ones in the past, and see how fast we were. Pick the 60th percentile.
  88. 88. @balazsbarn@nilanp Each end-to-end estimate is the sum of independently calculated component estimates: Estimate each step of the lifecycle Got the money! (2h) Money is on it's way! (1h) Money arrived! (2h) + + + KYC done! (15 min) + Total estimated delivery time: 6 hours 15 minutes Converted the money! (1h)
  89. 89. @balazsbarn@nilanp Why the 60th percentile? Why the 60th percentile? We moved from the 95th percentile to provide more accurate rather than safer estimations to match the main happy path payments.
  90. 90. @balazsbarn@nilanp MISSION ACCOMPLISHED We achieved something that brings value to all of our customers.
  91. 91. @balazsbarn@nilanp 2016 Q4: Speed team in Budapest inherits the estimator
  92. 92. @balazsbarn@nilanp What does speed mean to our customers?
  93. 93. @balazsbarn@nilanp #WHEREISMYMONEY?
  94. 94. @balazsbarn@nilanp Slower transfer means more contacts
  95. 95. @balazsbarn@nilanp Bad estimate results in contract rate 10x
  96. 96. @nilanp @balazsbarn RATIONal Contacts cost money. Charge as little as possible.
  97. 97. @balazsbarn@nilanp SPEED & ESTIMATION ACCURACY MATTERS TO OUR CUSTOMERS
  98. 98. @balazsbarn@nilanp Learning by doing.
  99. 99. @balazsbarn@nilanp Learning by doing. Bug fixes.
  100. 100. @balazsbarn@nilanp Learning by doing. Bug fixes. Investigating bad estimates.
  101. 101. @balazsbarn@nilanp Learning by doing. Bug fixes. Investigating bad estimates. Super hard without a database.
  102. 102. @balazsbarn@nilanp We built our own database.
  103. 103. @balazsbarn@nilanp We built our own database. Saved every estimate.
  104. 104. @balazsbarn@nilanp We built our own database. Saved every estimate. Saved the transfer.
  105. 105. @balazsbarn@nilanp We built our own database. Saved every estimate. Saved the transfer. Easy to compare actual vs estimated.
  106. 106. @balazsbarn@nilanp THE ESTIMATOR HAS OBVIOUS LIMITATIONS The estimations are poor, when we don't have sufficient data.
  107. 107. @balazsbarn@nilanp Card payment is introduced for Hungary, and the estimates are very pessimistic.
  108. 108. @balazsbarn@nilanp We don't have sufficient data on card payments to estimate!
  109. 109. @balazsbarn@nilanp If the customer pays by card, we know it's going to take X seconds to receive the money.
  110. 110. @balazsbarn@nilanp No need to rely on statistics!
  111. 111. @balazsbarn@nilanp Let's create a rule based engine!
  112. 112. @balazsbarn@nilanp If we have an applicable rule for a transfer, we don't rely on statistics.
  113. 113. @balazsbarn@nilanp TransferWise supports 700 routes.
  114. 114. @balazsbarn@nilanp TransferWise supports 700 routes. Most rules are currency specific.
  115. 115. @balazsbarn@nilanp TransferWise supports 700 routes. Most rules are currency specific. Our team can't do it alone.
  116. 116. @balazsbarn@nilanp Easy to use microservice & framework.
  117. 117. @balazsbarn@nilanp Easy to use microservice & framework. Any team can implement their own rules in fifteen minutes.
  118. 118. @balazsbarn@nilanp Estimation accuracy improved by 30%.
  119. 119. @balazsbarn@nilanp If we just keep doing what we are doing... That's pretty good.
  120. 120. @balazsbarn@nilanp If we just keep doing what we are doing... That's pretty good. If we just keep fixing routes, we are going to be awesome!
  121. 121. @balazsbarn@nilanp If we just keep doing what we are doing... That's pretty good. If we just keep fixing routes, we are going to awesome!
  122. 122. @balazsbarn@nilanp LET'S TAKE A STEP BACK
  123. 123. @balazsbarn@nilanp
  124. 124. @nilanp @balazsbarn RATIONal OUR PRICING IS CONSISTENT AND 10X BETTER OUR SPEED IS AN ESTIMATE
  125. 125. @balazsbarn@nilanp WE NEED TO CONTROL SPEED
  126. 126. @balazsbarn@nilanp HOW CAN WE DO THAT?
  127. 127. @balazsbarn@nilanp Instead of estimating, we target Got the money! (2h) Money is on it's way! (1h) Money arrived! (2h) + + + KYC done! (15 min) + Targeted delivery time: 6 hours 15 minutes Converted the money! (1h)
  128. 128. @balazsbarn@nilanp When we offer a speed to a customer, it has to be based on a plan.
  129. 129. @balazsbarn@nilanp This is where the magic happens
  130. 130. @balazsbarn@nilanp Money movement between two countries is never equal.
  131. 131. @balazsbarn@nilanp More money will move from the US to Japan than the opposite.
  132. 132. @balazsbarn@nilanp We have to provide liquidity!
  133. 133. @balazsbarn@nilanp We have to provide liquidity! Cool, what's the easiest way around this?
  134. 134. @balazsbarn@nilanp KEEP A LOT OF MONEY THERE
  135. 135. @balazsbarn@nilanp KEEP A LOT OF MONEY THERE
  136. 136. BUILD SOMETHING SUSTAINABLE
  137. 137. @balazsbarn@nilanp Liquidity need can be predicted for lower amounts.
  138. 138. @balazsbarn@nilanp What if we get a 500 000 USD transfer?
  139. 139. @balazsbarn@nilanp What if we get a 500 000 USD transfer? If we payout right away, other customers will have to wait.
  140. 140. @balazsbarn@nilanp Estimator talks to liquidity-service.
  141. 141. @balazsbarn@nilanp Estimator: When will you get the money, my friend? Liquidity: 4 PM Today. Estimator: Have the money by 4. I'm going tell the customer.
  142. 142. @balazsbarn@nilanp Now, we know when the payout will happen, before we gave a promise.
  143. 143. @balazsbarn@nilanp The estimator became the orchestrator of TransferWise.
  144. 144. @balazsbarn@nilanp It negotiates with all the components, and creates a plan before giving a promise.
  145. 145. @balazsbarn@nilanp SUMMARY
  146. 146. @balazsbarn@nilanp Start with something simple and learn on the job.
  147. 147. @balazsbarn@nilanp Think about debugging & monitoring.
  148. 148. @balazsbarn@nilanp Think about debugging & monitoring. (Think one more time).
  149. 149. @balazsbarn@nilanp Data analysis can lead to take baby steps & miss the game changer.
  150. 150. @balazsbarn@nilanp Be ready to change everything.
  151. 151. Thank you. @nilanp @balazsbarn

×