A Multi-Touch Attribution In Mobile Retrospective With Trendyol
0 likes632 views
The document discusses multi-touch attribution (MTA) in mobile marketing, particularly focusing on its implementation by Trendyol, Turkey's largest e-commerce company. It outlines the challenges and advantages of MTA, including the need for accurate measurement of marketing value and the complexities of tracking user engagement across different platforms. The document emphasizes the importance of adapting attribution models to align with marketing strategies and outlines future steps for improving attribution practices.
A Multi-Touch Attribution In Mobile Retrospective With Trendyol
- 1. Multi Touch Attribution in Mobile Adjust – Mobile Spree 2018
- 2. Agenda - Multi Touch Attribution • Some Background • Background on Multi Touch Attribution for Trendyol • Concept of 'Marketing Value’ • Results of MTA • Insights on MTA Results & Process • What should you expect to see? • What are the actions you can take? 2Mobile Multi Touch Attribution - Trendyol
- 3. Trendyol • Largest e-commerce company in Turkey by orders, userbase • Top 5 largest fashion e-commerce in Europe • 100% growth in its 8th and 9th year (so far), profitable for 4 years, • 2000 people team, has its own fashion collection for women, men, kids, home. • Covering 80% of Turkish female e-commerce shoppers in Turkey, • 90% GMV from repeat customers • ~75% of sales in mobile 3Mobile Multi Touch Attribution - Trendyol
- 4. Myself • In Trendyol for the last 1,5 years, leading a team of 50+ people of Analytics, CRM, Digital Marketing, Creative, Marcomm, Data Science • Has worked for Peak Games for 5+ years, responsible for 100M+ USD yearly marketing budget • Responsible for the growth of Trendyol, Dolap 4Mobile Multi Touch Attribution - Trendyol
- 5. From ROAS => Marketing Value • Problem: Comparing channels properly with respective to their short- mid-long term gains (CRM, Acquisitional, Transactional Mktg, SM...) • Problem: A Marketing source can bring revenue, new buyer, download, initiates a strategic objective. What is the real value? • Problem: Marketing objectives and priorities can change in time • Problem: Tracking capabilities differ in Web & Mobile, Are we treating mobile as we should? 5Mobile Multi Touch Attribution - Trendyol
- 6. From ROAS => Marketing Value • Old: Looking at ROAS only for a fixed attribution model in web & mweb, last click model in mobile • New: Assigning a value for attributed New Buyer, Download, Profit creating a ‘Virtual ROI’ => Marketing Value • New: Multi Touch Attributed New Buyer in Web & Mweb • New: Multi Touch Attributed New Buyer in Mobile • New: Multi Touch Attributed Revenue in Mobile 6Mobile Multi Touch Attribution - Trendyol
- 7. Marketing Value Challenges & Advantages Challenges • Finding the ’incremental value’ of new buyer, download • Mobile Multi Touch Attribution • Finding the right attribution coefficients & interval • Values may not be 100% correct. • Cross Platform and branding channels problem: Limited visibility on all the value • Soft Rules on MTA 7Mobile Multi Touch Attribution - Trendyol Advantages • Comparable channels and sources • Marketing aligned with strategic objectives • Marketing aligned with current priorities • More visibility on customer journey
- 8. Current Status on Attribution & Challenges Current Status • Different attribution models in different platforms • Problems in analyses on marketing efficiency • No neutral provider with MTA capabilities in mobile Process Challenges • Lack of data and tool availability, infrastructure • Lack of industry know-how and best practices in mobile • Source based problems in attribution funnel 8Mobile Multi Touch Attribution - Trendyol
- 9. Results (Mobile Engagement) # of attributed transactions Lookback window %difference with respect to last touch Network Last touch 7 days 15 days 30 days 7 days 15 days 30 days Push 292,811 300,100 297,866 293,926 2% 2% 0% Network A 84,994 84,817 85,169 86,137 0% 0% 1% Organic 78,865 83,168 86,970 92,320 5% 10% 17% Network B 78,260 85,624 88,135 91,947 9% 13% 17% Network C 67,354 60,950 54,052 47,572 -10% -20% -29% Network D 26,951 26,958 27,145 27,936 0% 1% 4% Network E 23,518 29,488 34,775 40,291 25% 48% 71% 9Mobile Multi Touch Attribution - Trendyol • Increase in organic helps us understand the real dynamics of channels, looking at ROAS conservatively. • Some channels stand still, they contribute all funnel evenly, some are penalized.
- 10. Results (New Buyer Acquisition) # of attributed new buyer Lookback window %difference with respect to last touch Network last touch 7 days 15 days 30 days 7 days 15 days 30 days Push 17,419 20,382 20,119 19,683 17% 16% 13% Organic 11,393 11,752 12,202 12,988 3% 7% 14% Network A 10,890 11,214 11,163 11,139 3% 3% 2% Network B 13,622 14,030 13,077 11,987 3% -4% -12% Network C 3,205 3,478 3,669 3,731 9% 14% 16% Network D 1,613 1,676 1,668 1,724 4% 3% 7% Network E 1,010 1,445 1,718 1,946 43% 70% 93% 10Mobile Multi Touch Attribution - Trendyol • Push campaigns persuade new downloads into a purchase • Some networks have a massive effect in new buyer acquisition • There are many organic sessions in new buyer persuasion process.
- 11. Insights on Results and the Process • Some sources are tricking «the last click attribution» • Some sources have good invisible effects on user journey: • Initiating a relation with a ‘product’ • Upper Funnel & Mid Funnel & Purchase Assists • New Buyer conversion • Not all retargeting is same, most retargeting on mobile clearly focuses on lower funnel => Risk of organic cannibalization is real • Male and female journeys are different both in time intervals, and in number of assists • iOS and Android users do not differ much, channels and ad units are the main differentiators 11Mobile Multi Touch Attribution - Trendyol
- 12. Insights on Results and the Process • Mweb journeys are difficult to track • Push tracking dilutes & disrupts the digital marketing tracking • Attribution time frame changes from a category to another • Attribution model will probably change from a category to another • We never see real organic and direct traffic in mobile from our current infrastructure – Organic traffic is underrated • Decomposing the sticky attribution of current last click attribution model • Needs more resources and is more difficult than it sounds :) 12Mobile Multi Touch Attribution - Trendyol
- 13. Next Steps • Including ‘Uninstalls’ in the journey • Assigning a marketing value for strategic objectives, new visits • Testing out attribution model coefficients • Ad Unit effect on attribution: Video vs. Rewarded Video vs. Static • Trying out different attribution window • Different attribution models and windows in different categories • Multi Touch attribution in downloads (?) • Cross Platform Attribution (?) 13Mobile Multi Touch Attribution - Trendyol
- 14. Why not download multi touch attribution? • Previous Gaming experience: Do we really need an MTA for download? • Majority of Downloads in a few steps, in a very short short time interval (Purchasing decision happens in 20+ steps, in 3-15 days) • Technical difficulties even bigger than engagement 14Mobile Multi Touch Attribution - Trendyol
- 15. You can survive without MTA when... • Your marketing is • All Download • All Acquisitional • All Transactional • Majority of spent is on a single provider • Your company has: • Small Userbase • Similar Maturity level across the userbase • Don’t have various strategic objectives for digital marketing • No spare resource for analytics & data engineering 15Mobile Multi Touch Attribution - Trendyol
- 16. Summary & Q&A 16Mobile Multi Touch Attribution - Trendyol • There is no one «true» model for attribution • No clear downside for doing MTA as long as you have resources • 80% Accuracy in MTA > Just Doing MTA >> Not Doing MTA • MTA should incorporate your marketing strategy • MTA coefficients and soft rules can change in time and in context
- 17. Thanks! Mobile Multi Touch Attribution