Making Programmatic Work
Download as PPTX, PDF6 likes1,285 views
This document discusses how to achieve marketing goals through programmatic advertising using data and algorithms. It emphasizes setting the right advertising goal, understanding the large programmatic opportunity in terms of available digital impressions, and managing the vast amount of data through predictive models and optimization to determine the right bid price for each impression and which impressions to purchase. The key to success is creating a strategic vision, setting a clear goal, accessing maximum supply, onboarding relevant data, using world-class algorithms, and providing insightful analysis.
Making Programmatic Work
- 1. Making Programmatic Work Achieving marketing goals through v data and algorithms Frith Fraser, Director, Platform Solutions, APAC
- 2. What is your advertising goal?
- 3. Optimizing to Goals, not Guesses ©2014 MEDIAMATH INC. 3
- 4. What is your true advertising goal? • The right goal for the right product • Return on marketing investment? • Offline attribution/custom attribution • Branding engagement metrics • Understanding audiences • Look beyond the click • Does CPC answer your marketing goal?
- 5. What does the programmatic opportunity look like?
- 6. 2.5 trillion impressions are available on a monthly basis globally In the UK, 107 billion monthly impressions available In the US, 900 billion monthly impressions available In Japan alone, 93 billion impressions and video units run through MediaMath’s TerminalOne System monthly In Australia, 40 billion monthly impressions available In Brazil, 115 billion monthly impressions available
- 7. Canada 160 million data points USA 13 billion data points More than 21 billion global data points with over 350 data partners UK 61 million data points France 45 million data points Germany 42 million data points Brazil 160 million data points China 16 million data points Japan 10 million data points India 50 million data points Australia 24 million data points
- 8. Seconds in the average human lifespan? Galaxies in the observable universe? ~ 150B Ants on planet Earth? ~ 2.5B ~ 1Q Quiz time
- 9. Let’s talk about data in RTB • 50 billion impressions per day • 200 variables per impression • 100 values per variable = 1,000,000,000,000,000 pieces of data per day (1015 = ONE QUADRILLION)
- 10. That’s a whole lot of data. How do we manage this?
- 12. The two buyer questions that matter: ©2014 MEDIAMATH INC. 12 Question 1: What is the right bid for each impression? Question 2: Which impressions should I buy?
- 13. What price should I pay? What is the right bid for each impression? Too high Overpay & underperform Too low Lose out & underspend Just right Maximise scale & performance
- 14. Question 1: A simple exercise What is the right bid for each impression? JPY1000 prize JPY500 Flip a coin to win the prize 50% chance Bid Price Goal Value x Action Rate = Bid Price
- 15. Question 1: The real thing What is the right bid for each impression? Goal Value x Action Rate = Bid Price 1% chance consumer takes desired action (purchase) Bid for an RTB ad ¥5000 value to advertiser (CPA) ¥50 bid price (breakeven) Your ad here
- 16. Answering Question 1: Video Publisher: YouTube Unit: 15 sec pre roll Time: 16.46 – 17.00 Age: 25-34 Gender: Male Price: JPY1500 A different model for every creative in every campaign of every advertiser – all in real time! Social Publisher: FBX Unit: Newsfeed Day: Tuesday Time: 5.00pm – 5.15pm Price: JPY300 Display Publisher: Rubicon Data: Rakuten Male Location: Tokyo Creative size: 160 x 600 Price: JPY300
- 17. Question 2: Which impressions should I buy? • ~100MM/day of RTB supply • Typical campaign spends ~$1K/day (0.001% of total supply) • Not buying the RIGHT 0.001% is throwing money away
- 18. Answering Question 2: Defining value Bid Price Market Price What it’s worth to me What I have to pay Relative = = Value • Cheap impressions are not necessarily good • Good impressions are not necessarily cheap
- 19. The benefits Buy in batch Buy what you want Fixed price, regardless of value Variable bidding, aligned with value Little/no insight into true drivers Full insights into “what” & “why” Manual, labor-intensive (~50/FTE) Fully automated, scalable (~5/FTE) Analyze 10-20 buys weekly Analyze 1MM opps. per sec. And the results are typically 10x better
- 20. Key learnings • Create a strategic programmatic vision • Set a true goal • Ensure access to maximum amount of supply • Ensure the ability to onboard data sets
- 21. Key learnings • Utilise a world-class algorithm • Use a predictive model to determine what each impression is worth • Use optimisation to determine which impressions to bid for • Provide meaningful and relevant analysis to your client • Build a knowledgeable and capable team
- 22. THANK YOU ffraser@mediamath.com facebook.com/mediamath @mediamathAPAC
Editor's Notes
- #8: Rakuten Data: to hit