Testing Into The Future - SEM's New Best Practices By Aaron Levy

50 %
50 %
Information about Testing Into The Future - SEM's New Best Practices By Aaron Levy

Published on October 25, 2018

Author: SearchMarketingExpo

Source: slideshare.net

1. #SMX #21B @bigalittlea TESTING INTO THE FUTURE- SEM’S NEW BEST PRACTICES

2. #SMX #21B @bigalittlea HI! I’M AARON! 12 years of digital-y stuff SEM Director at Elite SEM I teach at Drexel & UVM Hot sauce maker Ordained minister @bigalittlea | aaron@elitesem.com

3. #SMX #21B @bigalittlea Ad Age Best Agency Culture Ad Age Top 25 US Search Agencies US Search Awards #1 Best Agency Bing Independent Agency, Philanthropist of the Year Mashable #1 Best Tech Companies to Work For Entrepreneur #1 Top Company Culture Fortune Best Workplace for Millenials Crain’s NYC #1 Best Place to Work Award-Winning Digital Marketing Agency

4. #SMX #21B @bigalittlea ELITESEM.COM CONFIDENTIAL | 4 Segmented Keyword Pools top performers (exact) top performers (bmm) dsa + tofu tests average (exact & bmm)

5. #SMX #21B @bigalittlea Brand Non Brand

6. #SMX #21B @bigalittlea Want Your Stuff, May Already Have It Want Stuff You Sell Want Stuff From You OOOH, STUFF! Want Your Stuff From You Want Stuff, Not Sure What Flavor Want Stuff, Maybe Yours ¯_( )_/¯

7. #SMX #21B @bigalittlea#SMX #21B @bigalittlea Keywords Are One Lever An Old, Dated Lever

8. #SMX #21B @bigalittlea Have Your Stuff|E Want Stuff You Sell|E Want Your Stuff |E Taste Test|E Want Stuff From You|E OOOH, STUFF!|E Want Stuff, Maybe Yours|E ¯_( )_/¯ |E Have Your Stuff | BMM Want Stuff You Sell|BM Want Your Stuff |BMM Taste Test | BMM Want Stuff From You|BM OOOH, STUFF!|BM Want Stuff, Maybe Yours|B ¯_( )_/¯ |BMM Have Your Stuff|E Want Stuff You Sell|E Want Your Stuff |E Taste Test|E Want Stuff From You|E OOOH, STUFF!|E Want Stuff, Maybe Yours|E ¯_( )_/¯ |E Have Your Stuff | BMM Want Stuff You Sell|BM Want Your Stuff |BMM Taste Test | BMM Want Stuff From You|BM OOOH, STUFF!|BM Want Stuff, Maybe Yours|B ¯_( )_/¯ |BMM NEW TO SITE SITE VISITORS

9. #SMX #21B @bigalittlea

10. #SMX #21B @bigalittlea

11. #SMX #21B @bigalittlea ♀♂NEWDEEPBOUNCEDUMPNEWDEEPBOUNCEDUMP MORE MOBILE LESS MOBILE

12. #SMX #21B @bigalittlea CUSTBOUNCENEWDEEP MORE MOBILEY LESS MOBILEY CUSTBOUNCENEWDEEP MORE MOBILEY LESS MOBILEYYOUNG OLD

13. #SMX #21B @bigalittlea 2,048+ targeting signals

14. #SMX #21B @bigalittlea We Don’t Have The Same Tools Query-Level Bidding RTB Operating System Browser Recency + Frequency Search Partners Price Competitiveness & Seasonality*

15. #SMX #21B @bigalittlea#SMX #21B @bigalittlea Test Into A Smarter Future

16. #SMX #21B @bigalittlea Three Types Of AI-Driven Automation Bidding Targeting Everything

17. #SMX #21B @bigalittlea Three Types Of AI-Driven Automation Bidding Targeting Everything

18. #SMX #21B @bigalittlea AI-Driven Bid Automation Channel Goal tCPA T D Maximize conversions within target cost per acquisition (CPA) tROAS T S D Maximize revenue at target ROAS eCPC T S D TD: Maximize conversions at current CPA Maintain avg. CPC <= max. CPC S: Maximize number of conversions at current spend

19. #SMX #21B @bigalittlea AI-Driven Bid Automation Channel Goal tCPA T D Spend More tROAS T S D Spend More eCPC T S D TD: Spend More Spend More S: Improve ROAS #SMX #21B @bigalittlea

20. #SMX #21B @bigalittlea Culture of Testing: NNN Budgets Room for failure encourages experimentation 70% 20% 10% ProvenTactics EasyTracking “GottaBe There” OldStuff NewStyle Innovation

21. #SMX #21B @bigalittlea CPA Target: $100 Net Revenue Per Conversion: $150 Profit Per Conversion: $50 30 Conversions Per Day $3k Spend | $1.5k Profit Cost Of Testing – Cosmo’s Tours

22. #SMX #21B @bigalittlea Calculating Payback Periods Spend CPA Sales Revenue Profit Status Quo $42,000 $100 420 $63,000 $21,000 Learn Period $42,000 $115 365 $54,750 $12,750 Cost ofTesting ($8,250) Let’s assume a 14 day learning period 15% “tuition” as the tool learns

23. #SMX #21B @bigalittlea The test works! Cosmo sells 15% more tours at the same CPA Calculating Payback Periods Spend CPA Sales Revenue Profit Satus Quo $3,000 $100 30 $4,500 $1,500 Post-Learning $3,600 $100 36 $5,300 $1,800 Improvement + $600 - +6 + $675 + $225 Payback Period ROI Over aYear 37 days 10:1

24. #SMX #21B @bigalittlea Tolerance Forecasts Mitigate Risk ($40,000) ($20,000) $0 $20,000 $40,000 $60,000 $0 $40,000 $80,000 $120,000 $160,000 $200,000 NoTest Best Good Expected Bad Worst Didn't Work Test Period (Days) - - 7 14 14 21 21 Test CPA “Tuition” - - 10% 20% 30% 40% 40% Post Spend Lift - 50% 35% 20% 10% 5% -

25. #SMX #21B @bigalittlea Evaluating Payback Periods NoTest Best Good Expected Bad Worst Didn't Work Test Spend (daily) $3k $3k $3k $3k $3k $3k $3k Test Period Profit $1,500 $1,500 $1,091 $750 $462 $214 $214 COT (duration) $0 $0 ($5.7k) ($10.5k) ($14.5k) ($18k) ($27k) PostTest Spend $3k $4.5k $4k $3.6k $3.3k $3.1k $3k PostTest Profit $1.5k $2.3k $2k $1.8k $1.7k $1.6k $1.5k Break Even (days) - 0 11 35 97 240 Never Ever 12 Week Test Rev $0 $52.5k $37.6k $10.5k ($4k) ($22k) ($27k) (Expected Profit – Test Profit) * Test Period = Cost of Testing COT / (Post Test Profit – Expected Profit) = Payback

26. #SMX #21B @bigalittlea What Do Robots Care About?? Conversion Rate AOV That’s it.

27. #SMX #21B @bigalittlea Ignored • Bid Modifiers • Invisible Audiences Respected • +/- 100% Modifiers • MatchTypes (sorta) • Budgets (also sorta) • Targeting criteria #SMX #21B @bigalittlea

28. #SMX #21B @bigalittlea Set Up The Robots For Success Audiences: Add any/all audiences as observe, including in-market Recency is accounted for, no need to segment remarketing audiences by time Err big, not small – don’t make the robot guess Keywords: Err on the side of broad, not precise Eliminate outliers You can still smart-ify them, but in their own portfolio Eliminate crap

29. #SMX #21B @bigalittlea Recommended Setup Criteria tROAS tCPA eCPC Bids Average ROAS over last 30 days Average CPA over last 30 days Average CPC will stay below Max CPC Conversion Reqs Minimum Portfolio: 50+ Campaign: 15+ None! None! Google Recommends Portfolio: 150+ Campaign: 30+ Portfolio: 50+ Campaign: 15+ 15+ over Aaron Recommends Portfolio: 300+ Campaign: 60+ Portfolio: 150+ Campaign: 30+ 15+ over Other Notes Be patient! Keep changes to every two weeks and 10% thresholds Will take care of all mods except mobile *all recommendations over 30 day period

30. #SMX #21B @bigalittlea Properly Test With Drafts + Experiments Compress Campaigns There, I said it. Smush your middling performers together into a super campaign to maximize data. Minimize Funneling Segment match types at ad group level, not campaign. Ensure Data Parity Clone campaigns 30 days ahead of time. If you can’t wait, use new campaigns as “control.” Give It Room No limited by budget. Plenty of ads. No modifiers. Tons of audiences. Device splits only if you must.

31. #SMX #21B @bigalittlea …Does it work??? You gotta test J

32. #SMX #21B @bigalittlea Share these #SMXInsights on your social channels! #SMXInsights § Factor the cost of testing as you explore automation – what’s the cost of a loss or return on a win? § Modern tests won’t work on legacy structures. § S/he with the best data wins. Often, that does not mean you.

33. #SMX #21B @bigalittlea aaron@elitesem.com #SMX #21B @bigalittlea THANK YOU! MAY YOUR JOBS BE EASY AND YOUR REWARDS PLENTIFUL #SMX

34. #SMX #21B @bigalittlea source: enterprise marketplace Does eCPC Do… Anything? Short answer: Yes! It does what it’s supposed to with volume Clicks Cost CPC Orders Revenue ROAS AOV CVR Control 80.8k $29k $0.36 1,076 $68k 235% $63 1.3% eCPC 75k $30k $0.40 1,151 $74.4k 246% $65 1.5% Manual+eCPC Δ -7% 4% 12% 7% 9% 5% 2% 15%

35. #SMX #21B @bigalittlea source: enterprise marketplace tCPA vs. tROAS vs. hooman Cost CPC Conv. CPA Revenue AOV ROAS Manual $39k $0.39 56k $7 $90k $16 2.3 tCPA $42k $0.71 13k $3 $200k $15 4.7 tCPA Δ 9% 84% 136% -54% 122% -6% 104% Manual $43k $0.30 6.5k $7 $87k $13 2.0 tROAS $48k $1.17 3.3k $14 $47k $14 1.0 tROAS Δ 11% 283% -49% 119% -46% 6% -51% less variables + stability wins. note: test is old, tROAS better since

36. #SMX #21B @bigalittlea The Whole Shebang: Smart Display Campaigns

37. #SMX #21B @bigalittlea $100 goal CPA | $80 tCPA | $1,000/day budget | 30 day test Smart Display Test 1: Meal Kits Learned that our goals were too aggressive + optimized itself off 0 2 4 6 8 10 12 14 16 $0 $500 $1,000 $1,500 $2,000 Cost Conversions

38. #SMX #21B @bigalittlea $350 break even CPA | $300 tCPA | $350/day budget | 30 day test Smart Display Test 2: Luxury Hotels Google knows who likes luxury travel, buuuut they don’t know who has plans booked or needs a hotel. Whoops. 0 1 $0 $200 $400 $600 $800 Cost Bookings

39. #SMX #21B @bigalittlea Dropped target to $100 $300 break even CPA | $200 tCPA | $10k/day budget (starting) Smart Display Test 3: Subscription App Google had plenty of data to work with plus a commodity product. $2.7mm invested over the course of year at a >2:1 ROI (allll profit). 0 200 400 600 800 $0 $40,000 $80,000 $120,000 $160,000 $200,000 Cost Conversions

40. #SMX #21B @bigalittlea aaron@elitesem.com #SMX #21B @bigalittlea NOW WE’RE FOR REAL DONE COME TO THE NEXT #SMX!

#smx presentations

Add a comment