Optimizing customer segments at scale
•
0 likes•152 views
This document summarizes a presentation on optimizing customer segments at scale. The presentation covered five key lessons: 1) The importance of granular tracking to properly analyze performance; 2) Leveraging keywords and search logs to refine campaigns; 3) Testing a wide range of interest segments to identify high performers; 4) Tailoring ad messaging to customers and scaling dynamic text; 5) Developing a structured campaign framework for improved targeting, scaling, and optimization. The presenter provided examples and case studies demonstrating how to apply these lessons to increase profitability through optimized segmentation, keywords, interests, messaging and structure.
Report
Share
Recommended for you
Recommended for you
Recommended for you
Recommended for you
More Related Content
What's hot
What's hot (20)
Similar to Optimizing customer segments at scale
Similar to Optimizing customer segments at scale (20)
Optimizing customer segments at scale
- 1. Upside / 2013 1 AGILE MARKETING MEETUP MAY 2013 OPTIMIZING CUSTOMER SEGMENTS AT SCALE (AND SOME OTHER NEAT HACKS) JUNEY HAM CO-FOUNDER & PRESIDENT UPSIDE @juney @gainupside 1
- 2. Upside / 2013 A little bit about me 2 2
- 3. Upside / 2013 New company 3 3
- 5. Upside / 2013 5 My lessons learned in five buckets 1. Tracking 2. Keywords 3. Interests 4. Messaging 5. Structure 5
- 6. Upside / 2013 Proper tracking = proper optimization 6 NOT THIS AD GROUP-LEVEL TRACKING • Cannot breakdown which keyword in an ad group resulted in conversion • As keywords and ads change, ad group-level tracking makes tracking murky and imprecise Set up granular tracking (e.g., keyword + match type + ad) to minimize tracking issues. Create an internally-generated ID to uniquely track every atomic unit (use this to join against other data sets). KEYWORD-LEVEL TRACKING • Match type mismatch errors (e.g., joining broad match cost data with exact match revenue data) 6
- 7. Upside / 2013 Proper tracking = proper optimization 7 DO THIS KEYWORD MATCH TYPE AD NETWORK San Francisco Hotels Broad Best Hotel Prices Guaranteed. Save on Hotels in San Francisco! Search San Francisco Hotels Phrase Best Hotel Prices Guaranteed. Save on Hotels in San Francisco! Search San Francisco Hotels Exact Best Hotel Prices Guaranteed. Save on Hotels in San Francisco! Search San Francisco Hotels Broad Best Hotel Prices Guaranteed. Save on Hotels in San Francisco! Content Sample URL parameters ?utm_source=google&utm_medium=paid&utm_term={keyword}&utm_content={creative} &utm_campaign=<ADGROUPNAME>&Network={ifContent:Content}{ifSearch:Search} &SiteTarget={placement}&id=1234567890 ?utm_source=google&utm_medium=paid&utm_term={keyword}&utm_content={creative} &utm_campaign=<ADGROUPNAME>&Network={ifContent:Content}{ifSearch:Search} &SiteTarget={placement}&id=1234567890 ?utm_source=google&utm_medium=paid&utm_term={keyword}&utm_content={creative} &utm_campaign=<ADGROUPNAME>&Network={ifContent:Content}{ifSearch:Search} &SiteTarget={placement}&id=1234567890 Set up granular tracking (e.g., keyword + match type + ad) to minimize tracking issues. Create an internally-generated ID to uniquely track every atomic unit (use this to join against other data sets). revenue data (id=1234567890) cost data (id=1234567890) 7
- 8. Upside / 2013 8 Keywords are little units of intent and can be used effectively at scale SEARCH LOGS SEARCH QUERIES PARSE SEARCH LOGS TO CREATE NEW KEYWORD OPPORTUNITIES • Any site log (Apache, Rails logger, etc.) or analytics system (GA, SiteCatalyst) can be parsed to pull out a wealth of information, including: • pages visited • channels used • keywords searched • Test quality of keywords pulled from logs and build rules to filter out “bad seeds” • Search substrings • Gibberish • Blacklisted words (profanity, etc.) • If it makes sense, productize this process: • Pull all logs and do some basic scrubbing • Run terms through business rules & filters • Remove duplicates and add to correct ad group • Initial bid value assignment • Manual review • Launch! 8
- 9. Upside / 2013 9 Keywords are little units of intent and can be used effectively at scale LEVERAGE SEARCH QUERIES TO REFINE CAMPAIGNS For the ad group “Edinburgh Hotels”: 1. Add keyword negatives for irrelevant or poorly- converting phrases 2. Add new exact match keywords 3. Move queries to another campaign or ad group for better ad text targeting edinburgh hotel edinburgh cheap hotel edinburgh hotels discount edinburgh hotel jobs edinburgh scotland hotel edinburgh best hotels travel guide edinburgh hotels edinburgh hotel george SEARCH LOGS SEARCH QUERIES 9
- 10. Upside / 2013 Interests (and demographics) 10 POTENTIAL INTERESTS EXAMPLE INTEREST CLOUD Interests are unique to brand and customer so ... 1. Cast a wide net (test hundreds or thousands of potential interests, both broad and precise) 2. See what works for you 3. Optimize out the remaining non- performers 10
- 12. Upside / 2013 Messaging: don’t forget to put your “customer hat” on 12 vs. Which ad copy performed better? 12
- 13. Upside / 2013 Messaging: don’t forget to put your “customer hat” on 13 vs. Which ad copy performed better? 13
- 14. Upside / 2013 Messaging: don’t forget to put your “customer hat” on 14 vs. Which ad copy performed better? CTR 2%* CVR 3.2%* CPA $100* CTR 4%* CVR 4.5%* CPA $80* Beware of “too good to be true” messaging (it’s not always a race to the bottom) Price point likely resulted in different population of customers clicking on the ad, with different behaviors * illustrative example 14
- 15. Upside / 2013 Messaging: Scaling dynamic ad text through Ad Parameters 15 15
- 16. Upside / 2013 Messaging: Scaling dynamic ad text through Ad Parameters 16 16
- 17. Upside / 2013 Messaging: Scaling dynamic ad text through Ad Parameters 17 DYNAMIC COMPONENTS “FROM PRICE” (from $60/night!) NUMBER OF HOTELS (500 hotels in Los Angeles!) PERCENTAGE OFF (up to 50% off!) Problem: how to create relevant, targeted ad copy with the correct values, at scale? Solution: Google Ad Parameters! Usage: 1. Create new ads inserted with {param1} and {param2} Example: {param1:500} Hotels in Sydney Find Hotel Deals from {param2:$60}/Night! 2. Add param1 and param2 fields to each keyword, which triggers the proper parameter when the ad is fired 3. Optimize, try new variations, rinse, repeat! 17
- 18. Upside / 2013 The importance of structure: scale, tracking, messaging, targeting 18 Case study: • Started process of campaign structure overhaul • Used various data sources and analyses to determine to most effectively create targeted, granular, consistent structure • Created worldwide list of top destinations by language (English, Spanish, French, German, Dutch, Danish, Japanese, etc.) • Categorized destinations into tiers: • Tier 1: Major cities worldwide (New York City, London, Paris, etc.) • Tier 2: Regional + seasonal destinations (Edinburgh, Philadelphia, Phuket, Vail, etc.) • Tier 3: Long-tail cities 70K destinations * 500 patterns * 100 qualifiers = ~3.5B potential keywords ~3.5B keywords = 100+ accounts & righteous ire from Google account team 18
- 19. Upside / 2013 19 Optimization process: • Ran keywords through both internal vetting process AND Google Traffic Estimator to weed out low-relevance and low-volume queries • Used analysis of regional travel patterns and customer behavior to determine that travelers from different countries had significantly different travel behaviors, some that were specific only to that subset • US travel to Paris • UK travel to Nice, Provence • FR travel to Nantes, Deauville • Created flexible framework to handle differences in keyword density by origination • US English keywords for Miami (base): ~50 • UK English keywords for Miami (base): ~15 • SG English keywords for Miami (base): ~5 Result: Total keyword set reduced to about 200K and only four accounts per language Performance: Increased profitability by +300% The importance of structure: scale, tracking, messaging, targeting 19
- 20. Upside / 2013 20 Recap: Know your customer, track them, create hypotheses, test, repeat. TRACKING • Granular tracking is important to minimize reporting errors • Create a unique ID based on the combination of parameters you want to track (keyword + match type + ad + placement) KEYWORDS • Leverage site (search) logs to mine quality keywords • Search query reports are awesome for optimizing keywords • Productize processes to scale keyword nomination & creation INTERESTS • Zoom in then zoom out to find effective segments, then refine • Create custom interests using a combination of topics and precise interests • Applicable across channels, assets, etc. MESSAGING • Test things even if they are counter-intuitive • Scale dynamic messaging through Google Ad Parameters STRUCTURE • Sometimes, less is more • Drill down, then drill sideways to get more granularity • Create opportunities for flexibility to accommodate changes 20