ANZ Analytics
1 like826 views
This document provides an overview of Datalicious, an analytics company that offers data services including data platforms, insights reporting, and data applications to help clients drive smart data-driven marketing. It discusses Datalicious' history and clients, describes their wide range of data services, and covers topics like metrics frameworks, marketing funnels, and the importance of using multiple data sources and accounting for calendar events in data analysis.
ANZ Analytics
- 1. > ANZ Analy*cs Workshop < Smart Data Driven Marke.ng
- 2. > Short but sharp history § Datalicious was founded late 2007 § Strong Omniture web analy.cs history § Now 360 data agency with specialist team § Combina.on of analysts and developers § Carefully selected best of breed partners § Evangelizing smart data driven marke.ng § Making data accessible and ac.onable § Driving industry best prac.ce (ADMA) March 2011 © Datalicious Pty Ltd 2
- 3. > Clients across all industries March 2011 © Datalicious Pty Ltd 3
- 4. > Wide range of data services Data Insights Ac*on PlaAorms Repor*ng Applica*ons Data collec*on and processing Data mining and modelling Data usage and applica*on Web analy*cs solu*ons Customised dashboards Marke*ng automa*on Omniture, Google Analy*cs, etc Media aNribu*on models Alterian, Trac*on, Inxmail, etc Tag-‐less online data capture Market and compe*tor trends Targe*ng and merchandising End-‐to-‐end data plaAorms Social media monitoring Internal search op*misa*on IVR and call center repor*ng Online surveys and polls CRM strategy and execu*on Single customer view Customer profiling Tes*ng programs March 2011 © Datalicious Pty Ltd 4
- 5. > Smart data driven marke*ng Metrics Framework Metrics Framework Media ANribu*on Benchmarking and trending Benchmarking and trending Op*mise channel mix Targe*ng Increase relevance Tes*ng Improve usability $$$ March 2011 © Datalicious Pty Ltd 5
- 6. 101011010010010010101111010010010101010100001011111001010101 010100101011001100010100101001101101001101001010100111001010 010010101001001010010100100101001111101010100101001001001010 > Metrics framework March 2011 © Datalicious Pty Ltd 6
- 7. > AIDA and AIDAS formulas Old media New media Awareness Interest Desire Ac*on Sa*sfac*on Social media March 2011 © Datalicious Pty Ltd 7
- 8. > Simplified AIDAS funnel Reach Engagement Conversion +Buzz (Awareness) (Interest & Desire) (Ac.on) (Sa.sfac.on) March 2011 © Datalicious Pty Ltd 8
- 9. > Marke*ng is about people People People People People reached 40% engaged 10% converted 1% delighted March 2011 © Datalicious Pty Ltd 9
- 10. > Unique visitor overes*ma*on The study examined data from two of the UK’s busiest ecommerce websites, ASDA and William Hill. Given that more than half of all page impressions on these sites are from logged-‐in users, they provided a robust sample to compare IP-‐based and cookie-‐based analysis against. The results were staggering, for example an IP-‐based approach overes.mated visitors by up to 7.6 .mes whilst a cookie-‐based approach overes*mated visitors by up to 2.3 *mes. March 2011 © Datalicious Pty Ltd 10 Source: White Paper, RedEye, 2007
- 11. > Maximise iden*fica*on points 160% 140% 120% 100% 80% 60% −−− Probability of iden.fica.on through Cookies 40% 20% 0 4 8 12 16 20 24 28 32 36 40 44 48 Weeks March 2011 © Datalicious Pty Ltd 11
- 12. > Maximise iden*fica*on points Mobile Home Work Online Phone Branch March 2011 © Datalicious Pty Ltd 12
- 13. > Addi*onal funnel breakdowns Brand vs. direct response campaign People People People People reached 40% engaged 10% converted 1% delighted New prospects vs. exis.ng customers March 2011 © Datalicious Pty Ltd 13
- 14. New vs. returning visitors
- 15. AU/NZ vs. rest of world
- 17. > Exercise: Funnel breakdowns § List poten.ally insighcul funnel breakdowns – Brand vs. direct response campaign – New prospects vs. exis.ng customers – Baseline vs. incremental conversions – Compe..ve ac.vity, i.e. none, a lot, etc – Segments, i.e. age, loca.on, influence, etc – Channels, i.e. search, display, social, etc – Campaigns, i.e. this/last week, month, year, etc – Products and brands, i.e. iphone, htc, etc – Offers, i.e. free minutes, free handset, etc – Devices, i.e. home, office, mobile, tablet, etc March 2011 © Datalicious Pty Ltd 17
- 18. > Mul*ple metrics data sources Media and search data Website, call center and retail data People People People People reached engaged converted delighted Quan.ta.ve and qualita.ve research data Social media data Social media March 2011 © Datalicious Pty Ltd 18
- 19. > Importance of calendar events Traffic spikes or other data anomalies without context are very hard to interpret and can render data useless March 2011 © Datalicious Pty Ltd 19
- 20. Calendar events to add context March 2011 © Datalicious Pty Ltd 20
- 21. > Conversion funnel 1.0 Campaign responses Conversion funnel Product page, add to shopping cart, view shopping cart, cart checkout, payment details, shipping informa.on, order confirma.on, etc Conversion event March 2011 © Datalicious Pty Ltd 21
- 22. > Conversion funnel 2.0 Campaign responses (inbound spokes) Offline campaigns, banner ads, email marke.ng, referrals, organic search, paid search, internal promo.ons, etc Landing page (hub) Success events (outbound spokes) Bounce rate, add to cart, cart checkout, confirmed order, call back request, registra.on, product comparison, product review, forward to friend, etc March 2011 © Datalicious Pty Ltd 22
- 23. > Addi*onal success metrics Click Through $ Click Add To Cart Through Cart Checkout ? $ Click Page Page Product Through Bounce Views Views $ Click Call back Store Through request Search ? $ March 2011 © Datalicious Pty Ltd 23
- 24. Exercise: Sta*s*cal significance March 2011 © Datalicious Pty Ltd 24
- 25. How many survey responses do you need if you have 10,000 customers? How many email opens do you need to test 2 subject lines if your subscriber base is 50,000? How many orders do you need to test 6 banner execu*ons if you serve 1,000,000 banners March 2011 © Datalicious Pty Ltd 25 Google “nss sample size calculator”
- 26. How many survey responses do you need if you have 10,000 customers? 369 for each ques*on or 369 complete responses How many email opens do you need to test 2 subject lines if your subscriber base is 50,000? And email sends? 381 per subject line or 381 x 2 = 762 email opens How many orders do you need to test 6 banner execu*ons if you serve 1,000,000 banners? 383 sales per banner execu*on or 383 x 6 = 2,298 sales March 2011 © Datalicious Pty Ltd 26 Google “nss sample size calculator”
- 27. > Addi*onal success metrics Click Through $ Click Add To Cart Through Cart Checkout ? $ Click Page Page Product Through Bounce Views Views $ Click Call back Store Through request Search ? $ March 2011 © Datalicious Pty Ltd 27
- 29. > Exercise: Metrics framework Level Reach Engagement Conversion +Buzz Level 1, people Level 2, strategic Level 3, tac*cal Funnel breakdowns March 2011 © Datalicious Pty Ltd 29
- 30. > Exercise: Metrics framework Level Reach Engagement Conversion +Buzz Level 1 People People People People People reached engaged converted delighted Level 2 Display Strategic impressions ? ? ? Level 3 Interac*on Tac*cal rate, etc ? ? ? Funnel Exis*ng customers vs. new prospects, products, etc Breakdowns March 2011 © Datalicious Pty Ltd 30
- 31. > Exercise: Conversion Funnel March 2011 © Datalicious Pty Ltd 31
- 32. 101011010010010010101111010010010101010100001011111001010101 010100101011001100010100101001101101001101001010100111001010 010010101001001010010100100101001111101010100101001001001010 > Media aNribu*on March 2011 © Datalicious Pty Ltd 32
- 33. > Complex campaign flows = Paid media Organic PR, WOM, search events, etc = Viral elements = Sales channels YouTube, Home pages, Paid TV, print, blog, etc portals, etc search radio, etc Direct mail, Landing pages, Display ads, email, etc offers, etc affiliates, etc CRM Facebook program TwiNer, etc POS kiosks, Call center, loyalty cards, etc retail stores, etc March 2011 © Datalicious Pty Ltd 33
- 34. > Duplica*on across channels Paid Bid Search Mgmt $ Banner Ad Ads Server $ Email Email Blast PlaAorm $ Organic Web Search Analy*cs $ March 2011 © Datalicious Pty Ltd 34
- 35. > Cookie expira*on impact Paid Bid Search Mgmt $ Banner Banner Ad Ad Click Ad View Server $ Email Email Expira*on Blast PlaAorm $ Organic Google Search Analy*cs $ March 2011 © Datalicious Pty Ltd 35
- 36. > ANZ repor*ng plaAorms March 2011 © Datalicious Pty Ltd 36
- 37. > De-‐duplica*on across channels Paid Search $ Banner Ads $ Central Analy*cs PlaAorm Email Blast $ Organic Search $ March 2011 © Datalicious Pty Ltd 37
- 38. Exercise: Duplica*on impact March 2011 © Datalicious Pty Ltd 38
- 39. > Exercise: Duplica*on impact § Double-‐coun.ng of conversions across channels can have a significant impact on key metrics, especially CPA § Example: Display ads and paid search – Total media budget of $10,000 of which 50% is spend on paid search and 50% on display ads – Total of 100 conversions across both channels with a channel overlap of 50%, i.e. both channels claim 100% of conversions based on their own repor.ng but once de-‐duplicated they each only contributed 50% of conversions – What are the ini.al CPA values and what is the true CPA? § Solu.on: $50 ini.al CPA and $100 true CPA – $5,000 / 100 = $50 ini.al CPA and $5,000 / 50 = $100 true CPA (which represents a 100% increase) March 2011 © Datalicious Pty Ltd 39
- 40. > Reach and channel overlap TV/Print audience Banner Search audience audience March 2011 © Datalicious Pty Ltd 40
- 41. > Ad server exposure test Banner TV/Print Search Impression Response Response $ Banner Search Direct Impression Response Response $ Users are segmented before 1st ad is even Exposed group: 90% of users get branded message served Control group: 10% of users get non-‐branded message Banner Search Direct Impression Response Response $ March 2011 © Datalicious Pty Ltd 41
- 42. > Indirect display impact March 2011 © Datalicious Pty Ltd 42
- 43. > Indirect display impact March 2011 © Datalicious Pty Ltd 43
- 44. March 2011 © Datalicious Pty Ltd 44
- 45. > Indirect display impact March 2011 © Datalicious Pty Ltd 45
- 46. > Success aNribu*on models Banner Paid Organic Success Last channel Search Ad Search $100 $100 gets all credit Banner Paid Email Success First channel Ad $100 Search Blast $100 gets all credit Paid Banner Affiliate Success All channels get Search Ad Referral $100 $100 $100 $100 equal credit Print Social Paid Success All channels get Ad Media Search $33 $33 $33 $100 par*al credit March 2011 © Datalicious Pty Ltd 46
- 47. > First and last click aNribu*on Chart shows percentage of channel touch points that lead Paid/Organic Search to a conversion. Neither first Emails/Shopping Engines nor last-‐click measurement would provide true picture March 2011 © Datalicious Pty Ltd 47
- 48. > Full path to purchase Introducer Influencer Influencer Closer $ SEM Banner Direct SEO Online Generic Click Visit Branded Banner SEO Affiliate Social Offline View Generic Click Media TV SEO Direct Email Abandon Ad Branded Visit Update March 2011 © Datalicious Pty Ltd 48
- 49. > Poten*al calls to ac*on § Unique click-‐through URLs § Unique vanity domains or URLs § Unique phone numbers § Unique search terms § Unique email addresses § Unique personal URLs (PURLs) § Unique SMS numbers, QR codes § Unique promo.onal codes, vouchers § Geographic loca.on (Facebook, FourSquare) § Plus regression analysis of cause and effect March 2011 © Datalicious Pty Ltd 49
- 50. > Search call to ac*on for offline March 2011 © Datalicious Pty Ltd 50
- 51. March 2011 © Datalicious Pty Ltd 51
- 52. > PURLs boos*ng DM response rates Text March 2011 © Datalicious Pty Ltd 52
- 53. > Jet Interac*ve phone call data March 2011 © Datalicious Pty Ltd 53
- 54. > Unique phone numbers § 1 unique phone number – Phone number is considered part of the brand – Media origin of calls cannot be established – Added value of website interac.on unknown § 2-‐10 unique phone numbers – Different numbers for different media channels – Exclusive number(s) reserved for website use – Call origin data more granular but not perfect – Difficult to rotate and pause numbers March 2011 © Datalicious Pty Ltd 54
- 55. > Unique phone numbers § 10+ unique phone numbers – Different numbers for different media channels – Different numbers for different product categories – Different numbers for different conversion steps – Call origin becoming useful to shape call script – Feasible to pause numbers to improve integrity § 100+ unique phone numbers – Different numbers for different website visitors – Call origin and .me stamp enable individual match – Call conversions matched back to search terms March 2011 © Datalicious Pty Ltd 55
- 56. > Cross-‐channel impact March 2011 © Datalicious Pty Ltd 56
- 57. > Offline sales driven by online Adver*sing Phone Credit check, campaign order fulfilment Retail Confirma*on order email Website Online Online order Virtual order research order confirma*on confirma*on Cookie March 2011 © Datalicious Pty Ltd 57
- 58. > Full path to purchase Introducer Influencer Influencer Closer $ SEM Banner Direct SEO Online Generic Click Visit Branded Banner SEO Affiliate Social Offline View Generic Click Media TV SEO Direct Email Abandon Ad Branded Visit Update March 2011 © Datalicious Pty Ltd 58
- 59. Adobe campaign stack does not include organic channels or banner impressions and does not expire on any event, i.e. con*nues as long as the cookie is present. March 2011 © Datalicious Pty Ltd 59
- 60. > Where to collect the data Ad Server Web Analy*cs Banner impressions Referral visits Banner clicks Social media visits + Organic search visits Paid search clicks Paid search visits Email visits, etc Lacking organic visits Lacking ad impressions More granular & complex Less granular & complex March 2011 © Datalicious Pty Ltd 60
- 61. > Maximise iden*fica*on points Mobile Home Work Online Phone Branch March 2011 © Datalicious Pty Ltd 61
- 62. > Combining data sources March 2011 © Datalicious Pty Ltd 62
- 63. > Single source of truth repor*ng Insights Repor*ng March 2011 © Datalicious Pty Ltd 63
- 64. > Understanding channel mix March 2011 © Datalicious Pty Ltd 64
- 66. > Website entry survey De-‐duped Campaign Report Greatest Influencer on Branded Search / STS } Channel % of Conversions Channel % of Influence Straight to Site 27% Word of Mouth 32% SEO Branded 15% Blogging & Social Media 24% SEM Branded 9% Newspaper Adver.sing 9% SEO Generic 7% Display Adver.sing 14% SEM Generic 14% Email Marke.ng 7% Display Adver.sing 7% Retail Promo.ons 14% Affiliate Marke.ng 9% Referrals 5% Conversions arributed to search terms Email Marke.ng 7% that contain brand keywords and direct website visits are most likely not the origina.ng channel that generated the awareness and as such conversion credits should be re-‐allocated. March 2011 © Datalicious Pty Ltd 66
- 67. > Adjus*ng for offline impact -‐5 -‐15 -‐10 +5 +15 +10 March 2011 © Datalicious Pty Ltd 67
- 68. > Success aNribu*on models Introducer Influencer Influencer Closer $ Even 25% 25% 25% 25% ANrib. Exclusion 33% 33% 33% 0% ANrib. PaNern 30% 20% 20% 30% ANrib. March 2011 © Datalicious Pty Ltd 68
- 69. > Path across different segments Introducer Influencer Influencer Closer $ Product Channel 1 Channel 2 Channel 3 Channel 4 A vs. B New Channel 1 Channel 2 Channel 3 Channel 4 prospects Exis*ng Channel 1 Channel 2 Channel 3 Product 4 customers March 2011 © Datalicious Pty Ltd 69
- 70. Exercise: ANribu*on model March 2011 © Datalicious Pty Ltd 70
- 71. > Exercise: ANribu*on models Introducer Influencer Influencer Closer $ Even 25% 25% 25% 25% ANrib. Exclusion 33% 33% 33% 0% ANrib. ? ? ? ? Custom ANrib. March 2011 © Datalicious Pty Ltd 71
- 72. > Common aNribu*on models § Allocate more conversion credits to more recent touch points for brands with a strong baseline to s.mulate repeat purchases § Allocate more conversion credits to more recent touch points for brands with a direct response focus § Allocate more conversion credits to ini.a.ng touch points for new and expensive brands and products to insert them into the mindset March 2011 © Datalicious Pty Ltd 72
- 73. 101011010010010010101111010010010101010100001011111001010101 010100101011001100010100101001101101001101001010100111001010 010010101001001010010100100101001111101010100101001001001010 > Targe*ng and tes*ng March 2011 © Datalicious Pty Ltd 73
- 74. > Increase revenue by 10-‐20% Capture internet traffic Capture 50-‐100% of fair market share of traffic Increase consumer engagement Exceed 50% of best compe.tor’s engagement rate Capture qualified leads and sell Convert 10-‐15% to leads and of that 20% to sales Building consumer loyalty Build 60% loyalty rate and 40% sales conversion Increase online revenue Earn 10-‐20% incremental revenue online March 2011 © Datalicious Pty Ltd 74
- 75. > New consumer decision journey The consumer decision process is changing from linear to circular. March 2011 © Datalicious Pty Ltd 75
- 76. > New consumer decision journey The consumer decision process is changing from linear to circular. Online research Change increases the importance of experience during research phase. March 2011 © Datalicious Pty Ltd 76
- 77. > The consumer data journey To transac*onal data To reten*on messages From suspect to prospect To customer Time Time From behavioural data From awareness messages March 2011 © Datalicious Pty Ltd 77
- 78. > Coordina*on across channels Genera*ng Crea*ng Maximising awareness engagement revenue TV, radio, print, Retail stores, in-‐store Outbound calls, direct outdoor, search kiosks, call centers, mail, emails, social marke.ng, display brochures, websites, media, SMS, mobile ads, performance mobile apps, online apps, etc networks, affiliates, chat, social media, etc social media, etc Off-‐site On-‐site Profile targe*ng targe*ng targe*ng March 2011 © Datalicious Pty Ltd 78
- 79. > Combining targe*ng plaAorms Off-‐site targe.ng Profile On-‐site targe.ng targe.ng March 2011 © Datalicious Pty Ltd 79
- 81. ANZ Low Rate MasterCard
- 82. ANZ Business Debit Card
- 84. > Combining technology On-‐site Off-‐site segments segments CRM March 2011 © Datalicious Pty Ltd 84
- 85. > SuperTag code architecture § Central JavaScript container tag § One tag for all sites and placorms § Hosted internally or externally § Faster tag implementa.on/updates § Eliminates JavaScript caching § Enables code tes.ng on live site § Enables heat map implementa.on § Enables redirects for A/B tes.ng § Enables network wide re-‐targe.ng § Enables live chat implementa.on March 2011 © Datalicious Pty Ltd 85
- 86. > Combining data sets Website behavioural data Campaign response data + The whole is greater than the sum of its parts Customer profile data March 2011 © Datalicious Pty Ltd 86
- 87. > Behaviours plus transac*ons Site Behaviour CRM Profile tracking of purchase funnel stage one-‐off collec.on of demographical data + browsing, checkout, etc age, gender, address, etc tracking of content preferences customer lifecycle metrics and key dates products, brands, features, etc profitability, expira*on, etc tracking of external campaign responses predic.ve models based on data mining search terms, referrers, etc propensity to buy, churn, etc tracking of internal promo.on responses historical data from previous transac.ons emails, internal search, etc average order value, points, etc Updated Con*nuously Updated Occasionally March 2011 © Datalicious Pty Ltd 87
- 88. > Maximise iden*fica*on points 160% 140% 120% 100% 80% 60% −−− Probability of iden.fica.on through Cookies 40% 20% 0 4 8 12 16 20 24 28 32 36 40 44 48 Weeks March 2011 © Datalicious Pty Ltd 88
- 89. > Maximise iden*fica*on points Mobile Home Work Online Phone Branch March 2011 © Datalicious Pty Ltd 89
- 90. > Sample customer level data March 2011 © Datalicious Pty Ltd 90
- 91. > Sample site visitor composi*on 30% new visitors with no 30% repeat visitors with previous website history referral data and some aside from campaign or website history allowing referrer data of which 50% to be segmented by maybe 50% is useful content affinity 30% exis*ng customers with extensive 10% serious profile including transac.onal history of prospects which maybe 50% can actually be with limited iden.fied as individuals profile data March 2011 © Datalicious Pty Ltd 91
- 92. > Prospect targe*ng parameters March 2011 © Datalicious Pty Ltd 92
- 93. > Affinity re-‐targe*ng in ac*on Different type of visitors respond to different ads. By using category affinity targe.ng, response rates are lited significantly across products. CTR By Category Affinity Message Postpay Prepay Broadb. Business Blackberry Bold - - - + Google: “vodafone 5GB Mobile Broadband - - + - omniture case study” Blackberry Storm + - + + or hNp://bit.ly/de70b7 12 Month Caps - + - + March 2011 © Datalicious Pty Ltd 93
- 94. > Ad-‐sequencing in ac*on Marke.ng is about telling stories and stories are not sta.c but evolve over .me Ad-‐sequencing can help to evolve stories over .me the more users engage with ads March 2011 © Datalicious Pty Ltd 94
- 96. > Exercise: Targe*ng matrix Purchase Segments: Colour, price, Media Data Cycle product affinity, etc Channels Points Default, awareness Research, considera*on Purchase intent Reten*on, up/cross-‐sell March 2011 © Datalicious Pty Ltd 96
- 97. > Exercise: Targe*ng matrix Purchase Segments: Colour, price, Media Data Cycle product affinity, etc Channels Points Default, Have you Have you Display, Default awareness seen A? seen B? search, etc Research, A has great B has great Search, Ad clicks, considera*on features! features! website, etc prod views Purchase A delivers B delivers Website, Cart adds, intent great value! great value! emails, etc checkouts Reten*on, Why not Why not Direct mails, Email clicks, up/cross-‐sell buy B? buy A? emails, etc logins, etc March 2011 © Datalicious Pty Ltd 97
- 98. > Quality content is key Avinash Kaushik: “The principle of garbage in, garbage out applies here. [… what makes a behaviour targe;ng pla<orm ;ck, and produce results, is not its intelligence, it is your ability to actually feed it the right content which it can then target […. You feed your BT system crap and it will quickly and efficiently target crap to your customers. Faster then you could ever have yourself.” March 2011 © Datalicious Pty Ltd 98
- 99. > ClickTale tes*ng case study March 2011 © Datalicious Pty Ltd 99
- 100. > Bad campaign worse than none March 2011 © Datalicious Pty Ltd 100
- 102. > Exercise: Tes*ng matrix Test Segment Content KPIs Poten*al Results March 2011 © Datalicious Pty Ltd 102
- 103. > Exercise: Tes*ng matrix Test Segment Content KPIs Poten*al Results New Conversion Next step, Test #1A prospects form A order, etc ? ? New Conversion Next step, Test #1B prospects form B order, etc ? ? New Conversion Next step, Test #1N prospects form N order, etc ? ? ? ? ? ? ? ? March 2011 © Datalicious Pty Ltd 103
- 104. > Keys to effec*ve targe*ng 1. Define success metrics 2. Define and validate segments 3. Develop targe.ng and message matrix 4. Transform matrix into business rules 5. Develop and test content 6. Start targe.ng and automate 7. Keep tes.ng and refining 8. Communicate results March 2011 © Datalicious Pty Ltd 104
- 105. Contact us cbartens@datalicious.com Learn more blog.datalicious.com Follow us twiNer.com/datalicious March 2011 © Datalicious Pty Ltd 105
- 106. Data > Insights > Ac*on