customercentricsystems esteve almirall

Transcription

1 customercentricsystems esteve almirall

2 Advertising From Mass Media to Personalization

3 advertising\global spending> Digital, mobile & interactive are the key to industry growth rate (outpaced industry forecasts by 25%-40% in the last 2 years). Mature channels (print, direct marketing and TV) have single digit growth forecasts. Combined growth fo interactive formats over 20%.

4 advertising\audiences>

5 advertising\audiences> Consumers have more options for entertainment. TV vs. internet, game consoles, mobile devices, TV increasingly becoming a secondary background media More consumers spend significant amount s of time in Internet than with TV. Especially & gadgedtiers (early adopters w > 4 multimedia devices). Fragmentation of both audience and attention Mass adoption only in Social Networking w respondents under 35 Younger audiences are far less willing to pay.

6 advertising\audiences> main trends As users migrate to new screens so will do advertising dollars [ads sponsor 50% of TV in major markets] Synchronization Users in control VOD used by 50% of respondents in US & UK DVR penetration in USA almost 50% 50% of respondents watch at least 50% of TV contents on reply with ad skipping capabilities Multimedia devices Germany >> 70% mp3 40% internet phone Global >> 50% mp3 20% internet phone Mobile TV

7 advertising\audiences>

8 advertising\usergeneratedcontents>

9 advertising\usergeneratedcontents>

10 advertising\usergeneratedcontents>

11 advertising\usergeneratedcontents>

12 advertising\usergeneratedcontents> 26% US and 20% UK contribute to SN sites. [YouTube 32% find contents following recommendations from friends] Advertisers bypassing agencies [Current.tv pays $1000 for viewer created ads that they use] Companies bypassing agencies and working directly with consumers [Conde Nast has a poll of 10,000 consumers] Opening for aggregators Do you want to see the future? MySpace and FaceBook

13 advertising\measuerement> from impression to impact If consumer attention fragments measures have to change to be relevant Micro tracking from sampling to real time Micro / Individual targeting personalization Customization with automation - [later in this session]

14 advertising\measuerement> 2/3 expect 20% of advertisement revenue to shift from impression to impact based formats Measures related to old one-tomany formats have little sense in environments where you can interact (Second Life), network or trigger events.

15 advertising\marketplaces> Today most inventory systems involve few buyers and sellers. Emerging platform players lead by Google AdWords Improved inventory management Pricing transparency Streamlined buying/selling process Improved analysis capabilities Creation of large global marketplaces Panel expects 30% of advertising revenues to shift to open markets in the next 5 years

16 Marketplace closed open advertising\scenarios> future scenarios Open Exchanges Ad Marketplace Two major factors : User Control Consumers select, filter, interact, with mk messages Marketplaces From closed to open markets Evolution Consumer Choice producers User Control consumers

17 Marketplace closed open advertising\scenarios> Open Exchanges Evolution Ad Marketplace Consumer Choice evolution New technologies are assimilated by existing market incumbents who extend their influence to UGC and one-to-one markets. producers consumers User Control

18 Marketplace closed open advertising\scenarios> Open Exchanges Ad Marketplace open exchange Evolution Consumer Choice producers consumers User Control Shift from close distribution system to a widespread use of ad marketplaces. Reshuffle of power to exchanges that control a sizeable part of revenues. Inventory once available only to large advertisers becomes open and available to small buyers.

19 Marketplace closed open advertising\scenarios> Open Exchanges Ad Marketplace consumer choice Evolution Consumer Choice producers consumers User Control T o remain relevant distributors must offer more consumer choices like optin, permission marketing and targeted messaging. Opt-in information is combined with behavioral analysis and more sophisticated measures.

20 Marketplace closed open advertising\scenarios> Open Exchanges Evolution Ad Marketplace Consumer Choice producers consumers User Control ad marketplace Change in the backend systems towards open exchange combined with customers rejection of traditional advertising models. Consumers become involved not only in the production with UGC but also in the distribution with buzz/viral mk. Measurement systems are on-line and immediately impact the price of ads.

21 advertising\marketplaceexperiments> ad marketplace experiments Online: AdSense Contextual advertising sold by Google on their web sites Cable/Satellite: Astaound Me, Echo Star Leverage Google on-line on cable (Astound Me) or on satellite (Echo Star) Radio: dmarc (2006) Google AdSense in the radio industry Newspaper: PrintAds (2007) Google AdSense in the newspaper industry Product placement for film and TV Ad platform for the cable industry

22 advertising\responsescustomerchoice> responses to Customer Choice Pop-up messages while consumers are watching of fast-forwarding programs. Pay 1 minutes of free air times for every 5 spent interacting with ads (response rates of 5%, gave away 9 minutes in the first year). High definition on-demand movies with ads. Programming available on-line with ads.

23 advertising\technology> a technology driven process? New opportunities created by technology Internet Adsl 2 Mobile Devices Technology shaping the market Auction platforms UGC Blogosphere, You Tube Targeting recommenders, profiling Platforms for user interaction Facebook, Open Social Mobile The New Frontier Shift from Product/Services to Platforms

24 do you want 1 million dollars?

25

26

27

28 $1M 10% better

29

30

31

32 why?

33 recommenders\why>

34 recommenders\why>

35 recommenders\why> Scarcity Abundance

36 recommenders\why>

37 recommenders\why>

38 recommenders\why>

39 recommenders\why> where are recommenders used?

40 how?

41 recommenders\how> general idea User Profile compare Set of Items News Articles Books Music Movies Restaurants recommend items user will like

42 recommenders\how> personalization: general idea User Profile Interaction is adapted based on the data about an individual user. For example, personalized webpages, personalized information in museums, personalized news,

43 recommenders\how> user profile simple Interests, preferences, expertise level ratings complex demographical info (age, gender, location) purchase records, observed behavior Lifelog

44 recommenders\how> sources of user profile Entered explicitly by the user. Gathered implicitly by the system. Observing and recording the users behavior. Learning user s interest and preferences. Combination of both aproaches. Another dimension: public/private.

45 recommenders\how> acquired user profiles FROM Raw data. Generalization (find patterns and generalize) Statistical ML methods. Knowledge based ML methods. Relevancy of time (window of data and aging). Generic, reusable profiles still mostly theory (FOAF, ).

46 recommenders\how> recommendation algorithms Case-based /Stereotype based. Featured-based /Content-based. Collaborative Filtering.

47 recommenders\how> Case-based /Stereotype-Based Acquire a user profile. Classify user in a bucket (can be a hierarchy). Soccer moms, Produce case solutions adapting them to the concrete user. Recommend these solutions to users. e.g. recommender for tourists.

48 recommenders\how> feature-based /content-based filtering Learning from item examples 1. Look at all items a user likes. 2. Find patterns and generalize (usually clustering). 3. Recommend more items that fit the same patterns.

49 recommenders\how> feature-based /content-based filtering Learning from user examples 1. Given a category of items. 2. Given a set of users. 3. Generalize which users like a category of items. 4. Recommend items.

50 items recommenders\how> collaborative filtering users Sparse matrix of ratings or purchase patterns Algorithms recommend items based on item similarities (itemto-item) or user similarities (userto-user). Typically a correlation coefficient is used for recommendation.

51 recommenders\how> collaborative filtering PROs CONs Does not require analysis of the items (features). Better at qualitative judgments. Bootstrapping. Ratings required. Critical mass required.

52 recommenders\how> tunnel vision and the importance of serendipity Feedback loop System recommends items of type X User consumes items of type X Importance of exploration or serendipity (recommending items outside user s interest space).

53

54

55

56

57

58 Esteve Almirall