资源描述
DIGITAL STRATEGY IN THE AGE OF ARTIFICIAL INTELLIGENCEHOW TO TAKE ADVANTAGE OF AI IN ADVERTISINGARTIFICIAL INTELLIGENCE: IMAGINATION AND REALITYIn our imaginations, artificial intelligence often comes in blockbuster proportions: the computers takeover in “2001, A Space Odnullssenull,nullbattles for humanitnulls fate nullth multinullimbed robots in “nullhe nulltrinullnullthe meaning of love in “nullrnullnullIn realitnull, AI is more pervasive, if less epicnullAI algorithms help Inulls nullep nullue and nullatson nulln chess and “nullopardnullnulltournamentsnullnulloicenullctivated assistants give information, entertain, and control devices and homesnullSelfnullriving cars nullll go on sale to car services nenull nullear, nullneral nulltors promisesnull1AI, EXPLAINEDAt its core, artificial intelligence processes specific inputs then delivers specific outputsnullAsked the right nullestions and given the right instructions, its algorithms nullll provide solutions and suggest actions that produce desired resultsnullnullhe problems addressed bnullAI are sometimes ones that people could solve nullthout intervention, on a smaller scalenullAI could, for enullample, match one facial image to another, predict nenull nulleeks sales, or sanullnullether someones credit histornullshonulls thenullre likelnullto panulloff a loannullnullnullbreaking the problem into component parts, each relevant input can be nulleighted and a solution derivednullnullecogninullng a face, for enullample, is a combination of evaluating the curve of a nullnull, the angles of a nose, the color of someones enulles, and so on, nullth some factors being more pertinent to the solution than othersnull “AI is like a spreadsheet on steroids,nullsanulls Sara nullobertson, null of nulloduct nullgineering for nullanulls, and an enullert in artificial intelligencenullAI usuallnullnullorks in concert nullth machine learning so that algorithms get increasinglnullbetter at increasinglnullcomplenulltasks like pattern recognition and predictive analnullsisnullnullchine learning takes the results produced from a round of instructions, compares them to the predicted outcomes, evaluates, then sends information back on honullto adnullst and optiminullenullnullrther enhancement comes from processes such as “neural netnullorksnullnullconnected computing nodes nullorking together to increase processing Artificial intelligence (AI) has changed what we can achieve in media buying and planning, how to achieve it, and the metrics used to understand success. This paper lays out what AI is, strategies to apply it to advertising, and the best methods for gaining expertise, evaluating partners, and working with them to leverage the opportunities afforded by AI to achieve the best possible business outcomes.EXECUTIVE SUMMARY:AI is a computers abilitnullto choose and perform the right machine learning techninulles at the right time, successfullnull, regularlnull, and nullth a minimum of effortnull2ponuller nulland “deep learning,null nullich helps refine the understanding those neural netnullorks can producenullAI algorithms can nullork nullsupervised nullnullth people indicating nullich results are closer to the desired outcomes nullor nullnsupervised, nullere thenull enullecute and adnullst on their onull, usuallnullafter a period of human instruction nullsee rightnullAI todanullis used in nearlnullevernull industrnullin nullanulls nulle access evernulldanullnullIt helps search engines find their targetsnull enullecutes complenulldecisions for financial tradingnullponullers programming recommendations for services like nulltflinull2nullIt can also aid in content curation, enhance cnuller securitnull, improve nullarehouse inventornull management, assist salespeople in generating better leads, and help flnullairplanesnullAI FOR ADVERTISINGnullor advertising, AI is being used in a varietnullof nullanulls to improve effectivenessnullIt has been used to find and define audiences, refine creative messaging nullsee graphic page null, generate audience personas, and develop bidding strategies that optiminulle for clients stated goalsnull“AI has mannullseeminglnullsmall applications that currentlnull deliver digital marketing efficiencies for companies all over the nullorld nullfrom advanced consumer targeting and insights to highlnull personalinulled ad enulleriences,null sanulls Adam nullonull, Snull of nullsplanull at nullkuten nullrketingnullnullAt present, most advertisers arent taking advantage of the full capabilities of AInullInstead, thenulldeplonullit to achieve simple goalsnullnullt it can do much more than elevate discrete performance metricsnullnullen the mannullapplications of AI are used in concert, thenullcan add up to a significant transformation in digital advertising strategnullthat drives remarkablnullimproved resultsnullnullo accomplish that, nenullcampaigns must be conceived and built around the uninulle opportunities and strategies afforded bnullAInullnullhis renullires advertisers to shift their perspectivenullto refine and enulland their idea of marketing success and approach nenullcampaigns in cooperation nullth the nullanullAI nullorksnullnullhe most ponullerful nulland largelnullunfulfilled nullpotential of AI lies in the bigger picture, in its abilitnullto optiminulle tonullards business outcomes rather than simple metricsnullOne car maker, for enullample, sought to increase salesnullnullo do so, it evaluated the influence of individual performance metrics nullthin the contenull of that larger goalnullIt nulleighted factors such as nullebsite interactions, brochure donullloads, and shonulloom visits to determine messages that ultimatelnulllead to that outcome, and used artificial intelligence nullth machine learning to continuallnulloptiminulle tonullard better scores nulltherebnullmaking its marketing evermore effectivenullnullvsSUPERVISEDnullrain the model on data nullere the correct ansnuller is includednullnullrain the model on data nullth no ansnullers and see nullat it comes up nullthnullE.g. What bid price is linullelnull tnullnullnullthis inullressinullnullE. g. What dnullpenullle nullnullclicnullnullnullad hanulle inullcnullnullnullnullUNSUPERVISEDSUPERVISED VS. UNSUPERVISED, APPLIED TO ADVERTISING3AI IN PROGRAnullATInullnullogrammatic advertising is enullceptionallnullnullellnullsuited for AInullA nullorld nullth billions of impressions auctioned in fractions of a second and alnullanullsnullchanging circumstances creates a scale of multinullactorial problems that can onlnullbe solved effectivelnullnullth the help of AInull “nullo achieve an obnullctive becomes vernullhard nullthout the help of an AI platform that can do a lot of the heavnulllifting,nullsanulls nullanulls nullO nullcolas nulldonnull“nullhe amount of data, the combinations that can result, is gronullng enullonentiallnullto the point that a human nullll have trouble determining the right bidding strategnull to bunullmedia for a clientnullnullDESIGNING MEDIA STRATEGIES FOR AN AI WORLDIts a given advertisers nullant to reach the right person at the right time nullth the right message, and of course, at the right pricenullDIGITAL STRATEGnullIS EVOLVINGnullt, in the digital era, a lot of advertising media strategies have focused on targeting audience segments that seem to contain propitious prospectsnullnullssages are tailored to them via medium, platform, and screen, and, nullen possible, factor in behaviors and localenullAdvertisers run their messages, gather results, then trnullto optiminulle against marketing metrics such as completed vienullrates, time of enullosure, clickthrough rates nullnullnullnulland effective cost per thousand impressions nullenullnullnullnullet, evernullone of those measurements is an imperfect pronullnullfor nullats actuallnulldesired: salesnullnullnulloptiminullng tonullard nullnullamong a target audience, for enullample, a media bunuller manullbe getting a lot of the “rightnull people onto a nulleb page, but thats not necessarilnulla measure of sales efnulliencnullnullnullnullcontrast, artificial intelligence algorithms null correctlnullinstructed nullcan help optiminulle marketing plans tonullard better sales metricsnullAudience segments based on demographic, behavioral, and geographic characteristics can be highlnulleffective, but thenullnullll alnullanulls miss a large portion of potential purchase intenders nullo deviate from the standard definitionsnull nullgital media strategies that use AI to identifnulland locate prospects nullthout bias or assumptions nullll find customers, not nullst segmentsnullnullake, for enullample, a highnullnd home appliances maker that manullbe nulllling to spend null00 to sell one machinenullArticulated in that nullanullnullrather than in terms of metrics like demographics or nullnullnulldata nulltermining the bidding strategnullto achieve specinull outcomes becomes vernull hard nullthout the help of an AI platformnull4scientists and engineers can focus on optiminullng for the ratio of the cost to market the appliance against the number of appliances that marketing sellsnullnullith machine learning, AI manulleven make some counternullntuitive leaps, such as finding that a higher volume of less enullensive clickthroughs leads to better business results than more targeted nullnullnullnullus, a machinenullonullered bidding strategnullcan enullract manullmum value bnullconducting large numbers of realnullime tests in vernullgranular increments nullell benullond the capabilities of annull humannullnullere a programmatic specialist might test bids in five broad increments from nullnull00 to nullnull00, AI can run 10,000 tests in that range in increments of fractions of a pennnullto hone in on preciselnullthe most costnullffective bidnullIn this process, a feedback loop is created in nullich the outcomes continuallnullimprove in lightningnullast increments that add up to big resultsnullAI, APPLIED TO DYNAMIC CREATIVE OPTIMIZATIONTESTAds run based on models in small samplesMEASUREnullfectiveness evaluated against outcomesLEARNInputs refined to make improved effectivenessMODELnullogrammatic specialists say what should workVARIAnullESnullOLORnull GEOGRAPnullnull nullONTENTnull5AI can be used to help with dynamic creative optiminulltion (nullnull), continually increasing effectiveness by choosing combinations of creative components, testing outcomes, and improving. nullorking with skilled professionals, AI used for nullnullwill create many multiples more creative possibilites than humans could at a fraction of the cost.EXECUTING AN AI-POWERED STRATEGYOne of the biggest challenges in emplonullng AI is defining the desired outcomes in nullanulls the machines can understand and properlnullact uponnullAn advertising bidding strategnull that applies AI likelnullstarts nullth programmatic specialists nullo have a thorough understanding of the current advertising landscape and can: Design. nullild a uninulle plan based on the renullested parameters and desired business outcomes, including forecasting the likelihood of delivernullnull Inullnullnullnnull nulln complenullmodeling on large data sets to generate performance predictions and custominulle bidding strategies in nullSnull nulldemandnullide platformsnull Onullnullnullnulle. Improve the enullecution of the AI to make it more effective and to achieve better outcomesnull Annullnullnulle. nullevienullthe results earned from the AI in reports that are uninullelnull formatted to provide insights into the effectiveness of the algorithms and connullurationsnull Tnull nullnullTO SnullnullSSnull TALENTnullhose strategies renullire not onlnull access to premium marketing opportunities but also the skills to handle premium technologiesnull In addition to marketing enullertise, programmatic specialists nullll need data science and engineering acumen to articulate the steps in nullanulls that lead AI to the best resultsnullSnullennullsnull can define the fundamentals of the pronullct, making sure its soundnullnullhenullnullll:null nullvise a proof of concept.null Test algorithms against those proofs.null nullork with the data.null nullcide on inputs and outputs .nullhen the scientists nullll choose an algorithm thenullbelieve nullll nullork for the case, make it better, run enulleriments, and make recommendations for engineers to enullecutenullEngineenull nullll then: nullut the algorithms into production and use. nullather and further refine the data.null nullke the algorithms repeatable. null nullt them to scale.null nullork with servers to make sure they can handle the demands put on them.null nullgure out how to handle edge cases that may not have been predicted.null nullke recommendations for further improvements, and for the scientists to test.nullile mannullof the best algorithms used in AI are freelnull available via a communitnullof leading scientists, it takes great enullertise to assemble, custominulle, and deplonullthem intelligentlnullnull“nullou cant nullst donullload an algorithm and enullect it to nullork out of the bonullnullnullobertson sanullsnull “It needs to be custominulled to nullour business, nullich is nullnulldata scientists and engineers nullo specialinulle in AI are so innullemandnullnullPARTNERING nullAssuming such talent is not available innullouse, the task for marketers becomes honullto evaluate potential AI partnersnullnullderstanding the imperatives described above, thenullmight ask if a potential partner can:null nullve thousands of examples of machine learning models and algorithms theynulle run.null Indicate what revenue their AI installations have produced.null nullow when their AI outperformed manually traded campaigns. (nulle case studies.)null nullndle multiple competing renullirements across multiple geographies and languages.nullerhaps most cruciallnull, the potential partner must give enullamples of nullanulls thenullve custominulled their algorithms to not onlnullnullork nullth clients nullIs but also their proprietarnull, firstnullartnulldatanullAfter all, that data helps fuel the AI and give brands a competitive edgenull“It manullbe easnullto replicate softnullare, but it is impossible to replicate someones proprietarnull data,nullnullobertson notesnull“Its the custominulltion that gives algorithms their real ponuller and makes them most anullesomenullnullnulllgorithmsnull need to be custominulled to nullour business, nullich is nullnull data scientists and engineers nullo specialinulle in AI are so innullemandnullnull Sara Robertson VP nullnullPnullnullnullnullEngineenullng, Xnullis6CASE STUDIESCASE STUDY null TELCOBETTER OnullnullOnullS nullING AI FOR RETARGETINGA telecommunications company historically had a complex setup, with nullnullnull campaigns being manipulated in a programmatic advertising platform. nulltiminullng towards performance renullired timenu
展开阅读全文