Gary Angel, Pres­i­dent and CTO of Sem­phonic, a Web ana­lyt­ics con­sul­tancy, recently hosted an Omni­ture Devel­oper Con­nec­tion User Group in San Fran­cisco. Gary co-founded Sem­phonic and leads their con­sult­ing efforts for com­pa­nies like Amer­i­can Express, Charles Schwab, Intuit, Genen­tech, Nokia, Sears and Turner Broad­cast­ing. Gary has pub­lished arti­cles on Web and SEM Ana­lyt­ics in DM News, Amer­i­can Demo­graph­ics, CRM Guru, CRM Buyer, IMe­di­a­Con­nec­tion, Busi­ness Geo­graph­ics and Busi­ness Insur­ance. As an early-adopter of the Omni­ture APIs, Sem­phonic has been build­ing inte­grated solu­tions that com­bine Web ana­lyt­ics with enter­prise data to help cus­tomers real­ize new mar­ket­ing inno­va­tion and effi­ciency through the Omni­ture plat­form. After the San Fran­cisco User Group, Gary and I con­nected to talk more about Semphonic’s use of the Omni­ture APIs:

Q:  At the Omni­ture devel­oper user group, you men­tioned that cus­tomers are pulling you into inte­gra­tion projects that really advance beyond cur­rent web ana­lyt­ics par­a­digms. What are cus­tomers ask­ing you to inte­grate and why?

A: One of the things that make inte­gra­tion projects chal­leng­ing is that they tend to be pretty unique. I guess I’d say that inte­gra­tions tend to fall into one of three basic types. The most com­mon is prob­a­bly the inte­gra­tion of mul­ti­ple man­age­ment report­ing streams. The vast major­ity of our cus­tomers are multi-channel. For many of them, their source of ulti­mate truth are CRM and finan­cial sys­tems – and data from these sys­tems forms the bulk of what gets reported up to senior man­age­ment. But the top end of the fun­nel – where online behav­ior occurs – needs to be rep­re­sented there as well. The API makes it pos­si­ble to pull that online behav­ioral infor­ma­tion at the same time as you are tap­ping into these other sys­tems – and push­ing the infor­ma­tion into a sin­gle uni­fied (and auto­mated) report. The sec­ond type of inte­gra­tion, less com­mon but to me more inter­est­ing, is when we pull infor­ma­tion so that we can ana­lyze it in more detail. In a way, our SAINT appli­ca­tions are like that – we take advan­tage of the full-feature reg­u­lar expres­sion libraries in .Net to built rule-based SAINT tables auto­mat­i­cally – some­thing that just isn’t pos­si­ble oth­er­wise. Finally, we’ve worked on a few inte­gra­tions where the goal was sim­ply to pull data out of the ana­lyt­ics and into some piece of an online sys­tem (usu­ally to push or opti­mize some part of con­tent). That’s use­ful but also very pro­vi­sional – ulti­mately you’d hope that peo­ple use more com­plete tools for that kind of work.

Q: For those who aren’t famil­iar with the SAINT (Site­Cat­a­lyst Attribute Import­ing and Nam­ing Tool) API, can you share an exam­ple of the kind of data you use it for and why?

A: SAINT is used to gen­er­ate lookup tables in Site­Cat­a­lyst. Prob­a­bly the most com­mon use is for com­pa­nies to assign friendly names and rollup cat­e­gories to cam­paigns based on a cam­paign code cap­tured from the URL. But it can be used for almost any kind of vari­able. Our clients often use SAINT tables for prod­ucts and mer­chan­dis­ing cat­e­gories, arti­cle names, video meta­data and more. This all works fine where you con­trol the meta­data that comes in (like prod­uct id to prod­uct name). But it’s nice to be able to cat­e­go­rize exter­nal vari­ables as well — things like SEO key­words, arti­cle tags, and inter­nal search key­words. To effec­tively ana­lyze any of these, you need to cat­e­go­rize them. In the­ory, SAINT let’s you do this. But SAINT tables are strictly a 1-to-1 lookup. You can’t apply any logic like “if the key­word phrase con­tains XXX assign it to Y Cat­e­gory.” That makes SAINT very dif­fi­cult and man­ual for this type of appli­ca­tion. We use the SAINT API to pull all the val­ues from a Site­Cat­a­lyst vari­able and then apply a series of reg­u­lar expres­sion or lookup rules to cat­e­go­rize each value — then auto­mat­i­cally gen­er­ate the SAINT table. It’s almost the only prac­ti­cal way to do analy­sis on large open-ended fields like SEO key­words or inter­nal search key­words and it’s also very use­ful for sophis­ti­cated SEO reporting.

Q: Sem­phonic was a very-early-adopter of the Omni­ture APIs. In your blog, you write about the increas­ing matu­rity of the Omni­ture APIs. Where have you seen improve­ments and how is that help­ing you serve cus­tomer needs better?

A: It has improved a lot – and in many of the ways you’d most expect as a soft­ware sys­tem achieves matu­rity. The doc­u­men­ta­tion and exam­ples are a lot bet­ter now. That’s a big deal when you’re first start­ing out with a new soft­ware tool. We hap­pen to be a .Net shop, and when we first got started the secu­rity model was just a killer. It took for­ever just to fig­ure out how to authen­ti­cate. That’s really frus­trat­ing of course because you feel like you’re just spin­ning your wheels. It’s also really hard, as a con­sult­ing firm, to jus­tify those hours. You can’t tell a client you couldn’t fig­ure out how to logon to the sys­tem – so you end up hav­ing to eat all those “edu­ca­tional” hours. Now there are good exam­ples that make this pretty easy. The extent of the APIs has also improved a lot. The cov­er­age is pretty darn good now. Finally, the cost model and token model have improved and been clarified.

Q: What would you like to see on the roadmap for the Omni­ture APIs?

A: Like most devel­op­ers, I’d still like to see a throt­tling mech­a­nism as opposed to the cur­rent token mech­a­nism for con­trol­ling usage. I’ve also cam­paigned for a sys­tem where 3rd Par­ties like us could buy tokens and use them for cus­tom apps to sim­plify our client arrange­ments. In terms of the APIs them­selves, I guess I’d like to see direct access to the event level data ala the Data Feed. I think it would be great to be able to cus­tomize and launch data feed requests and get back that true server-call level data.

Q: How do you see cus­tomer require­ments evolv­ing, Vis a Vis Web ana­lyt­ics inte­gra­tion with multi-channel mar­ket­ing efforts, in 2010?

A: At the upper-end of the matu­rity curve I think 2010 is going to see a lot of move­ment on the customer-level inte­gra­tion of web ana­lyt­ics data. To be hon­est, I thought a lot of this would hap­pen in 2009, but it was such a tough year for big projects that I think most orga­ni­za­tions ended up just push­ing these types of projects out. My sense is that lot’s of com­pa­nies are ready to com­bine key online events with their other customer/visitor data. Doing that often means mov­ing data in both direc­tions: cer­tain kinds of cus­tomer data need to move out to the web ana­lyt­ics solu­tion so that you can under­stand who you’re pro­fil­ing and what their online behav­ior means. Then, as behav­ior man­i­fests itself online, you need to be able to move that data back into your CRM and mar­ket­ing ware­house sys­tems for out­bound messaging.

Q: What’s the most inter­est­ing use of Web ana­lyt­ics data you’ve seen, out­side the report­ing envi­ron­ment of Sitecatalyst?

A:  I have two ways of answer­ing this ques­tion. The most “pow­er­ful” uses of web ana­lyt­ics data I’ve seen are usu­ally inte­gra­tions that take advan­tage of event-based mar­ket­ing. Vis­i­tor X did this online – respond with these changes to the site, this out­bound email, this cus­tomer com­mu­niqué. This is pow­er­ful stuff and it takes a lot of mar­ket­ing oper­a­tions work – but truth to tell it’s not that inter­est­ing ana­lyt­i­cally. It’s usu­ally just cherry-picking – because the best oppor­tu­ni­ties for this kind of inte­gra­tion are obvi­ous. Ana­lyt­i­cally, I con­tinue to believe that one of the most inter­est­ing ana­lyt­ics projects we do is full behav­ioral seg­men­ta­tion. Typ­i­cally, we take a full Omni­ture data feed for a month or two and then build behav­ioral pro­files of every vis­i­tor in tools like SPSS or SAS using clus­ter analy­sis. When we can, we’ll also inte­grate online sur­vey data. This type of behav­ioral pro­fil­ing is fas­ci­nat­ing work – but I also think it’s gen­uinely pow­er­ful. It’s the best way I’ve found to make web ana­lyt­ics data actu­ally come alive for marketers.

Q: You’ve recently been writ­ing about appli­ca­tion mea­sure­ment and how that’s dif­fer­ent than Web Ana­lyt­ics. As more apps move into the cloud, what should devel­op­ers be think­ing about in terms of mea­sur­ing their applications?

A: I wrote five long blogs on this — so it’s a chal­lenge to shrink it down to a bite-sized answer. But here are a cou­ple big-picture things. First, you’ll find that mea­sur­ing appli­ca­tions requires a pretty fun­da­men­tal shift in mea­sure­ment think­ing. The basic web site mea­sure­ment stuff (pages,
clicks) really don’t apply. Instead, you need to think about cap­tur­ing func­tional usage, appli­ca­tion states, and per­for­mance infor­ma­tion. Unfor­tu­nately, you still need to trans­late this par­a­digm back into some­thing that works in the ana­lytic solu­tion and that can be pretty dif­fi­cult. It’s also impor­tant for devel­op­ers to real­ize that mea­sure­ment inte­gra­tion takes real plan­ning and test­ing cycles — and there aren’t sim­ple auto­mated solu­tions for test­ing. So it’s vitally impor­tant to inte­grate the mea­sure­ment into the early stages of devel­op­ment and build a care­ful test-plan — oth­er­wise you’re likely to end up leav­ing most of the impor­tant mea­sure­ment on the app cutting-room floor.

Q: What advice would you give to a new Omni­ture Developer?

A: If you’re pick­ing an envi­ron­ment, PHP is the best sup­ported and doc­u­mented. Def­i­nitely start with one of the sam­ple pro­grams in the gallery – obvi­ously you should pick one from your envi­ron­ment. I find it’s just a lot eas­ier to get started when you can begin by mak­ing tweaks to exist­ing code that com­piles and works. I also think it’s worth start­ing with the Site­Cat­a­lyst Report­ing APIs – or at least under­stand­ing them – even if you’re focused else­where. I find we end up using these even when we are doing an appli­ca­tion mainly focused on some­thing else (SAINT for instance). Site­Cat­a­lyst is still at the heart of the envi­ron­ment and those APIs are def­i­nitely worth under­stand­ing.