Verification of Session flow in Ecommerce – GTM and Google Analytics way

Let’s jump into Google Tag manager (GTM) wagon and contribute to solutions we came across our cases.

In our company culture is always present question on data quality we work with. Google Analytics data as primary storage point is crucial to decision making and data analysis. Problem we faced at one point in time was – how to detect breaking steps in sales funnel per user session? Google Analytics can give pretty nice information on selected data summarized and split into steps – but we needed to  understand more per user session. Beside this problem we had to be certain that data stored in ecommerce part is in line with internal sales system.

We can not guarantee that data stored in Analytics regarding sales (ecommerce) part is 100% correct, as sales process is not hosted on client servers and in complete client control – that we can analyze server logs and all process steps.

System in place, with multi step goals, different payment processors depending on currency used became serious challenge involving wide range of talents, tools and resources.

Main goals of GTM and GA setup was

  • identify steps in process where users stop or generate error (with complete trace per user and session and position in process)
  • verify  ecommerce data collected on thankyou page with processing system(s) information

Second goal concerning verification of GA ecommerce data and external system was finished with standard reporting set from GA. On the other hand, first request showed to be bigger issue.

Google Analytics is storage engine, that does computation of multiple metrics attaching them to specific dimensions. In this process happens sampling. If you have big set of data, in order for system to be responsive and fast – sampling must occur. We had to deal with specific people and their problems on daily basis in order to debug process and avoid sampling.

In order to be able to analyze anything we had to add additional information using session level custom dimension information. As GA is prohibiting storing any kind of personal information in GA we had to rely on unique session identification provided by server. Using this variable (and including it within data layer pushed to GTM) we were able to follow specific user and his flow through site, identifying steps in sales funnel where user stops or goes further.

On top of other data layer variables – user language, site language, location and pages in sales funnel we were able to detect faulty scenarios in sales process.

In terms of outcome – detecting faulty scenarios in sales funnel worked flawlessly giving us and developers insight where to look for breaking points. This lead to cutting development cycles from more than 24 hours (mostly waiting for processing logs) to less than 2 hours.

For second problem – quick integration with SuperMetrics for Google Docs solution and active follow up on client side provided insights in quality of data stored in Analytics. At the end small difference occurred between Analytics and sales system, showing that root cause of difference exist due to nature of processing system – even if transaction information sent to website cleared as OK in next few minutes payment processor can decline payment and cancel sales.

If you need more details or more graphic explanation of process – respond :).

Google Analytics – Model dodeljivanja (Attribution Model) – drugi deo

U prvom delu smo odgovorili na neka jako važna pitanja, prvenstveno na pitanje šta je to model dodeljivanja. Napravili smo razliku između običnih konverzija i konverzija u Multi Channel Funnel izveštajima. Analizirali smo putanje konverzija i definisali vrste interakcija.

U drugom delu ćemo se baviti pitanjima poput vrsta attribution modela i odgovorićemo na pitanje kako izabrati odgovarajući Attribution model. Videćemo kakve zaključke možemo izvesti iz različitih odnosa vrsta konverzija.

Vrste Attribution modela

Attribution modeli se generalno mogu svrstati u dve kategorije :

  • Osnovni
  • Prilagođeni

Osnovni Attribution model

Osnovni Attribution model definiše distrubuciju interakcija (ili dodirnih tačaka ) na putu do konverzije. Postoji nekoliko osnovnih tipova Attribution modela:

  • Last interaction attribution model – Ovaj model dodeljuje 100% kredita poslednjoj interakciji. Google Analytics koristi ovaj model po defaultu.
  • First interaction attribution model – Ovaj model dodeljuje 100% kredita prvoj interakciji.
  • Linear attribution model – Ovaj model dodeljuje jednak kredit svakoj interakciji na putu do konverzije.
  • Time Decay attribution model – Ovaj model dodeljuje više kredita interakcijama koje su vremenski najbliže trenutku ostvarivanja konverzije.
  • Position based attribution model – Ovaj model dodeljuje 40% kredita za prvu interakciju, 20% kredita za srednju interakciju i  40% kredita do poslednje interakcije .
  • Last non direct click model  – Ovaj model dodeljuje sve zasluge za konverzije svim interakcijama do poslednje koja nije direktni klik na putu konverzije.
  • Last Adwords Click – Ovaj model dodeljuje sve zasluge za konverzije poslednjem Google AdWords kliku.

Napomena : izbor attribution modela uglavnom zavisi od poslovnog modela klijenta i ciljeva oglašavanja.

Prilagođeni Attribution model

Kao što samo ime kaže ovi modeli su razvijeni od strane ljudi poput nas. Kada izgradite sopstveni attribution model, možete da kreirate sopstvena pravila za dodeljivanje kredita različitim interakcijama na putu do konverzije. Ova pravila su poznata pod imenom “custom credit rules” u Google analitici.

Kako da napravite svoj Attribution model u Google Analitici

Ako imate pristup Google Analitici, pratite sledeće korake kako biste kreirali sopstveni Attribution model.

1. Idite na Model Comparison Tool (ispod Conversions > Attribution na vašem Google Analytics nalogu)

2. Iz padajućeg menija ‘select model’ izaberite ‘create new custom model’.

create new custom attribution model

3. Imenujte svoj model i izaberite osnovni model (linear, time decay, position based). Podesite kredite za konverzije, odredite lookback window. Ova pravila definišu kako treba da se kredit svake interakcije distribuira na putu do konverzije. Svako pravilo se zasniva na jednom ili više uslova.

create or edit attribution model

Napomena: Model Comparison tool vam omogućava da kreirate i primenite prilagođene attribution modele. Takođe vam omogućava da uporedite do tri modela istovremeno.

Kako da izaberete Attribution model

Attribution model birate na osnovu poslovnog modela i reklamnih ciljeva vaših klijenata. Attribution model koji izaberete ima veliki uticaj na obim i vrednosti konverzije. Tako oni u velikoj meri utiču na vrednost vaših marketinških kanala. I obim konverzija i vrednost konverzija mogu varirati od jednog do drugog Attribution modela.

Na primer, ako izaberete ‘First interaction attribution model’ onda se svim kanalima marketinga koji iniciraju konverzije pripisuje visoka vrednost konverzije. Ako izaberete ‘Last interaction attribution model’ onda će svim kanalima marketinga na kojima su završene konverzije biti pripisane konverzije sa visokom vrednošću.

Slede neka opšta pravila koja možete primeniti dok odlučujete koji Attribution model treba da izaberete:

  • Ako ste novi igrač u vašoj niši onda vam treba veća svest o brendu nego vašim direktnim konkurentima. Posledično, vaši ciljevi oglašavanja će biti više orijentisani na građenje brenda. Dakle, potrebno je da dodelite veći prioritet interakcijama koje iniciraju proces do konverzije. Iz tog razloga, First interaction attribution model je pogodniji za vas.
  • Ako imate poslovni model gde je podjednako važna svaka interakcija na putu do konverzije onda je Linear attribution model je prikladan za vas. Na primer, ako pokrenete uslugu podrške klijentima onda je svaka interakcija sa klijentima jednako važna za vas. U tom slučaju možete da koristite linearni model interakcije.
  • Ako prodajete proizvode kao što su FMCG (Fast Moving Consumer Goods) koji podrazumevaju najmanji nivo razmatranja od strane kupca, onda je Last interaction attribution model  prikladan za vas, jer ne morate da dodelite više značaja prvim i srednjim interakcijama u vašem putu do konverzije.
  • Ako imate poslovni model ili ciljeve oglašavanja koji vrednuju prve i poslednje interakcije više od srednjih onda je Position based attribution model pogodan za vas.
  • Ako želite da razumete motive kupovine vaših klijenata tokom promotivne kampanje onda biste želeli da dodelite više značaja interakcijama koje se javljaju u najkraćem vremenu konverzije, jer su više relevantne nego interakcije koje su se dogodile pre nekoliko dana. Posledično, Time Decay attribution model  je više prikladan za vas.
  • Direktan saobraćaj ne predstavlja izvor saobraćaja niti marketinški kanal. Tako da, direktnom saobracaju ne bi trebalo dodeliti značaj na putu do konverzija. Posledično, treba da koristite Last non direct click model koji daje sav značaj konverzijama do poslednjeg indirektnog klika na putu do konverzije.
  • Koristite poslednji AdWords click model, ako želite da date Google AdWords-u veći prioritet za konverziju.

Vrednosti konverzije

U multi channel funnel izveštajima postoje 3 vrste konverzija:

  1. Assisted Conversion Value
  2. Last Interaction Conversion Value
  3. First Click conversion value

Assisted Conversion Value je ukupna vrednost ostvarena pomoću asistiranih konverzija . Što je veća vrednost ovih konverzija to je važniji marketinški kanal u asistiranju na putu ostvarivanja konverzije.

Last Interaction Conversion Value je ukupna vrednost povezana sa poslednjom interakcijom pre samog ostvarivanja konverzije. Što je veća vrednost ovih konverzija to je važniji marketinški kanal pri samom završetku proseca ostvarivanja konverzije.

First Click conversion value je ukupna vrednost povezana sa prvim klikom u prosecu ostvarivanja konverzije.  Što je veća vrednost ovih konverzija to je važniji marketinški kanal u pokretanju procesa konverzija.

Kako se Assisted / Last Interaction Conversion izračunava i kakve zaključke možete iz toga izvesti?

Ovaj odnos se izračunava kao :

= Broj asistiranih konverzija / Broj Last Interaction konverzija

  • Ako je vrednost ovog pokazatelja blizu nule onda to znači da marketiniški kanal najviše doprinosi na samom kraju puta do konverzije, pre svega u završetku konverzije.
  • Ako je vrednost ovog pokazatelja blizu 1 onda to znači da marketiniški kanal jednako doprinosi u oba slučaja –  pri asistiranja i pri završetku konverzija.
  • Ako je vrednost ovog pokazatelja više od 1 onda to znači da marketiniški kanal pre svega doprinosi pri asistiranju konverzija.

Kako se First / Last Interaction Conversion izračunava i kakve zaključke možete iz toga izvesti?

Ovaj odnos se izračunava kao :

= Broj asistiranih konverzija / Broj Last Interaction konverzija

  • Ako je vrednost ovog pokazatelja blizu nule onda to znači da marketinški kanal najviše doprinosi na samom kraju puta do konverzije, pre svega u završetku konverzije.
  • Ako je vrednost ovog pokazatelja blizu 1 onda to znači da marketiniški kanal jednako doprinosi u oba slučaja –  pri pokretanju i pri završetaku konverzija.
  • Ako je vrednost ovog pokazatelja više od 1 onda to znači da marketiniški kanal pre svega doprinosi pri pokretanju konverzija.

Ovim dolazimo do kraja priče o Attribution modelu. U nekim od narednih postova bavićemo se temama kao što su Campaing Source Attribution, Cost Data import i šta oni predstavljaju. Videćemo na konkretnom primeru kako upotreba različitih marketinških kanala doprinosi većoj vrednosti konverzija i kako pravilna analiza konverzija pomoću Multi Chanel Funnel izveštaja daje pravu sliku o povraćaju uloženih sredstava kroz različite marketinške kanale.

 

 

Problemi u merenju srpskih medijskih sajtova

Dragi medijski sajtovi u Srbiji… imate problem (u većini slučajeva).

Napomena: Huge Medija nije ni u kakvim ugovornim obavezama sa bilo kojim od spomenutih sajtova.

Pre nekoliko meseci (tačnije kraj novembra 2013-te) na Blogomaniji 2013 želeo sam na radionici posvećenoj analitici – predstavim probleme koje mogu zadesiti sajtove svake namene, strategije koje mogu da se preduzmu onda kada pogledamo analitiku (podsećalo je na šahovsku simultanku),  podsetim prisutne kako da izađu iz problema koji se zove srpski sajtovi unutar Chrome-a koje Google detektuje kao hrvatske, kako domaći medijski sajtovi ne prate analitiku kako bi trebali i zašto su u problemu… Ovaj članak se vraća na poslednju stavku – problem u kome se nalaze medijski sajtovi.

Šta se promenilo za 3 meseca? Blic je promenio način merenja tj unapredio je postojeću analitiku. Bravo. Ali to je to. Niko ništa.

U nastavku je pregled nekoliko najvećih sajtova kao i implementaciji Google Analitike na istim (zaključno sa 1.3.2014.).

Dodatak: lista je povećana sa još 8 sajtova koji imaju bitniju posetu (hvala na komentarima)

Implementacija koda za praćenje poseta sa Google Analitikom - najposećeniji informativni portali

Implementacija koda za praćenje poseta sa Google Analitikom – najposećeniji informativni portali

Šta možemo zaključiti iz ove slike ili tabele sa pregledom načina implementacije?

Medijske sajtove (uglavnom) ne zanima publika koja ih posećuje. Svetli primeri su NaDlanu, Novosti (uz čudnu implementaciju zajedno sa Gemius kodom za praćenje) i Blic. Zašto? Implementacijom naprednog Google Analytics koda obezbedili su svojim marketing menadžerima ili predstavnicima prodaje uvid u demografiju svojih posetilaca.

Šta se dešava sa ostalima? Kod za praćenje je jednostavno stavljen kako bi davao osnovne podatke o posetama. Pitanje je i kako se podaci i koriste. RTS čak ima prastati JavaScript kod za praćenje koji uskoro više neće biti validan.

Zašto postavljam pitanje o tome kako se koriste podaci koji itekako mogu pomoći vlasnicima sajtova da bolje shvate svoju publiku i načine monetizacije svog sadržaja

? Zato što su ovi podaci na raspolaganju svakome, besplatni, zavise od profilisanja posetilaca i njihovog prethodnog ponašanja a ne upitnika koji treba da popune. Zato što dobijanje ovih informacija zahteva tako malo truda. Zato što su pored demografije dobijamo i interesovanja i teme. Moćno zar ne? Ali, kao što vidimo, niko ne koristi.

Poslednjih 8 meseci interfejs Google Analitike vrišti o tome da je moguće videti demografiju i interesovanja posetilaca web sajta jednostavnom promenom koda na web sajtu. U pitanju je sledeća promena.

Neophodna promena Google Analytics koda za pobedu

Neophodna promena Google Analytics koda za pobedu

Da li su sajtovi spremni za prelazak na Univerzalnu Analitiku? Prema svemu gorenavedenom – teško, sem nekoliko svetlih primera.

A šta je Univerzalna Analitika – poslednja verzija Google Analitike, sa novim (boljim) načinom merenja, koja će postati jedina moguća – nakon migracije starih podataka. Pitanje je da li su podaci koji sada postoje dobri. Kakva je svrha migracije podataka u novu formu ako stari sadržaj nije kvalitetan.

Naravno – ono što ne vidimo je kako su nalozi za analitiku namešteni i kako primaju podatke sa gorenavedenih sajtova. Prateći iskustvo i situacije na terenu – videli smo previše stvari. Zaključke i probleme kao i rešenja smo opisali u prethodnom članku (engleski).

Promenite ili adaptirajte svoje kodove – iznenadićete se novim kvalitetom informacije koje možete dobiti i kako vaš marketing može bolje razumeti šta redakcije rade i to bolje prodavati.

Vidimo se uskoro na netu. Uskoro.

Misbehaviors in Google Analytics

Often professionals working with Google Analytics tend to forget the basics of implementation. Not that they are bad at this – but most of the time they don’t think at all of problems that can happen during the website development, developers gone wild… and they tackle problem at the end – when their service is called upon.

Lets try for beginning, to set things up from the start – before professionals take over and do their magic. Data of website owners/companies should be OK. First step in setting things right is proper implementation of tracking code.

Working with programmers and development teams, it came to my knowledge that most of the time – tracking code implementation is just a line in their checklist. No real thinking about what has to be there, which code to use etc… But this line item happen all the time and developer habits die hard. Team or programmers modulize implementation and their only concern is to place tracking ID. No mention of changes in tracking code or methods for collecting data.

At this moment in time (end of February 2014) we have 2 possible ways – Classic and Universal. Stated in Developers part of Google Universal Analytics Upgrade Center in Phase 3 Universal Analytics will be Out of Beta and it will  consist of all beautiful stuff we have in Classic Analytics (remarketing and support for display advertising (demographics is already implemented)).

However, nothing is stoping you or any other website owner to have this data already. Most of “old” websites use some sort of implemented GA (for Google Analytics), either as Sync or Async version, with or without Display Advertising support (so called dc.js implementation) or just plain Universal Analytics code (analytics.js).

But why this blog post?

Developers do their job and move on. Implementation of GA is no simple task and so many things can go wrong. We have sevelar point of failure and we will cover it below. In process of implementation there are several things

  1. creation of property (either Universal or Classic)
  2. adjustments to code – extras (optional)
  3. placement of final code on template pages at the right place
  4. verification of how things were implemented and do they work

Point 4 is solvable with Google Tag Assistant – nice little piece of extension for Chrome. It can detect various implementation ways, all problems and can suggest solution. When verification exist, Point 3 is easy to change if there is any problem. Which moves us to Point 1 (and 2) which is origin of problem.

How and what do you want to track?

Choosing Analytics property

Choosing Analytics property

Using Universal is a clear win as this is clear path of Google Analytics toward better understanding of traffic and analysis available for all. But until Phase 3 of Universal Analytics we can not expect to have remarketing (essential for great AdWords campaigns), demographics and interests (however we started noticing this data in Universal some time ago).

Anyway, all Classic properties created now or in past will be migrated with all data. Until then, one of the clear paths we like to do is to have 2 properties per website in order to have complete view on visitors and their actions and habits. Back to code that has to be implemented. Whichever type of analytics you choose, have in mind that developers must implement it properly. How does it look like?

Different codes for different properties

Different codes for different properties

Each code must be implemented in its basic form properly and at the right place. Take a look at difference between Classic and Classic with Display Advertising enabled – just one line of code, that makes great difference if you have Google AdWords campaigns and you analyze data and are in desperate need of demographics.

What is wrong than?

In our experience – 5 out of 10 clients we saw in the past month have Universal Analytics activated on property level but actual code on site is Classic. This is one problem.

We’ve seen also usage of Universal code with Classic property. It seems that someone jumped into Universal wagon without changing property type or upgrading. Anyhow, these things can be seen if you compare how you defined property and code implemented on site.

Also, we see wrong position of either Analytics code on web page. In order to verify that everything is OK there must be clear oversight of all parts from creation of property, implementation and checking if it is placed properly. We strongly encourage clients if they are at classic analytics (ga.js) implementation to move to async code (and check current status) and to add display advertising option  (dc.js) in order to understand their audience (and have great advertising).

In the future this property will migrate to Universal and there will be no harm d0ne. If clients have Universal Analytics we let them be, but advise to add additional code in order to take advantage of good remarketing (Classic code with Advertising support – dc.js implementation). And it is simple as that.

Control and understand – that we do not live in perfect world. Understand that developers with low priority tasks just put code where it should be without consulting with specialist. Accept the fact that there is so much data within Google Analytics that can explain a lot of business issues and help them solve.

But first of all – do you know was the implementation done right? We know that misbehaviors exist and that we have to deal with them at the start – not at the end.

— If you are ok with your current setup you should go further.

Superweek 2014

We’ve been to SuperWeek 2014 in Hungary, with support from Google, and it was a blast. In case you did not know, it is one of a kind conference dedicated to Web Analytics. Created by Bánóczy Zoltán and his company AALL, SuperWeek is a place to be if you deal with web analytics in any professional way, and in case you are living in Europe.

Tagline of conference is:

SUPERWEEK conference is a unique, anual European gathering of data evangelists, analysts and thought leaders of the Modern Web Industry.

I would say:

Place to learn and to share, place where you can see where to aim for, place to find and educate clients.

Being there is great investment and of great value. Not all things are publicly available (certain presentations and talks) and are limited to ones who were there.

My credentials

My credentials

The Speakers

All the people you know from their blogs, companies… the ones that matter. And certanly Avinash. Being The Star of this community, he was mentioned non stop, day before his arrival. There was almost worship moments with mentioning of his name. Certanly, Mr Kaushik is the person who is, based on his history and effort and results, the one who moves industry forward with his exposure and evangelizm of actionable measurement and involvement with clients. About his speach – little later.

Boys and gals gave us their overview of the market, tools, insights, strategies and overall – where do we move as industry and how to deal with clients (both internal and external).

By personal opinion, first day was packed with great sessions, and the last 2 days were just like cherry on the top. Great talks and exhange with speakers beside Avinash.

Content

Good people gathered all available presentations and data on Dropbox so feel free to use them (beside official one download link above). Some of them are usable only to people who were there as there is a big context behind presentations. This is one of the reasons to be next year in Hungary. Hear the people, ask the questions, interact.

The conference

It would be shame to take out certain speakers and their themes as it would be long list of lectures and presentations I attended, and I am afraid to miss someone. However, I can pinpoint themes and takeaways suitable for companies, like the one I represent.

Speakers at SuperWeek by  Søren Mortensen

Speakers at SuperWeek by Søren Mortensen

This region of Europe still does not have the agency of size like the one we’ve seen on SuperWeek, which is a problem. We can not relate to problems these kind of companies cope with. Most of questions going to speakers were oriented toward practical problems agencies and professionals have. Big companies have big problems, big headcount and clients that have big problems. We have less bodies in office, many clients and wrong ratio between quality and profit. To be more descriptive – we have problem puting price of our services.

As there is no mature market for this kind of services (in region), and the clients have just entered this land of web analytics – everything is very fragile and full of child-disease-like situations. To us – anything that produces value and not-so-obvious consulting or business measurement (read : how to put a price on new visitor) is new and strange, and never before seen. SuperWeek was and will be a place to educate ourself and clients, place to learn how to move forward.

The tools and services that make life easier – is important thing that I can take away from SuperWeek. Open discussion and panels with sharing of best practices – the best.

The venue

At first, mentioned mountain in Hungary (Galyatető) sounded like a bad idea. But being there – it is a great spot. Like a retreat, with no interuptions, distractions – total devotion to things that matter.

SuperWeek 2015 – see you there next year.