Analytics 2.0

Much more than just tagging

Analytics 2.0 - Much more than just tagging

The Data Driven culture componen for startups

use data wiselyTwo years ago I’ve been invited by Lorena Suarez to be part of the mentors that help Wayra’s startups improve their analytics skills. At the beginning I was a little confused since I normally work with big companies that not only have an information caos but in several cases they don’t have a data driven culture, which is normally the biggest problem. When the information began to abound big companies started a chase to get the best scientist/Analysts to solve the problem. Later on they began to notice that the problem was not getting solved and the confusion turned no them.

I’ve mentioned several times an interesting report from IBM called “Global CMO’s Study” in which they pointed out that the biggest problem the CMOs where facing is the amount of data, “we are drowning in data”. The problem is not getting the better professionals but changing the company’s culture. Most of the biggest companies today were born in a time where information was very scarce, so is normal expecting that they don’t have the Data Driven Culture component into their culture DNA. The thing is that adding the data driven component to a current company culture is normally a work that takes time, a lot, that undoubtedly worth.

So as I was saying at the beginning, when Lorena Suarez called me with that idea I was little confused at the beginning but after hearing his idea everything became very clear. If we add the data driven component when they are building their culture then is gonna much more effective than doing it later. Indeed since information today is not as expensive as it use to be if those startups can take advantage of the information they can compete head to head with the industry leaders.

That was the project we begun two years ago with the people from Wayra and replied with other startups from different part of the world. The main important field in which we work is in:

Avoiding inferences: As mentioned in my book Meta Analytics, people tend to infer. You might say, but what’s wrong with infer? Inference is the act or process of deriving logical conclusions from premises known or assumed to be true. The problem is the “Assumed” word. The world is going very fast today and we have never time to have “known premises” so we just “asume” them. How? Basically we asume that any comment from a friend or an expert, note in a newspaper, post in a blog, tweet or a facebook post is information. TODAY NOBODY EVEN CARE THE METHODOLOGY WITH WHICH THE INFORMATION WAS GENERATED. Methodology is the most important thing because is the way we know if we can use or not that source. Are you going to make a decision based on a report with an uncertainty of 40%? Or even worst a source that you can’t even calculate the uncertainty? Of course not.

Avoiding inferences

However that’s exactly what people do today. If I see a post in facebook I don’t have time to analyse if it is correct or not, so basically I “believe” is true because “sounds” like it is true, or just don’t “believe” on it because sounds weird (to me).

Inferences are normally based on “Assumed” premises and that’s why it brings a lot of troubles. It brings trouble to all the decision makers not only entrepreneurs. So what we do is give them the tools to identify and avoid inferences. We are not even talking about processing or analysing information until now.

Top 5 inferences an investor/mentor hear from an entrepreneur during the first 5 minutes:

1. We don’t have competition/our product is unique.

2. People like / People don’t like.

3. Everybody knows that…

4. We can get this done in a week.

5. We have to improve our product/launch this features.

As you can see, with information all the above assumed premises can be solved.

How to make a self explaining reports

Aristotle_Bust_White_Background_TransparentThe Digital Analytics Association defines Digital Analytics as the measurement, collection, analysis and reporting of internet data for purposes of understanding and optimising digital usage.

It’s not a minor thing that analysis and reporting are separated. Analysis and reporting are two really different things.

1. Analysis: Is the process of breaking a complex topic or substance into smaller parts to gain a better understanding of it. The technique has been applied in the study of mathematics and logic since before Aristotle (384–322 B.C.), though analysis as a formal concept is a relatively recent development.

The word comes from the Ancient Greek ἀνάλυσις (analusis, “a breaking up”, from ana- “up, throughout” and lysis “a loosening”).[2]
As a formal concept, the method has variously been ascribed to Alhazen,[3] René Descartes (and the contemporary philosopher René Dechamps :-) ), and Galileo Galilei. It has also been ascribed to Isaac Newton, in the form of a practical method of physical discovery (which he did not name or formally describe). Analysis definition here.

2. Reporting: A report or account is any informational work (usually of writing, speech, television, or film) made with the specific intention of relaying information or recounting certain events in a widely presentable form. Written reports are documents which present focused, salient content to a specific audience. Reports are often used to display the result of an experiment, investigation, or inquiry. The audience may be public or private, an individual or the public in general. Reports are used in government, business, education, science, and other fields.

Reports use features such as graphics, images, voice, or specialized vocabulary in order to persuade that specific audience to undertake an action. One of the most common formats for presenting reports is IMRAD: Introduction, Methods, Results and Discussion. This structure is standard for the genre because it mirrors the traditional publication of scientific research and summons the ethos and credibility of that discipline. Reports are not required to follow this pattern, and may use alternative patterns like the problem-solution format. Report definition here.

As you can see, in the above definitions there are to important points.

The analysis definition says that in the Analysis process it’s one person breaking up the information in smaller parts trying to identify insights, with his mental model. In that process the person generates constants questions and answers that defines the final the context in which some identified insight make sense. I call that the P.I.S. or personal information sense. The biggest challenge that an Analyst have is to transfer the P.I.S. to the reported person through the report. In the Report definition it says that you have to present the report with the method, however the P.I.S. is not the method. We can identify the P.I.S. like the mental process than with methodology.

43824002

Why is this important? I witness many times the presentation of a report by, in the best case, the analyst or, in the worst case, by another person (the cool person that presents reports to the client, because the “nerd” should not be in touch with clients (???)). Most of the cases I was not able to understand what was that slide trying to say or how the person was relating that “insight” with the shown information. The answer is that the context is given by the P.I.S. that in most cases is not transferred from the analyst to the reported person.

To do this you can use a simple method. Analytics is part of the logic  (from the Ancient Greek: λογική, logike) which describes the use of valid reasoning. So the P.I.S. can be transfer using the Logic methods in each slide (that should transfer one clear and useful insight). One of those methods is the syllogism (Greek: συλλογισμός – syllogismos – “conclusion,” “inference”) is a kind of logical argument in which deductive reasoning is used to arrive at a conclusion based on two or more propositions that are asserted or assumed to be true.

There are infinitely many possible syllogisms, but only a finite number of logically distinct types, which we classify and enumerate below. Note that the syllogism above has the abstract form:

Major premise: All M are P.
Minor premise: All S are M.
Conclusion: All S are P.

The premises and conclusion of a syllogism can be any of four types, which are labeled by letters as follows. The meaning of the letters is given by the table:

code quantifier subject copula predicate type example
a All S are P universal affirmatives All humans are mortal.
e No S are P universal negatives No humans are perfect.
i Some S are P particular affirmatives Some humans are healthy.
o Some S are not P particular negatives Some humans are not clever.

In Analytics, Aristotle mostly uses the letters A, B and C (actually, the Greek letters alphabeta and gamma) as term place holders, rather than giving concrete examples, an innovation at the time. It is traditional to use is rather than are as the copula, hence All A is B rather than All As are Bs. It is traditional and convenient practice to use a, e, i, o as infix operators so the categorical statements can be written succinctly:

Form Shorthand
All A is B AaB
No A is B AeB
Some A is B AiB
Some A is not B AoB

So each slide can transfer the P.I.S. by using, for instance, syllogisms that explains the valid reasoning that we are transferring in each slide and which will help the reported person understand why you are saying that.

The valid reasoning should be supported by valid statistical models. So you can say, every time we show product B in the conversion funnel of product A the ticket value increases 20% (Square R 0.8, anova model), so if we increases the product B x% more in the same process, we will get a X% higher average ticket value with a confidence level of 95%. In this example, we are using a valid reasoning in parallel with a valid statistical method and we know the risk we are assuming by making that decision (increasing X% the appearance of product B).

Marketing is about people buying intention. Conjoint Analysis 101.

understanding-human-brain-neuroscience-tell-enough-satel-marcus

We’ve already talked that companies have one main objetive, earning money. If we add the variable ‘t’ we can split this main objetive in two, present and future earnings. The present earnings is the Company’s current economic result, while the future earnings are determined by the buying intention.

On the other hand the above mentioned buying intention it’s the result of several variables that occurs together like:
1. The person have the need or intention to buy something in particular.
2. The company have the product that can satisfy the generic need.
3. The person have the money to buy the product.
4. The brand communicates something that make the person feel special.
5. The company service makes people feel special.

In few words, the product/brand have a balance of attributes that makes a person want it. The buying intention is based on personal experiences and marketing communications.
That’s why it’s important to know what combinations of attributes and their magnitudes are the closest one to satisfy an specific type/segment of client.

Conjoint Analysis is a statistical technique used in market research to determine how people value different features that make up an individual product or service. The objective of conjoint analysis is to determine what combination of a limited number of attributes is most influential on respondent choice or decision making. A controlled set of potential products or services is shown to respondents and by analyzing how they make preferences between these products, the implicit valuation of the individual elements making up the product or service can be determined. These implicit valuations (utilities or part-worths) can be used to create market models that estimate market share, revenue and even profitability of new product designs or services.

Conjoint Analysis steps:
1. Desegregate the product or service into attributes. For example in flight tickets some attributes are price, if the ticket is refundable, if there is any cost on changing the departure date, etc. If we are talking about laptops those attributes can be price, screen size, ram memory, hard drive, weight, time of battery, etc.
Each attribute can be broken in levels like regular hard drive or flash memory.
2. Then a range of products (with different attribute combinations) are shown to the respondents in different ways. It can be the product (or service) it self, a picture of it, a prototype, etc.
3. Respondents rank the products or services.
As the number of combinations of attributes and levels increases the number of potential profiles increases exponentially. Consequently, fractional factorial design is commonly used to reduce the number of profiles that have to be evaluated, while ensuring enough data are available for statistical analysis, resulting in a carefully controlled set of “profiles” for the respondent to consider.

help-179-1

Get it? Good. Now imagine the conjoint analysis potencial with Social Analytics or pols in your website (Integrated with you web analytics platform). You can release to the internet several images or competitive products and do conjoint analysis by classifying the comments left by the internet (Facebook, twitter, etc) users and generate a product with the set of attributes and magnitudes that fit best your market requirements with a plus advantage.

The regular conjoint analysis is based on a short list of attributes that you or a previous market research defined. With social media or with pols you can define the set of attributes based on your clients requirements and then select the magnitude and level of attributes that will become the best option for your client.

You can use almost real time conjoint analysis for:

1. Flight tickets.

2. Mobile phones.

3. Phone plans.

4. Financial services.

5. Retail (delivery, quality of products, price, etc).

6. And even for web analytics platforms (real time, olap cubes, custom variables, integration with other platforms, free vs paid, etc).

Information Chaos? Analytics Governance is your only salvation

Most companies were founded in a time when there was a shortage of information. Getting the information was extremely expensive and not very useful since it normally came along late (normally just to justify a decision what was already made) . Today getting information is not expensive any more. Actually you have information everywhere and almost in real time for a very low cost. This avalanche of information instead of improving the decision making process is making it harder or in most cases impossible.

market research

Today’s managers and executives say that they are drowning in data. Most of they already attempted to solve the problem by hiring better prepared resources in matters like Big Data, Statistics, Analytics, Data Mining without actually achieving any interesting result. All the mentioned unsuccessful efforts focused on improve the decision making processes in a world with information overload ended up generating a big frustration.

Big Data

The thing is that the problem is not based on skills but in the culture. Companies born a world where the information was expensive and low. So those companies where not generating the Data Driven Culture component into their DNA. Companies are systems, and as such they have “parts” that interact each other with a main objective, those interactions generate information that tells you why your company is achieving or not its goals and how to improve that situation. Which means, which “part” of the system is not interacting properly (bottle neck) preventing the company as a whole to achieve a better result. Some of those parts are “Humans”, people that are interacting each other “thinking” that they are improving the company situation.

Digital Analytics

However not all the good intentions generate great business results, which explains why hiring “some” big data genius will not fix your problems (I mean, not just that). Is not about how good you analyse the available information but how much of the information from your system interactions (the ones that helps you understand which resource/resources are preventing the company to reach or improve its results) is available. A data driven culture requires that each resource from the company shares the part of “reality” that creates in the company.

Decision making scenario = Σ information(interaction 1) + information(interaction 2) +  information(interaction x)

The above mentioned formula tells you how important is sharing the information. Normally that information is kept unintentional hidden by each employee in a paper notebook next to their computers.

Data Driven Culture

The solution is adding to the company’s culture the Data Drive Culture component to their DNA. That’s not something that you do changing all the people in your company but generating a properly environment where people can develop their skills understanding what that means in the current company structure.

Analytics structure

The company’s objective must be so clear that all people’s activities should interact creating like the steps of the stair that ends in the business objetive. The biggest challenge is coordinating all those activities. Analytics Governance are the is the set of activities focused on understanding which resources are part of the Company’s system, design the measurement systems and generate a cyclic process towards the improvement of the decision making processes, making clear which resources are not bringing the company closer to its objetive and why.

Analytics goal

Those companies that have an Analytics Governance plan are those that will make the difference. Regarding the new companies, I highly recommend to starting up with Analytics Governance activities, so their culture DNA will have from the beginning the Data Driven component :-)

Data Driven culture

HOWA Buenos Aires 2013!

Hands on Web Analytics

HOWA Buenos Aires

Organised by Intellignos, December 10th at the headquarters of the Faculty of Economics of the University of Palermo, the seventh edition of
HOWA Buenos Aires, the most important analytics event of the year.

Register Here for free

When? The 10th of December 2013 from 9am to 6pm.

Where? Universidad de Palermo, Larrea 1079, Ciudad de Buenos Aires.

Schedule

9 a 9.30 hs Presentación y repaso de temas del año – Auditorio
Information Gandalfs Analytics Ninjas
9.35 a 10.20hs Cross device Tracking  Cross device análisis
10.25 a 11.15 hs  Herramientas de usabilidad  UX Analytics
11.20 a 12.05 hs Coffe Break
12.05 a 13.00 hs eCommerce Tracking Análisis y optimización de conversiones
13.05 a 14.05 hs Armado de sistemas de información integrados – Auditorio
Networking / Libre
14.10 a 14.55 hs Cross device Tracking Cross device análisis
15.00 a 15.45 hs Herramientas de usabilidad UX Analytics
15.45 a 16.40 hs Coffe Break
16.45 a 17.30 hs eCommerce Tracking Análisis y optimización de conversiones
17.35 a 17.55 hs Cierre del evento – Auditorio

Google Analytics Summit 2013

As every year I’m here, in this case with Richard Dawson (Intellignos) and Diego Salama (Mercado Libre) at the Google Analytics Summit, in this case in it’s 2013 version.
This post is in real time so you will see things in draft and not finished…don’t worry, will be done after the event is finished.

Google Analytics Summit 2013

Great introduction by Paul Muret, Vice President of Engineering at Google Analytics talking about history and the Analytics Market.
Paul gave the voice to Bebak Pahlavan the Director of Product Management of Google Analytics who after saying that they launched more than 70 new feature, will introduce 14 new Features:
1. Auto-event tracking en Google Tag Manager.
2. Premium Service Level Agreement for Google Tag Manager.
3. Upgrade to Universal Analytics for the standard accounts!
4. Management UI and API.
5. New ABC report.

ABC Report Google Analytics
6. New Unified Segments.

Google Analytics Unified Segments

7. Audience Data and reporting!

Google Analytics audience analysis
8. Audience data within unified segments!
9. Export GA hit data into google big query for premium customers
10. Double click campaign manager integration – view through, click through data.
11. Double click data import into Multi channel funnels.
12. Google Play integration with GA Analytics to analyze the impact on downloads.

Plays with Analytics
13. Analytics Academy.

Oct 1: Course opens for registration!
Oct 8: Units 1-4 will be available, Google Group opens for discussions.
Oct 15: Live Hangout #1, Units 2-6 open for access.
Oct 22: Live Hangout #2
Oct 30: Course closes, get your certificate by this date!
14. In-Product help videos

SDX – more granular and complex querying of unsampled data
The upcoming BigQuery integration is a planned feature for Google Analytics Premium that allows clients to access their session and hit level data from Google Analytics within Google BigQuery for more granular and complex querying of unsampled data. For those unfamiliar with Google BigQuery, it’s a web service that lets you perform interactive analysis of massive data sets—up to trillions of rows. Scalable and easy to use, BigQuery lets developers and businesses tap into powerful data analytics on demand. Plus, your data is easily exportable.

Google Analytics big Query

APIs For Enterprise
Large companies have unique needs; they have many websites and many users. In the past, it could take many hours to setup Google Analytics. With our new Google Analytics Enterprise APIs, IT teams can programmatically setup and configure Google Analytics accounts, saving time, and giving them more time to analyze data.

 

 

After the break Tom Davenport, Professor, Author and Senior Advisor to Deloitte Analytics is presenting Marketing Analytics 3.0. Speaks how old is Big Data and ask to the attendants “Who would anytime say I work with small data?” :-)

Tom Davenport at GA Summit
Marketing Analytics 2.0 is the big data era which is
1. Complex, large, unstructured data about customers
2. New analytical and computacional capabilities.
3. “Data scientists” emerge
4. Online and digital marketing firms create data-based products and services.

Marketing Analytics 3.0
Fast, Pervasive Digital marketing:
1. A seamless blend of traditional analytics and big data
2. Analytics integral to marketing and all other functions
3. Rapid, agile insight and model delivery.
4. Analytical tools available at point and time of decision
5. Analytics are everybody’s job
6. Heavy reliance on machine learning “However we are very sceptical about it’s potential”
7. In-memory and in-database analytics.
8. Integrated and embedded models.
9. Analytical apps by industry and decision.
10. Focus on data discovery.

GE has been creating the new analytics and industrial internet model and invested 2 Billon on that.

“75% of marketers don’t know their ROI”

Recipe for a 3.0 world
1. Start with an existing capability for marketing data management and analytics.
2. Add some unstructured large volume customer data.
3. Throw some product/service innovation into the mix.

Now is the turn of Russell Ketchum Product Manager at Google Analytics In-app measurement: Going native.

Conversions, are they doing what matters?

[In Apps]: when you’re looking at behavior metrics, you’re looking at what people did / how they’re using your app.

Acting on an idea

Users should spend time with their data

The drawer is the fastest way to data

Improving the drawer helps users get to data

LUNCH TIME!

Lunch time at GA Summit

At 13.40hs cames along Jody Sarno, Customer Insight Senior Analyst at Forrester to talk about solving marketing challenges: How attribution can help.

Clear up the confusion. Marketing Mix modelling (MMM) is the process of using statistical analytics to estimate, optimise and predict the impact of paid, owned media.

Key Findings

1. Marketers leverage attribution to uncover marketing and consumer trends.

2. Opportunities. Data integrations, change management and customer purchase path.

End of Jody conference

While the next speaker begins you can take a look and register at the brand new Google Analytics Academy.

Now Bill Kee, Head of Attribution Products at Google Analytics talks about how to make attribution works.

  • In 2011 multichannel funnels
  • In 2012 Attribution Modelling tool
  • In 2013 Data Driven Attribution

There are two important new integration, the first is Youtube Display Network with Google Analytics and the second is Double Click with Google Analytics allowing to understand the full clickstream (user journey) of a user.

Data driven attribution model.

Calculate the impact of each touch point. “All models are wrong! But some are useful.” Bill Kee – Head of Attribution Products, Google.

Bill invite Melissa Shusterman, Strategic Engagement Director at MaassMedia to talk about a case study.

Melisa says that attribution allow them to optimise all displays campaigns an not just the ones that drives conversions…(sorry I don’t understand what she wanted to say).

1. Initial Analysis, click throughs.

2. View Throughs conversions, confusing. This contradicts click throughs. conversions.

Key areas of attribution:

1. Last Touch Sales.

2. Attributed Sales.

3. Percent Non-last touch sales.

4. Cost per attributed sale.

Results: The traffic decreased but the conversions with the Data Driven Model increased.

Making it work

  • Set up acurate weighted goals
  • Allow advertisements time to work
  • Dont’ think too small
  • Kill poor performers
  • Sell attribution-  Help display compete
“Data Driven Attribution forces Display teams to get smarter” 

Next presentation is Steve Yap, Head of Emerging Products at Google. Required to win: The integrated analytics imperative.

Integrations thorough principle. Today’s market and todays consumer demand more from us. They want relevancy, engaging creative and meaningful content.

Principles.

1. Whatever we built has to be the best in the market.

2. The system have to work well with one other and be better together.

3. They are easy to use.

Progression toward action like Doubleclick, Teracent, Invitemedia and Google Analytics.

 

if you wanted to know what they are thinking just go and ask them

Multivariate or A/B testing is very cool and useful. Today’s solutions are very flexible and allows you to test several pieces of marketing (ads, lading pages, etc) in real time without the need of being an expert. You have tools for any particular need, free solutions like Google Content Experiments; paid solutions for a low price like Convert, Optimizely, Unbounce and VWO; and paid solutions for bigger budgets like Adobe Test & Target, Sitespect and Autonomy optimost among others.

Multivariate testingMultivariateTesting

Even though this kind of solutions are great and useful it is important to know what are they useful and are not useful for. A/B testing and Multivariate solutions are useful to test a “Short list” of changes or optimisations after you already know them. What they can’t do is let you know what are those changes you have to test. Let’s put it this way, if someone tells you “what do you prefer, that I give you a punch in your eye or in your stomach?” You will probably chose the eye or the stomach depending on what bother you less.  However that doesn’t mean that you like to be beaten.

Before running A/B testing or a multivariate testing I highly recommend you to run surveys to let your users chose the short list to be tested (without they knowing that). If you just show them just a few combinations that you have in mind, you could be leaving aside some other possibilities that can drive you in a better manner to your business result.

A/B testing or multivariate testing are a behavioural source of information. Behavioural information can just answer you “WHAT” the user is doing but not “WHY”. It is important to know what kind of information source can answer different questions so you can go there and get the answer. The “WHYs” can just be answered by Attitudinal information sources like surveys.

You have also a full range of surveys solutions both Free like Survey Monkey and Google Forms; and paid like Zoomerang. So, if you wanted to know what they are thinking just go and ask them, no excuses.

The US ICT Market optimistic about Analytics and BI investments

According to a report by Forrester Research the total Total U.S. ICT market in 2011 was $962B with the majority being generated from software sales ($208B) followed by Telecom Services ($199B) and IT Consulting and Systems Integration Services ($188B). The following graphic show the U.S. ICT market for 2011.

Information platform used

Forrester forecast for analytics and BI are more optimistic.  When customer experience and engagement is taken into account, the forecast seems high. Salesforce knows how to translate trial users into customers. The question is can they do this fast enough in 2012 throughout the enterprise and mid-tier accounts to keep up their sales growth on track while reducing churn and increasing profitability.
Investment in Analytics
Smart Computing is defined by Forrester as platform technologies including specialized analytics, BI, service-oriented architecture (SOA) infrastructure, virtualization software, rules engines, and awareness-based technologies. Forrester is very optimistic about this area with a growth rate second only to cloud computing. Its index of the market is based on Informatica, Pegasystems, and Tibco Software.

Information integration where are you?

The following chart presented by eMarketer shows the answers made by  the Forrester Conference attendees.

Information sources

The question was “What is the best source for generating data on their customers”, the answers are not a surprise at all. They answer as each source provides enough information itself about their customers, but they don’t! Web analytics just tells you what your customers are doing but how are you gonna make a decision when you don’t know why?