Datalicious Blog

We're hiring! Looking for a Coder & Systems Engineer with Go/Linux Skills ($60-90K)

Thu, 18 Apr 2013 00:30:00 -0700

Datalicious is a leading Sydney Data Agency with a number of exciting Big Data opportunities looking for a strong technical all-rounder with at least 3-5 years of systems development & UNIX experience to hit the ground running.

We are a fast paced data and analytics agency that is currently growing at a rapid rate.. We have a big list of blue-chip clients and provide our team members with the opportunity to learn new skills and gain experience in actual Big Data.

Our systems are written in Python, Go and Bash, but we are in active development and your approach to problems is more important than knowledge of specific languages.

You will

Get your hands dirty helping to maintain and improve existing products
Proactively present data-driven innovative solutions to our clients' problems
Learn new languages and systems where appropriate

Required skills

OS: General knowledge of Linux/Unix systems. Should have basic understanding of processes, threads, standard userspace utilities.
Shell scripting: Essential.
Languages: Must be proficient in at least one general-purpose open-source dynamic / scripting language (Ruby, Python, Perl, any lisp, etc. [1]) and at least one systems language (Java, C, C++, Go, Objective C, etc. [2]).
Databases: Understanding of RDBMS, preferably some experience with "NoSQL" systems as well. PostgresSQL, MySQL.

You should must have these attributes

Ability to balance active development and operational maintenance
Excellent attention to detail and a desire to understand systems at every layer of the stack
Excellent time management skills and be willing to multi-task
Be a self-motivated learner
Be very strong in at least one programming language

Desirable Skills

Systems Administration: Configuration management, remote execution, monitoring
Revision Control: Git, mercurial, SVN.
Infrastructure: Cloud IaaS platforms, particularly AWS.
Open Source: Ability to demonstrate contributions - small or large - to real code bases, involvement in FOSS communities, account on GitHub or LaunchPad, etc.
Linux Environments: Experience with the LAMP stack or composable systems in general.

Bonus Points!

You’ll get bonus points if you have done any work with concurrency, distributed systems, quorums, etc. Also a bonus if you have an interest in algorithms, 'Big Data', R, SPSS and/or have some stats knowledge.

How we will reward you

Competitive salary depending on skills and expereince
Exposure to a growing list of interesting blue chip clients
Flexible working hours in a dynamic team environment
Young start-up with a "work hard, play hard" company attitude
Training on industry leading analytics, marketing and big data platforms
Plenty of development and career opportunities in fast growing business
Great office space in the city that we share with other agencies
Free coffee and toast to kick-start your day

Don't forget to check us out on JobAdvisor, Facebook and LinkedIn to find out more about your potential team members and what they think about working at Datalicious.

If the above sounds interesting, please apply via Jojari (preferred) or email us at jobs@datalicious.com so we can have a look at your resume and arrange a quick initial phone interview to ask you a few questions before we meet for an interview.

Permalink | Leave a comment »

We're hiring: Looking for a Project Manager to join our team of data geeks in Sydney ($60-90K)

Thu, 18 Apr 2013 00:30:00 -0700

Datalicious is one of Australia’s leading data agencies. We’re currently looking for a Project Manager who has experience in delivering digital and technical projects and who understands what it takes to keep our clients and our staff happy.

This is a great opportunity for an experienced project manager/producer with 5+ years experience. The candidate should have experience in digital tech projects or data driven projects with a high level of complexity and numerous stakeholders.

Datalicious is a fast paced data and analytics agency that has started to develop its own products in order to help us grow faster, both in Australia and overseas. There is never a dull moment here at Datalicious and you will have the opportunity to spend your time working on a huge variety of projects - Big Data, Media Attribution and Web Analytics projects are just some of the types of projects we are working on for our clients. We’d also love to get you involved with some of our R&D projects and train you up in areas that are of interest to you.

Required skills

5 years experience in a Project Management / Producer role, preferably in digital
Ability to inspire / manage our teams of internal / external resources
Ability to manage resource allocation and workflow in the agency
Best practice knowledge in Project Management
Covering everythig from timelines, documentation and task allocation
Strong scope and change management skills
Brilliant attention to detail
Excellent communication skills
Excellent documentation skills
Ability to be pro-active with solutions and recommendations
A sense of humour and dedication to our growing team

How we will reward your efforts

Exposure to a growing list of interesting blue chip clients
Flexible working hours in a dynamic team environment
Young start-up with a "work hard, play hard" company attitude
Training on industry leading analytics and marketing platforms
Freedom to assist in developing overall agency production processes and workflows
Plenty of development and career opportunities in fast growing business
Great office space in the city that we share with other agencies
Free coffee and toast to kick-start your day

Don't forget to check us out on JobAdvisor, Facebook and LinkedIn to find out more about your potential team members and what they think about working at Datalicious.

If the above sounds interesting, please apply via Jojari (preferred) or email us at jobs@datalicious.com so we can have a look at your resume and arrange a quick initial phone interview to ask you a few questions before we meet for an interview.

Permalink | Leave a comment »

Free cloud tools for creating infographics

Thu, 04 Apr 2013 18:36:00 -0700

Infographics have become a good way of disseminating public data in a visually compelling medium. We also think they can be a good way to get business data in the hands of non-analysts, in a format that makes it easy to understand the story behind the data. There are a lot of tools available via the cloud and a number of them are suitable for blogs or internal reports. We've reviewed four 'free' tools today.

1. Infogr.am

Getting started is super easy and the user interface is well designed. It's really interactive infographic-centred rather than data-centred - you start with one of 6 templates/styles, then add in your data and comments. The templates are well designed but with limited customisation, making them simple for non-designers to use.

Infogr.am could be a good free tool for creating a public one-off chart to embed or share on Facebook, Twitter or Pinterest. The privacy controls are limited so it's better suited to visualising data and infographics for public use.
http://infogr.am/

2. Venngage:
An infographic tool developed by the team from Visual.ly - an online dataviz sharing community, which aims to make reports easier for analysts who don't want to install expensive software. There's a free demo mode with most functionality, but the Professional version offers export options and the ability to publish documents privately. It also allows you to use connections to Social Media analytics with some standard report templates set up.

In my experience all the features you'd want appear to be available, but there were a few bugs with data uploads that affected my ability to demo the tool.
http://venngage.com/

3. Easel.ly:
The UI of the tool is easy to navigate and quick to adjust to. It offers a number of layout options, including specific mobile and landscape sizes. Easel.ly has a good library of scaleable graphics including backgrounds, shapes and vector icons. You can download as a jpg or share a link to the image with private/public sharing options.

However, unlike some other cloud infographic tools, it doesn't have the ability to input tables of data into specified charts, so you'll need to create bar charts and scale your visuals by hand. This might not be an issue for the more creatively driven user for but for accurate charts and ease of use, not having standard chart formats is an obstacle for creating templates or updating the underlying data in the future.

This tool is free but only available in beta, so it's possible that more functionality might be added in the future that improves its usefulness in producing ongoing reports, so it's one to keep an eye on.
http://www.easel.ly/

4. Piktochart

This cloud tool has recently been re-developed and has a lot of functionality. It has good control over all visual elements including layout, fonts, backgrounds and so on. Piktochart comes complete with a big library of scalable vector icons as part of the free version, with more available on the Professional plan. It accomodates simple charts and can import summary data from CSV files.

The free templates are fairly simple but not as constrained as Infogr.am - they are starting points for developing your infographic, rather than a straightjacket with limited options. As with Venngage, having a lot of user freedom could mean it's easier to break the 'fundamental laws' of good design, but there is a much bigger range of templates available in the Pro version. The templates get you started in the right direction.
http://piktochart.com/

Our pick?

Piktochart is the tool we think fits our needs for infographic design, based on good functionality and ease of use, lots of customisation in visuals, layout control and good value from both the free and professional versions.

Have you got any good infographic tools we missed? Leave a comment and we'll check it out.

Looking for someone who can help you communicate the insights in your data? Contact Datalicious to discuss how we can turn your data into actionable insights.

Permalink | Leave a comment »

MapReduce, Hadoop, and doing it yourself

Wed, 27 Feb 2013 02:34:00 -0800

MapReduce, Hadoop, and doing it yourself

The Data Team attended the fantastic Big Data Meetup organised by Rami Mukhtar from Ambiata last night. I've been a very times before and always enjoyed it, but it's become particularly relevant now that Rami has decided to focus the group more specifically on the technical side of our corner of the industry. It's an open group, and anyone with an interest in data processing should consider attending. You do not need to be a database engineer or "guru" of any kind to understand what's going on. The talks last night were useful without the use cases being explicitly relevant to my own work, which says a lot about the quality of the speakers.

From our own internal discussions and my general impression from mingling with the nascent 'big data community' over the last few months, I've gathered that there are quite a few people working in this space whose understanding of the Hadoop platform and MapReduce in general is not strong. This may be in part because something like Hadoop necessarily subverts the skills barrier between application/system programmers, web developers, DBAs and operational staff (a Good Thing in and of itself).

In light of that, I thought I'd write a post - a short essay, really - outlining my personal conceptual understanding of MapReduce and Hadoop, with the hope that it will improve someone else's understanding.

As my own knowledge of this field is very recent, I'm hoping that I can express some of the concepts without the baggage of all the history of data warehousing.

Disclaimer: much of my knowledge of Hadoop beyond the difficulties of getting it up and running is second-hand. My own understanding of MapReduce is from a combination of close reading and the heavy use of other systems based on it (Riak, Mongo, CouchDB, application libraries in various programming languages).

Disclaimer #2: I will only be very briefly touching on both the benefits and drawbacks of Hadoop and "Big Data" platforms in general. If you're trying to make decisions about your own operational environment, it's safe to say that you should worry about the unknown drawbacks until you're absolutely sure that you don't need to.

Introduction

When analysing large datasets for online marketing, almost all the questions you have from your dataset involve aggregation of data by some key (often a "user id").

It's important that the term "aggregation" not be overloaded by any familiarity you have with SQL. You can think of this alternatively as "sessionisation" of data.

Essentially we want to take very large sets of data and organise them into "paths". These paths are a time-ordered set of all available data for a given user that lead to a particular data point.

That's the fundamental operation, but to make it valuable there's an awful lot of complexity around it in practise.

Even so, there's a lot to be gained from understanding that the concept of "pathing" is fairly generic, as there is existing theory - and even code libraries - focused on dealing with this problem. Most of our work has been designed with the underlying generic aspects of the problem in mind.

An excellent introduction to these concepts can be found in the book "Mining of Massive Datasets" by Anand Rajaraman and Jeffrey David Ullman. It's useful even if, like me, your eyes glaze over when you see anything that looks like formal symbolic logic, or just straightforward undergraduate maths.

MapReduce

Rami's talk on Tuesday night focused on Ambiata's use of the Scoobi project. Scoobi is a DSL (Domain Specific Language) implemented in the Scala programming language for working with Hadoop clusters. Code written in Scoobi is turned into the parts of a MapReduce job that runs on Hadoop.

Hadoop queries are built around a general concept or model called "MapReduce". The idea MapReduce of is to break a problem down logically into two main stages. Very simplistically, these are:

map - transformation, selection and filtering of data
reduce - aggregation

The lifetime of a job in any MapReduce system has many other stages before, in between, and after those stages, including but not limited to: planning, shuffling, sorting, error condition handling, provisioning, storage of temporary files, etc. Some of these relate to how all the data gets shunted around and managed, others relate to optimisation. In general, the conceptual model can be limited to MapReduce, or MapShuffleReduce, for the purposes of development and design.

Think of a Hadoob "job" as a query in a traditional database. The programmer / DBA (how happy I am to see this distinction torn down) writes jobs and asks the Hadoop cluster - i.e., the database - to execute it. Hadoop will take these jobs and distribute the various components of them to many - dozens, hundreds, even thousands - of machines working on each stage of the process.

The way the user expresses what they want from the database is very different to a traditional approach driven by SQL.

Let's say someone wants to know the total of the ages of all users, apparently. In SQL you might write this:

SELECT SUM(age) AS total_age FROM users;

Here's an example of how you would do it instead in a fictional JavaScript MapReduce [1] framework:

// Here we assume that the MapReduce framework will look up
// global namespace to find the 'map' and 'reduce' functions,

// It will call map() for every user object found in our table, file, etc.:
var map = function(user) {
    return user.age;
};

// And it will call reduce() a single time with the result of all the map calls:
var reduce = function(user_ages) {
    var sum = 0;
    user_ages.forEach(function(age) {
        sum += age;
    });
};

Languages like Clojure and Python let the programmer do this work very concisely. Respectively:

(reduce + (map :age users))

reduce(lambda x, y: x + y, map(lambda x: x.age, users))

The reason this model can be mapped so easily to different languages and frameworks is that it's inspired by concepts from functional programming, which emphasises (among other things):

functions that are general enough to be easily and fluidly composed with each other
functions that are "pure": they take some input and produce some output and do nothing else, and always produce the same output for a given input

The "functional" underpinning of the MapReduce framework means that queries have particular predictable semantics. This predictability allows a well-designed database to execute a query in parallel very efficiently. This is hard to do with queries written in a language like SQL for a number of reasons. Simplistically, a language like SQL doesn't have enough "barriers" between different stages of programming, and too much implicit transferral of control and state, to allow it to be effectively broken up into many different pieces of work procedurally and run in parallel [2]. In MapReduce, this capability is built-in from the beginning.

Hadoop

That all sounds great, but it's not so simple.

MapReduce has this neat functional underpinning, but in traditional Hadoop you have to write a whole lot of ugly object-oriented code in Java and jump through lots of hoops to what what seem like really basic things. Systems like Scoobi and Hive exist to build a cleaner interface atop that.

An illustrative example is the canonical "Word Count" problem. The Java solution:

http://wiki.apache.org/hadoop/WordCount

A Scoobi version I found:

https://gist.github.com/schmmd/1942948#file-src-main-scala-wordcount-scala

The difference is pretty clear (if you set aside how weird Scala code looks).

This is why all the companies using Hadoop for big data can, to a certain extent, throw additional machines at a problem. It's why their data can live on fifty different machines, and travel through five hundred machines at query time. Each individual Hadoop cluster at Facebook has ~2000 machines. Supposedly, with a brilliantly architected network, you can get all of those machines doing lots of work at once on a single query.

In almost all small- and mid-sized Hadoop clusters, engineers hit bottlenecks with the internal network and things like operating system I/O barriers far sooner than they can utilise a good fraction of their theoretical processing capacity. Even in clusters of 5-6 machines this can be a serious issue.

There are also serious problems with network contention in Hadoop clusters. From what reading I have done, things like network bandwidth, disk I/O and operating system buffers become bottlenecks before it's possible to use a good chunk of the theoretical compute capacity of a cluster - even in relatively small clusters of half a dozen machines. The clusters of thousands of machines and Yahoo, Facebook, and so on would involve huge investments into the optimisation of the network and the hardware configuration of the machines.

Trade-offs

A key factor in our platform decisions is the issue of "throughput". Here at Datalicious we need to produce very large data extracts from our processing platform and deliver them to the Insights team for analysis in various statistical packages.

There's a lot of pressure on the database to stream data out while the pipeline is in operation. The majority of well-published use cases of Hadoop and almost all enterprise big data platforms can demonstrate fast query results, but the actual data representing the query result is often trivially small (relative to the input data). Our pipeline actually produces more data than goes into it! At least locally, this gives us a fairly niche set of requirements.

For this reason, our current solution to our largest problems is bespoke. We have developed our own framework with a combination of languages and database tools, most notably core GNU/Linux tools, Go and Python

This means that we can utilise the hardware to the fullest, and gives us tremendous efficiency in the small. After appropriate optimisations, we were able to fully utilise the storage available storage throughput as well as all 8 cores of a production system during processing.

Despite not being on a data platform, there's a lot to be gained from tools that have developed around the Hadoop ecosystem in particular. Flume and Zookeeper are both projects that we're taking an interest in for part of our pipeline. Another interesting tool is the Storm Project. Storm is designed to be a "real-time" analogue of Hadoop, but it has a comparatively light footprint in some ways, and makes far fewer assumptions about its environment.

The ~80 lines of largely declarative Scala/Scoobi code that Rami showed is equivalent to a far greater chunk of our own code. It's widely accepted in general that more lines of code = more bugs. On the other hand, it's trivial for us to produce mockups and write unit tests for our code, and we have pretty decent test coverage, at least of the core logic.

In addition to that, we a great deal of code and configuration for managing the processing and infrastucture. I can't speculate as to what Ambiata's Hadoop cluster would have in terms of custom configuration, but the advantage for Hadoop there is that there is tooling to help manage the configuration. Obviously most of this work is simply what Hadoop does out of the box - albeit at a high operational overhead.

It's crucial to note that our this is all distinct from the code that encapsulates all the "business logic", which is where we see the real value. There are many factors that complicate what we wish could be a simple declarative query.

That's all the stuff that we have fine grained, testable control over with our imperative code. This is all the work that essentially needs to be redefined for each client, and it involves bespoke configuration and code. It's not "generic", so something functional like Scoobi, or MapReduce with its proliferation of libraries, does not have as much of an advantage.

Additionally, much of the work we do to our data is fiddly - the rules are recursive to some extent, and there are all sorts of weird branching edge cases. It could still be done in a higher-level query language, but there's a tipping point where our solution is actually cleaner and more straightforward. I believe we're already at that point.

And as for parallelisation - thanks to some particular characteristics of our data, we have our own solution which avoids all sorts of nasty issues with windowing and so on that come up when data like this is processed on a MapReduce cluster.

If you're interested in our Media Attribution offering for optimising media spend, contact us!

Daniel Bryan is a Senior Technical Analyst at Datalicious, and is responsible for the implementation of data processing and aggregation architectures for a number of projects and clients.

[1] there are actually many MapReduce implementations where JavaScript is a supported query language, but my example was not based off of one in particular. For those curious about MapReduce in JavaScript, underscore.js is a browser library providing some rich functional programming constructs, including map(), reduce() and filter().

[2] a lot of work has been put into this, and systems like Dremel / Google BigQuery, Hive, and a number of databases (e.g. Greenplum, a fork of the open-source PostgreSQL) have come a long way in allowing SQL or SQL-like queries to run parallel, but there are some serious limitations and trade-offs.

Permalink | Leave a comment »

Guest Post: How to use data and become the SEO expert that packs a punch!

Wed, 19 Dec 2012 21:33:00 -0800

I’m Tom Sheppard Check me out on Linkedin!

All the grueling interviews are behind you, today is the first day in your new SEO role. It’s time to get into some serious SEO work and actually make a difference to the company that has hired you. Sounds simple enough right?

Well here’s how I use DATA to make smart decisions.

No messing around, log straight into Google Analytics or the next available web analytics package and start segmenting the google / organic dimension like crazy. That’s assuming google / organic is the biggest referral of sales out of the organic bucket. Exclude the brand keyword terms associated with your company. If there are a few variants of a brand term, consider using a pipe symbol between them and “matching RegExp” to pull out those unwanted keywords…

For more info about Regex click here

Now that we have stripped out the brands terms, the real fun begins…

Start by exporting the raw non-branded data into excel along with the conversion data.
Cross-reference a fresh ranking report to the keywords exported.
Cross-reference a search query report from Webmaster Tools, to get the approx CTRs of the keywords.
Calculate what each keyword is worth in terms of dollar values.

You will end up with something like this:

Now if we sort by revenue to give us the most valuable terms, we can work on making a difference pretty quickly.

Consider how much additional revenue you can generate by working through your keywords this way.

For Example: The super cool ipad term is worth 5 times more per sale, vs. cheap phones and it offers more opportunities to grow given it’s rank & low CTR.

Going back to the revenue-ordered table, search each keyword, ask questions (like those below) & make notes. This will ultimately form the basis of your action list.

Questions to ask

What else is on the SERPS for that term: News results / videos / images / maps etc
Title tag: Is it matching the words in my analytics report? Is the keyword at the start of the title? Is the title tag too long?
Meta Description: What is being displayed? Is it compelling enough? Is the CTR high or low vs. other terms in similar positions? Are your USPs mentioned?
What is on the landing page? Does it match the user’s query? Are the on page factors there? Is the page fast or slow?

Here is a worked example on how data begins to pack a punch

After asking the above and re-looking at the mock data – you find [tablet deals] interesting due to:

It's Keyword ranks #5 with significant sales, the most in fact.
Has a fantastic Meta description given the CTR and the position it is ranked.
No universal search results being displayed, and the landing page converts fantastically.

All bodes well for some optimization work.

By digging a bit more, a link analysis report (above) shows there are few internal links pointing at that landing page and it still ranks well. By tweaking the count of links flowing towards the “tablet deals” page to be similar to the “optimized tablet deals”, (red striped pattern) we might hope to see signs of us climbing up the SERPS. Be sure to monitor rankings & traffic once the change is implemented. The end result = more traffic and greater potential for increased sales, all thanks to every SEO’s favourite tool…. DATA.

If you are desperate for more SEO & Data info check out the video “Better Rankings in Day” – it’s a really great resource.

Bio: Tom Sheppard is an experienced Search expert, with specialist SEO, SEM and Web Analytics experience from Australia and the UK. If you would like to connect please do so via linkedin or on Google+.

Permalink | Leave a comment »

Should Ricky Ponting have retired? Find out, we analysed and visualised his test batting career in R

Wed, 12 Dec 2012 22:16:00 -0800

A few weeks ago some friends were debating whether Ponting should retire. There were arguments in favour, that he still had something to offer the team not only as a batsmen but also a fielder. The naysayers said that he just wasn't what he used to be, that his best days were behind him, and that he should make way for the future of Australian cricket. Turns out he decided to go anyway, but we thought it worthwhile to see if he made the right decision...

We found this great website howSTAT! that had the batting records of the world's Test batsmen, and scraped it off using this new Chrome extension called Scraper .

Typically someone's batting record gets presented like the chart below, but when batsmen have had many innings this chart gets rather unintelliglble. It's even more unintelliglble when you're trying to compare one batsman's chart to another.

Cricket, being the birthplace of statistics (after governments keeping tabs on their citizens), deserves better analyses than mere bar charts and batting averages. So we loaded the data into R and used the ggplot2 package* to examine how Ponting's career has ebbed and flowed over time. A first, we thought a moving average might be a good way to identify trends and you can see the semblance of a trend in the chart below.

Turns out, ggplot2 has this great smoothing function that helps to identify trends in data. Applying this function to Ponting's career, we see that Ponting has had two purple patches with the bat. His first peak in form was the latter half of 2002 and the whole of 2003, with centuries against Pakistan, England, West Indies, Zimbabwe, and culminating in two double tons against India at the Adelaide Oval and the MCG. Unfortunately the first double ton wasn't good enough as Dravid stole the show and helped India to a 4 wicket win. The second did, however, help Australia win comfortably.

The second peak was the summer of 05/06 against the West Indies and South Africa, making 149 and 104 not out in Brisbane against the West Indies, and 120 and 143 against South Africa at the Sydney Cricket Ground (in the latter match, Australia chased down the 288 set by South Africa in the fourth innings).

Now as for whether this was the right time to retire, we certainly see a recent downturn in Ricky's form (even with the double ton against India at the Adelaide Oval) that does suggest that his decision was a good one. He's been able to leave the game before things turned ugly.

With that settled we decided to turn our attention to how Ponting's record compared to other recent Australian captains. The chart below shows that Ponting certainly didn't have the fairy-tale ending of Steve Waugh, Greg Chappell. or even Allan Border. However, it certainly wasn't as quick a deterioration in performance as Mark Taylor or Kym Hughes. Unfortunately, Kym Hughes copped a back-to-back series against the West Indies in their prime with Holding, Marshall, Garner, et. al**

Let's also see how Ponting fares against some other recent greats. He certainly has the same double hump career pattern as Lara (although Lara's form peaks were wider apart). Kallis is also in his second peak and we hope that he can maintain his form for some time yet. The most striking trend is Tendulkar's decline in form. Surely, this current series against England in India is his farewell tour.

And what of the others in the current Australian cricket team? Mike Hussey really did turn things around against South Africa so let's hope he can continue this form against Sri Lanka. We all know that Michael Clarke has been batting well but we see that when he is down on form, he's really down on form. And Watson, he's hit his lowest point, in a similar place to where Brad Haddin found himself. He better get some runs in Hobart...

* We've been learning R recently by following Hadley Wicham's Data Analysis course at Rice University. It's great.

** We have done a bit more smoothing work in this chart so that the batting records can be more easily compared and, no, Kym Hughes didn't score negative runs in his final few innings - this is just an artifact of the smoothing method).

Permalink | Leave a comment »

Shining a light on the mysterious dark social and direct traffic channels for media attribution

Mon, 19 Nov 2012 23:39:00 -0800

In the past few weeks, a variety of our clients have been forwarding us questions about this "Dark Social" phenomenon, recently highlighted in an article that appeared in The Atlantic magazine. The basic idea is that:

A lot of what happens on the web is social in nature (tweeting, sending emails, instant messages, facebook posts, etc between people); and
The extent of this activity, and it's impact on your business/site, is difficult to quantify.

For publishers of content such as The Atlantic, and businesses that use content to drive traffic, it's crucial to know how much of your content "out there" is creating a buzz and driving visitation.

Typically, one would just measure the amount of visits that came via referral from particular domains. However, referrals can still be only a small proportion of your traffic, leaving a still large portion of traffic denoted as "Direct" or "Typed/Bookmarked". So, how could one figure out the true source of this traffic?

A great idea (credited to Chartbeat) outlined in the article was to examine the landing pages of these "Direct" visitors, and if their landing page was something like this

"http://www.theatlantic.com/technology/archive/2012/10/atlast-the-gargantuan-t.../"

then clearly that person arrived at the site via a link forwarded to them by someone else or posted, or tweeted or whatever. It's very unlikely that someone would type this into a browser. These visitors were denoted as arriving from "Dark Social" (We think a better, but less catchy, name would just be "Social Web").

Now that we've identified our Social Web traffic, the next question is to work out the source of the rest of this Direct traffic. Theoretically, this traffic already knows our brand but we need to work out how they already know our brand.

The first step is to implement what we call path to purchase tracking, which essentially keeps a log of traffic sources and Display Media impressions delivered to a particular person. Quite often, someone will visit a website Direct who have previously been exposed to Display Impressions or completed a Generic Search.

A second step might be to find out which of your Direct visitors are your customers/subscribers. By definition they must already know you, regardless of any marketing activity. You can do this if you have any online self-service or login mechanism.

A third step is serve a small pop-up question to your "Direct" traffic, and ask them how they found out about you, whether that be a previous retail experience, friend and family, or a television ad, for example.

The final result is that you've got your Direct traffic split into categories that might look like this

Social Web
Display Impressions
Generic Search
Existing Customers
Retail (bricks and mortar)
Television
Word-of Mouth

Any remainder might just be bundled into the catch-all of "Brand Equity".

Now that each visit is categorised and measured more accurately, we can assign conversion credits more accurately (with a multi-touchpoint/path-to-purchase view), and subsequently make better assessments about the impact of our multi-channel marketing strategies.

Permalink | Leave a comment »

Multi-channel media attribution or why measuring the last click is just not enough to maximise media ROI

Sun, 18 Nov 2012 02:43:00 -0800

Media attribution is an incredibly hot topic around our office at the moment, and it's also a topic we are very passionate about.

If you think your company should be analysing all campaign touch points leading up to a conversion, not just the last, then this video will be of great value to you. Christian Bartens our Managing Director helps show you how to record and analyse multi-channel media attribution data in your existing web analytics platform to generate some quick wins, as well explain the pros and cons of more professional long-term approaches. The slides are available for download from Scribd.

This video was filmed at our 'Boosting Media Value Event' held in Sydney in October 2012. We'll be holding another event in the New Year so stay tuned for updates.

Video recorded and edited by:

Permalink | Leave a comment »

TrinityP3: The media and marketing value chain or how to get value from agency relationships

Sun, 18 Nov 2012 02:41:00 -0800

For those of you who couldn't attend our 'Boosting Media Value' event held in October, here is the first keynote presentation from Darren Wooley, Managing Director of Trinity P3.

In this video, Darren shares insights and practices to understand where media and marketing value is created, lost and hidden. He speaks very passionately about Marketers who are bombarded with claims and recommendation on the best way to invest their often largest investment - their media budget - by people with a vested interest in the way they will spend this money. Darren makes recommendations on practical ways for advertisers to unlock the value in their media investment. The slides are available for download from Scribd.

If you enjoye this video presentation then please stay tuned, we'll be holding another event in early 2013.

Video recorded and edited by:

Permalink | Leave a comment »

ThinkVine: Agent based marketing mix modelling and optimisation in today’s consumer-driven world

Sun, 18 Nov 2012 02:28:00 -0800

We are very proud to be the Australian partner of ThinkVine, a marketing mix modelling agency from the US. We recently had the CEO come all the way from the US to present the keynote speech at our Boosting Media Event.

In this video Mark helps you understand why agent-based modelling is a smarter approach to marketing mix optimisation. The video also includes simple explanations of this cutting edge approach and interesting case studies from the US. The slides are available for download at Scribd.

If you are interested in marketing mix modelling, please get in touch and we can arrange a demonstration.

Video recorded and edited by:

Permalink | Leave a comment »

We're hiring! Looking for Business Development Director to join our team of geeks in Sydney ($100k+)

Thu, 01 Nov 2012 19:57:00 -0700

THIS ROLE HAS BEEN FILLED!

Are you an experienced business development or sales wizard with a strong marketing technology background and proven revenue track record that is looking for a new opportunity in one of the fastest growing industry sectors?

Datalicious is a fast paced data and analytics agency that has started to develop its own products in order to help us grow faster, both in Australia and overseas. We’re now looking for a self-motivated individual that can operate in a start-up environment to build-up and lead our sales team so we can start capitalising on all the opportunities that present themselves.

Please note: Sales people without at least $2m current annual targets need not apply!

What you should bring to the table

+ Strong background in marketing technology and/or solution sales

+ Keen interest in all things data, analytics and marketing optimisation

+ Ability to not only convert leads but also proactively develop new ones

+ Proven track record of annual sales revenue of at least $2 million or more

+ Demonstrated ability to operate in a start-up without formalised support

+ Experience in building-up and managing small and effective sales teams

+ Experience in developing a partners sales channel would also be useful

+ Entrepreneurial spirit, business or marketing degree a definite bonus

+ Ability to turn product feature list into convincing business benefits

+ Ability to transform custom projects into out-of-the-box service packages

+ Ability to customise sales pitch for different client and industry needs

+ Ability to manage sales process from opportunity to approved estimate

+ Proven experience dealing with clients and partners at the highest level

+ Dedication to providing lightening fast professional responses to all leads

How we will reward your efforts

+ Exposure to a huge list of blue chip clients

+ Flexible working hours in a dynamic team environment

+ Young start-up with a "work hard, play hard" company attitude

+ Potential for public speaking engagements to build personal profile

+ Potential for a wider international role in the not too distant future

+ Freedom to manage, grow and shape your team as you see fit

+ Attractive salary base plus signifiacnt commission based bonus

And don't forget to check us out on JobAdvisor, Facebook and LinkedIn to find out more about your potential team members and what they think about working at Datalicious.

If the above sounds interesting, please email us at jobs@datalicious.com so we can have a look at your resume and arrange a quick initial phone interview to ask you a few questions before we meet for a proper interview.

Permalink | Leave a comment »

We're hiring! Looking for a Technical Analyst to join our team of data geeks in Sydney ($60-80K)

Thu, 01 Nov 2012 19:15:00 -0700

THIS ROLE HAS BEEN FILLED!

If you are a website or database developer (or an analyst with exceptional technical skills) with anywhere between 1 to 4 years of work experience (ideally in a digital agency environment) who would like to get into marketing analytics then you should apply. More experienced developers are welcome to apply as well.

Please note: Web developers without any web analytics skills at all need not apply!

What you should bring to the table

+ Keen interest in all things web analytics and online optimisation

+ Basic understanding of marketing principles provides a head start

+ Entrepreneurial spirit, business or marketing degree would be good

+ Strong JavaScript and SQL coding skills, PHP and HTML a bonus

+ Experience with either SiteCatalyst or Google Analytics a must

+ Experience with Google Web Site Optimizer would also be useful

+ Basic understanding of HTTP and FTP protocols an absolute must

+ Attention to detail and experience with testing processes essential

+ Experience dealing with large and complex datasets would be useful

+ Intermediate to advanced data manipulation and ETL skills also help

+ Experience with other database related technologies another bonus

+ Flexibility, lateral thinking and attention to detail are crucial

+ Ability to respond to client deadlines when necessary

+ Basic Excel, Word and PowerPoint skills

How we will reward your efforts

+ Exposure to a growing list of interesting blue chip clients

+ Highly flexible working hours in a dynamic team environment

+ Young start-up with a "work hard, play hard" company attitude

+ Training on industry leading analytics and marketing platforms

+ Freedom to experiment with emerging technologies and new tools

+ Plenty of development and career opportunities in fast growing business

+ And of course you get a salary, maybe even a bonus plus other perks

And don't forget to check us out on JobAdvisor, Facebook and LinkedIn to find out more about your potential team members and what they think about working at Datalicious.

Permalink | Leave a comment »

Internet Explorer 10 and the Do Not Track function

Thu, 01 Nov 2012 16:08:00 -0700

Along with the release of Windows 8 last week, Microsoft has also released their latest version of Internet Explorer – IE10. Version 10 is currently only available on Windows 8, but will be released mid-November for Windows 7. Vista and XP will not be compatible with IE10.

Besides enhanced support for everything HTML5, CSS3 and JavaScript, most users will see a great increase in speed in web browsing. Although IE10 is still not the fastest browser around, the difference between the latest versions of the “Big Three” (Chrome, Firefox and IE) is getting smaller.

Visually, the update includes edge-to-edge browsing. You won’t see any more navigation bars or tabs by default - the browser fills your entire screen.

The biggest addition to Microsoft’s browser is one that, though normal users will encourage, could have a big impact on the online marketing industry. With IE10, Microsoft implants the Do Not Track (DNT) function. This function does exactly what is says: it doesn’t track what a user is doing on a website. While most browsers have this function available, a user has to manually activate it. In IE10, the DNT function will be on by default (unless a user manually turns it off).

A research team from Microsoft found out that 75% of Internet users in North America and Europe don’t want to be tracked; which seemed reason enough to add the DNT function to the browser. Brendon Lynch, Microsoft’s Chief Privacy Officer stated, “Consumers want and expect strong privacy protection to be built into Microsoft products and services”.

Large corporations are already picking up on the fuss about the latest updates, Yahoo has written a complaint letter to Microsoft, saying it will ignore the request from the DNT function and continue tracking user activity. This letter has been signed by more than 30 brands; amongst them are Procter & Gamble, Ford, Coca Cola, Unilever, and American Express.

So what‘s going to happen? In the end, it seems that not much will change - Yahoo will ignore IE10’s DNT request, continuing to track users and serve ads. Furthermore, Apache, the world’s most used webserver, already has a work around that ignores DNT requests.

Permalink | Leave a comment »

Want to start your own data business? Read about how Datalicious started and attend start-up bootcamp

Wed, 31 Oct 2012 18:42:00 -0700

I don't usually like writing about myself but I heard about this start-up bootcamp event in Sydney coming up on November 10 and was looking for a reason to blog about it as I think everyone should try and start their own business. Unfortunately data and analytics have very little to do with entrepreneurship apart from the fact that Datalicious used to be a start-up as well so I thought I'd share how the company got started. If you have no ambitions to become your own boss tune out now.

Datalicious actually goes all the way back to 1997 when I started my first business right after high school together with some friends of mine, The Dive And Travel Company. Yes, that's right, I'm a PADI/CMAS scuba diving instructor by trade and only immersed myself in spreadsheets later in life. Stupid decision, I know! Now you might ask yourself what that has to do with Datalicious? The answer is persistence. Even with the best of ideas you simply don't create a successful business without it - and without failing a few times.

Although the scuba diving school still exists today I have to admit it was a bit of a failure at the time. We had fun and learned a lot but we certainly didn't make enough money for all the time we invested. Instead of packing it in and moving on I then jumped onto the Internet bandwagon and started another company, ScubaCrawler.com. The community website was meant to become the central information and planning hub for divers around the world but unfortunately never really got off the ground. We actually ended up in court suing our developer at the time because we were stupid enough to pay him in advance for skills he didn't really have. I know, well done - but at least it put me off wanting to run my own business. I was forced to get a proper university degree and a series of real jobs which ultimately made me end up in Australia and provided me with the necessary skills and opportunity to start Datalicious in 2007.

The point of all this is that it might take a few attempts - persistance - before you get it right. I failed twice before I finally made it work with Datalicious and now SuperTag, maybe. Mind you, even that started as a shaky part-time gig working from home that could have gone either way for a very long time and in the end took five years of hard work with lots of ups and downs. If you persist the right opportunity will present itself at some stage - when that happens you just have to run with it, take a risk and give it a go. In fact, it will only be the first of many opportunities that present themselves because if you keep going you will continue to create new and gradually bigger opportunities for yourself. In Datalicious' case, it means using our local data and analytics experience and turning it into products to start an international SaaS business. I would like to claim that this was my grand master plan from the very beginning but that wouldn't be true - it's simply the result of persistence, a series of seized opportunities, lots of work and a little bit of luck.

Ok, enough of the pep talk but I urge you to give it a try and start your own business! After all, what can possible go wrong? I am touching wood right now. Anyway, I'm sure there's a business idea lingering somewhere in that brain of yours so sign-up for the upcoming start-up bootcamp event in Sydney now and see what happens.

Permalink | Leave a comment »

We're hiring! Looking for a Senior Marketing Analyst to join our team of data geeks in Sydney ($70-$90K)

Thu, 04 Oct 2012 02:43:00 -0700

THIS ROLE HAS BEEN FILLED!

After over a year of smashing data as a Marketing Analyst at Datalicious I am moving on to a new and different role in the world of education (a company called Smart Sparrow), and we are looking for a skilled replacement.

Working at Datalicious is quite a unique experience. You will never get bored doing the same old job; every client is different and the challenges vary immensely. This type of variety is unrivaled and makes the work truly engaging. I have worked on projects ranging from data visualisations of Tweets and liveable Melbourne suburbs, to homepage optimisation and analysis for the big banks, to large scale CRM and transaction analyses using machine learning, to analyses on what ads on the web are working for the big telcos (media attribution). You get the picture ...

The role has offered me a great learning experience; from three-day presenting workshops, to online machine learning and database courses, to accredited Tableau and Splunk training. So you will be constantly picking up new skills. The whole team has just recently decided to move towards using R as their common analytics tool; so this was a great opportunity to learn a new programming language.

As a function of the range of jobs we take on at Datalicious, you will also use a variety tools to do your analyses. This will allow you to build up a great analyst arsenal. We love experimenting with new (and established) cutting edge technologies such as Splunk and R, but the list goes on.

In short, you'll have fun in this young and passionate team of nerds, you'll learn lots of great data related stuff, and you'll always have a new challenging task at your desk. I will miss it.

More specific details of the role ...

Senior Marketing Analyst

The chief purpose of the role is the analysis of large, complex customer and web analytics datasets with the aim of reporting marketing performance as well as developing communications and contact strategies for our clients. The role would ideally suit someone who would like to develop their marketing nous and have the opportunity to entrench their data analysis skills with customer and web data.

What you should bring to the table:
+ Extensive experience with at least one language such as R, SPSS, SAS, SQL, or Matlab
+ Experience with advanced statistical analysis or implementation of complex business rules
+ Keen interest in all things web analytics and marketing optimization
+ Experience dealing with large and complex datasets essential
+ Basic understanding of SQL and BI infrastructure would be good
+ Proven ability to turn data into insights and actionable recommendations
+ Proven ability to effectively communicate insights to key stakeholders
+ Solid understanding of marketing principles would be very useful
+ Entrepreneurial spirit, business or marketing degree would be good
+ Experience in dashboard design and development would be nice
+ Hands on experience with Tableau software would be even better
+ Experience with Omniture and Google Analytics would be useful
+ Flexibility, lateral thinking and attention to detail are crucial
+ Ability to respond to client deadlines when necessary
+ Reasonable Excel, Word and PowerPoint skills

How we will reward your efforts:
+ Exposure to a growing list of interesting blue chip clients
+ Highly flexible working hours in a dynamic team environment
+ Young start-up with a "work hard, play hard" company attitude
+ Training on industry leading analytics and marketing platforms
+ Freedom to experiment with emerging technologies and new tools
+ Plenty of development and career opportunities in fast growing business
+ And of course you get a salary, maybe even a bonus plus other perks

Permalink | Leave a comment »

Guest post: A breakthrough approach to marketing mix optimisation in today’s consumer-driven world

Tue, 02 Oct 2012 18:35:00 -0700

By Mark Battaglia, CEO at ThinkVine

Marketers say, “you had me at hello” when the conversation is about complexity in the marketplace. No one doubts that there is more diversity among consumers, more fragmentation in media consumption behavior, more marketing choices and more ways to buy. And, no one doubts that the pace of change is accelerating.

The explosion in digital and emerging media channels has also brought a huge volume of rich, but complex, data sources. With big data comes the promise of better decision-making information and measurement of marketing performance.

Over the past few years, many ways to collect, store, report on and visualize data have been developed and proven. Most give us looks in a rearview mirror – they tell us what already happened in a small part of the world. But, despite increasing demands for accountability, we rarely use data to look ahead to see, via broad or narrow views, where we should go. Simple historical reports with counts and averages by tactic and for all the customers in the database, even if they are presented in a whiz-bang dashboard, may be easy to understand; help make the next campaign a little more efficient; or prepare metrics for the next monthly report, but they don’t supply the information decision-makers really need in today’s complex world, much less tomorrow’s.

It is possible now to employ an agile, objective, cross-channel marketing planning process that uses the readily available data and new types of advanced analytics to provide the forward-looking information marketers need to develop an effective and efficient annual marketing plan that spans all consumer groups, all types of marketing, and all channels, and to adjust that plan throughout the year as the market changes – even if doing so means going outside our collective comfort zone. Because the market has changed and continues to change, and because data is the new currency of the realm, the media industry must embrace data and technology not only to improve tactical efficiency but also to inform marketing strategy.

To learn more about ThinkVine, visit www.ThinkVine.com, or follow us at www.twitter.com/ThinkVine or www.linkedin.com/company/ThinkVine.

Permalink | Leave a comment »

We're hiring! Looking for a Head of Data to join our team of data geeks in Sydney ($100-140K)

Wed, 26 Sep 2012 20:50:00 -0700

THIS ROLE HAS BEEN FILLED!

Are you a senior technology lead or a data scientist with a strong technical background and proven experience in big data as well as team management?

Datalicious is a fast paced data and analytics agency who has a genuine interest in developing products and solutions for our blue chip clients based on new cutting edge technologies. We’re now looking for a self-motivated individual who has the right innovative problem-solving attitude as well as proven project and team management experience to lead our technology team in Sydney.

Please note: managers without serious hands on technical expertise need not apply!

What you should bring to the table

+ Keen interest in all things data, analytics and marketing optimisation

+ Proven ability to store, process and manipulate large data sets

+ Demonstrated ability for lateral thinking and problem solving

+ Project and team management experience, large/mid scale projects

+ Understanding of basic marketing principles would come in handy

+ Entrepreneurial spirit, business or marketing degree a bonus

+ Experience in data warehouse specification and delivery

+ Experience with big data and database related technologies

+ Some HTML, JavaScript and PHP skills would help as well

+ Experience with CRM, ad serving, web analytics and testing useful

+ Ability to identify client opportunities and turn them into projects

+ Ability to translate business requirements into project deliverables

+ Exceptional organizational, presentation and communication skills

+ Risk management to maximise data quality and minimise data loss

+ Attention to detail bordering on OCD absolutely crucial for success

+ Demonstrated ability to manage client and partner relationships

+ Demonstrated ability for developing processes and documentation

+ Confidence in identifying, presenting and selling-in solutions to clients

+ Demonstrated ability to manage resources and client expectations

+ Demonstrated ability to manage R&D projects with new technologies

+ Ability to accurately estimate and manage jobs within technology team

+ Ability to respond to client deadlines when necessary

How we will reward your efforts

+ Exposure to a huge list of blue chip clients

+ Flexible working hours in a dynamic team environment

+ Young start-up with a "work hard, play hard" company attitude

+ Training on industry leading analytics and marketing platforms

+ Freedom to experiment with emerging technologies and new tools

+ Potential for public speaking engagements to build personal profile

+ Potential for a wider international role in the not too distant future

+ Ability to build new technology that will meet a genuine market need

+ Freedom to manage, grow and shape your team as you see fit

+ Attractive salary base plus performance based bonus and perks

Permalink | Leave a comment »

ADMA guest lecture slides on measurement and metrics at Multi-Channel Marketing Certificate

Wed, 26 Sep 2012 19:48:00 -0700

In this increasing complex media world, the ADMA Multi-Channel Marketing Certificate is an awesome foundation for any Marketer and teaches how to develop effective strategies in this environment. We especially like the course as we get to do a guest lecture on measurement in a multi-channel world every-time (yay, pat on the shoulder)! So check out our slides below, download them and save them somewhere for later use - hopefully you get some ideas and any questions just email us.

201209 ADMA Multi-Channel Measurement V1.pdf Download this file

Permalink | Leave a comment »

Datalicious, TrinityP3 & ThinkVine to host exclusive media optimisation event on October 4th

Tue, 18 Sep 2012 05:30:00 -0700

Attention all heads of marketing: Want to learn about media optimisation and how to boost media ROI?

To help celebrate ThinkVine’s visit to Sydney, the teams at Datalicious and Trinity P3 are coming together to host a 360° media event to help you get the most out of your media spend.

Held at The City Hotel on Kent Street, the event will showcase new strategies for boosting the return on your media investment. The afternoon promises to be very insightful and will cover some of the latest best practices and techniques from experts in the industry.

The presentations will be followed by networking drinks in The City Hotel Beer Garden. Read on for more details!

Where: The City Hotel, 347 Kent Street Sydney, Sydney
When: October 4th, 2:00pm to 5pm (followed by networking drinks)
Pirce: $15.00 (all proceeds will be donated to Unicef Australia)

Read more and register now.

Key speakers

Darren Woolley, Managing Director - TrinityP3
The media and marketing value chain: Understanding where value is created, lost and hidden
Christian Bartens, Managing Director - Datalicious
Multi channel media attribution: Why measuring the last click is just not enough
Media Attribution Case Study - TBC
Real world example of techniques in practice
Mark Battaglia, CEO - ThinkVine
A breakthrough approach to marketing mix optimisation for today’s consumer-driven world

Come join us for drinks post event and make the most of this chance to pick the brains of some genuine experts in the industry!

Read more and register now.

See the full gallery on Posterous

Permalink | Leave a comment »

Towards a single customer view: Understanding the problem before launching into solutions mode

Fri, 14 Sep 2012 00:03:00 -0700

The concept of a 'single customer view' is applicable to a few different problems in web analytics and online marketing. We've blogged before about using Splunk as a platform to build a single customer view from web data, while decorating it with offline data (such as call centre conversions, classification schemes and cost data).

This is part one of a two-part post. Part two will cover the slightly more technical side of things: what's involved in implemented the necessary tracking, and the platforms and tools that are available for building a solution.

Many web analytics platforms offer a limited form of something like a 'single customer view' implicitly: the concept of the "Visit" and the "Visitor" has been around for a long time; generally speaking, it's difficult to understand unique users' behaviour over anything more than the very short-term, and a conversion or even a funnel start is seen to be the very end of a path, rather than a part of a customer's ongoing relationship with your business. The real understanding of paths is usually locked away behind the built-in model of metrics, traffic sources and so on.

With greater popularity and understanding of the paradigm of 'big data' and the proliferation of platforms capable of aggregating unstructured or semi-structured data, it has become possible to construct a single customer view encompassing the entirety of a customer's relationship with a business' online activities, covering display ad impressions, organic and paid search traffic, EDM, partner referrals, post-impression activity and more.

With the SuperTag and its global container tag, it's possible to implement tracking across all parts of your business' site and ensure that all the data can be matched with information from other platforms such as ad servers and call centres.

Here are a few points to consider before embarking on a single customer view project:

How much detail do you need in your single customer view? Do you want to see a full picture of all touchpoints across channels leading to conversion, with all the detail on each touchpoint preserved? Or do you just want to see an overview of the common patterns of activity (for example, Display > Display > Organic branded search Visit > Direct Visit > Conversion)? This has a significant impact on the nature of the process you need to do and how quickly you can generate reports.
What will be your 'unique key' for tracking a user? Often in practise, this may involve several keys depending on how different systems can be integrated. You probably have a third-party cookie with your ad server that you can use to match up users across platforms. Think about how you could connect this with the on-site tracking - perhaps by passing a first-party cookie through as a parameter in your click and conversion tags on your sites. Also consider whether it is feasible for your offline and online systems to pass data to each other. Don't get too bogged-down in the details of this - just consider any possible blocking points here and try to flag gaps.
What kind of data extracts are offered by your existing or potential analytics, search marketing and ad serving platforms? Can they offer automated scheduled extracts? Are they provided in a machine-readable format (such as plain, predictably structured CSV)? What is the data retention period?
What would be involved in getting tracking up and running to fill in your data gaps? Do you have a container tag such as the SuperTag implemented to provide reasonable turnover time on updates (timely with marketing campaigns, new landing page launches, etc.) and ensure that tracking covers your whole site? How confident are you that your business rules are reflected in the tracking implementation?
Perhaps most importantly: how confidently can you define your conversion points? Are you more concerned about driving traffic to your 'funnel start' point, or are you focused on which touchpoints actually lead to final sales, regardless of the funnel process? If you want to report on some combination of these, think about how they relate to each other, and consider defining a single metric as the 'canonical' conversion which can be the final go-to for determining the real cost and value of a touchpoint in a path of a user interacting with your business over time.

No amount of technical implementation will help if the goal of the project is not clear: understand the conversions and the conceptual points where value exists in your web presence and the goals of a 'Single Customer View' project will fall out of that.

by Daniel Bryan

Permalink | Leave a comment »