John D'Arcy

John's bio

John D’Arcy is a Lead Consultant at Foviance and has helped clients measure and optimise their marketing communications for over 15 years and is an authority on web analytics, data visualisation and statistical modelling. He has used techniques such as segmentation, targeting and econometric modelling to drive increases in marketing ROI in the Media, Technology, Automotive, Financial and FMCG industries. John’s analysis has been used for projects as diverse as developing the measurement strategy for the launch of a major online store to the optimisation of pan-European automotive sales regions. John’s analytics teams have consistently uplifted their client’s revenue by over £1 million each month.

This article, written by John D’Arcy, was originally published on Clickz.com on 02/05/2012 and is republished here with permission.

Dashboards, infographics, and visualisations have been on my mind a lot this year. I’ve always loved turning data into pretty colors and patterns, which is great as the number of dashboard projects for consultants like me has gone through the roof in the past couple of years.

The projects I’ve worked on, even just in 2012, have been at times affirming as an analyst, have driven some key business relationships, but sometimes have been very frustrating. I’ve been made to feel like Leonardo da Vinci when in one project I simply merged two data sets with a customer ID, drew a two axis line chart, and saw the client nearly fall off their chair in amazement. But I’ve also been involved in projects that made me feel like Leonardo the Turtle, wading through a sewer of data and misunderstood objectives.

It’s made me think a lot this year about the responsibilities and checklists from both a client and agency perspective. To get the best out of your dashboard project, whether it has two data sources or 202, here are a few key learnings and things to think about for both sides.

For agencies or those internals teams tasked with building the tool:

• Be a Doubting Thomas and don’t scope out your project until you physically see the data.
  Clients love to say to you, “Here’s my data, I just want you to pop that in a beautifully visualized set of charts.” But under the bonnet of any tool, web analytics, sales, or customer relationship management (CRM) could be a whole mess of data that will take weeks to get to a point where you can easily visualise. This is particularly true of multi-data source dashboards, say ad serving and web analytics data. If you don’t have a long-term relationship with your client and they want you to scope without seeing the data then walk away now.
• Deliver a data audit.
  This should include profiling the data you receive so that you can understand underlying data distributions and also a quality assurance review of calculated metrics. Too often clients believe their data house is in order and when resulting visualisations don’t look the way they want, the person building the dashboard gets blamed.
• Visualisation is less important than data structure.
  Did I really just say that? Stop scratching your head deciding if you agree with Stephen Few about pie charts and start thinking about the database or data tables you are storing your data in. Does it need to be based on a data cube? Are you likely to have to build a gazillion summary tables? Will your dashboard have to read new data sources in the future and is your data layer flexible enough to do that quickly?

Learnings for clients or the teams, like marketing, who are going to be using the tool:

• Beware the funky white elephant.
  “Sex sells” and you want a bit of wow factor to impress your boss. But your company is already littered with reporting packages. Ensure that the visualisation tool that you use can be shared across a number of users and that internal security settings won’t mess up everyone being able to view the pretty Flash charts. To be frank, if you are impressed by a moving chart, then you shouldn’t be the person deciding what software to use.
• The value of the data layer is probably 10 times the value of the dashboard.
  If you are merging multiple data sets and building a multi-channel dashboard, you are very likely to be pooling data that has previously been sat in disparate data silos. If this is the case, the data layer you get built for you, say a SQL database or even a set of Excel workbooks, is going to contain data over and above that which is summarised in your visual layer. You must get ready to mine this data like you would have mined each data source separately. Do you have the people in place to do that?
• Cost of ownership is likely to be way more than cost of development.
  If you haven’t thought yet about who is going to own, run, and analyse your dashboard then stop now.

I haven’t even begun to talk about which is the right visualisation tool or what metrics are the right ones to include. From a structural perspective, data is a messy plate of spaghetti. But if you are prepared to audit properly, spend time developing the right metrics, and ensure there is an after-life for your dashboards then the visual impact will be dwarfed by the business impact from all your hard work.

This article, written by John D’Arcy, was originally published on Clickz.com on 04/04/2012 and is republished here with permission.

A few weeks ago when playing around with voice of the customer data for a client, I found a great stat that their customer satisfaction score highly correlated to the day of the week. People’s satisfaction with the website and products varied significantly depending on the day that they filled in the survey. Can you guess the day of the week that produced the lowest scores? People hate Mondays. And they were taking that fact out on our survey. More people were satisfied on Thursdays, Fridays, and Saturdays before the rate started a depressing decline to a Monday low.

The rate on Mondays was really dragging down scores for overall satisfaction and NPS (Net Promoter Score). It almost felt that people’s Monday blues in getting back to work after a great weekend was biasing the results. This was a new revelation as the data had only ever been provided at an aggregated monthly level before. We extracted two years’ worth of data and did some stats tests to prove the data was unbiased; we proved that there were no outliers producing false averages and that the differences we were seeing in the scores were statistically significant.

We created a nice data visualisation showing smiley happy customer icons running up to the weekend and sad depressives on Monday. We played around with the axis to highlight the differences while ensuring we didn’t over-egg the results. The visualisation generated lots of hilarity, engagement, and debate. Were people just really hacked off on a weekend? If you asked the same person the same question on a Thursday would they reply in a different way than on a Monday? Was it that just moody people answered surveys on a Monday? Would we see this trend for other brands? Were happy people off spreading the love and didn’t have time to answer surveys until the end of the working week? Was this actually useful for our client?

We didn’t know the answers to those questions. The services our client sold need to be available all week; it wasn’t the sort of product or brand where communications tailored for the day of week would be appropriate. It was daft of us to recommend only launching new products on a Friday. The client couldn’t tailor products to the day of the week. We had a great, funny 15-minute debate with the client; he was more engaged with the results than we have seen for a while. And then we wondered what to do next.

But there were definite learnings from this exercise:

1. Useless but true isn’t always bad. Rightly so we try and live by the mantra of focusing on “actionable analytics” but sometimes it’s fun to dig into data to find trends that wow people – even if you can’t get your head around how to take action on it. We don’t often get a chance to be playful and that’s something as analysts we shouldn’t lose or we risk delivering staid analysis. In this case we found how to present a regular report in a new way. And, that kept our client engaged and assured they weren’t looking at the same report we’d shown them last month.

2. The power of analysis is in the detail. Once we got our hands on daily data, we could find a lot more information. With any piece of analysis, get down to the lowest level you can. Just running some distribution and x-tab charts across the whole data set didn’t just uncover a single funny trend. We now know to check there are no biases in the future. More importantly though, by understanding all the variations in the data set an analyst has more confidence in the overall results. This leads to being able to be more forceful with recommendations and so the analysis becomes more actionable.

3. No surprise that visualising these distributions was key to getting people debating. The visualisation was simple and to the point. It made one statement, “I don’t like Mondays” and triggered a debate. Everyone got engaged because everyone could imagine themselves in that situation saying, “Yeah, I tend to be in a bad mood on a Monday too.”

4. Finally: Is this a trend for every type of survey? I’ve seen other analysis that backs up the distribution we saw. Why not try it out by asking for a pay raise on a Friday? Your boss is probably in a better mood and will be more satisfied with your overall work!

This article, written by John D’Arcy, was originally published on Clickz.com on 21/03/2012 and is republished here with permission.

2012 will be the year of the election, you couldn’t win without “Big Data.” It will be the year when Big Data could help you increase your margin by 60%. Big Data also means I could track everything my customers do and store it forever. Big Data sounds brilliant doesn’t it? But admit it; we still don’t really know what it practically means for most of us.

When I first heard the phrase “Big Data” a year ago I didn’t think, “Hey, that sounds cool.” I actually cynically thought, “Here we go again. The big technology companies have thought up a new tag line that will hook us all into thinking we need to replace all our old systems with new, faster, expensive servers.” I worried that the analytics industry was about to go down another cycle of technology-led implementations rather than thinking about how to use data we already have in an imaginative, insightful way. But there’s still time for us to grab the phrase and define what it should mean. Here are some thoughts: (more…)

Foviance is delighted to have been recognised as a Google Analytics Certified Partner (GACP), achieving company level accreditation to package, sell and deliver analytics with Google’s trusted badge of certification.

Google Analytics is the leading free-to-use enterprise-class web analytics solution. In the right hands it will provide insights into website traffic and marketing effectiveness. It’s powerful, flexible and simple to understand, ensuring businesses are in a better position to target advertising, strengthen marketing campaigns and improve those all important conversion rates.

Of course Google Analytics can be implemented out-of-the-box, but that level of solution would not leverage the power of the data that our customers require. Over the last few years our consultants have demonstrated an advanced ability to use Google Analytics to help improve our clients’ website conversion rates and ultimately their profitability. Our detailed knowledge of implementing and using Google’s tools frequently help us to optimise the ROI of those customer brands.
(more…)

We’ve had a stats love-in at Foviance this week, with a gathering to review Hans Rosling’s recent BBC4 program The Joy of Stats and a lunchtime debate about data, statistics and the role they play in understanding our world.
(more…)

There’ve been two news worthy football events in the last week. The first was my return after 2 years ‘resting’ to my local 5-aside league, the second was the return to our TV screens of Uefa’s Champions League. While I stood around in the pouring rain wishing I was a decade younger on Tuesday, on Wednesday I rested my aching legs on the sofa and watched Arsenal demolish Braga. Grudgingly I’ll have to admit they were quite good.

Given my interest in football, advertising and data I often think about the numbers behind the game. Sometime that is match statistics, sometimes it is financial data. (more…)