Subscribe to the Insightr Blog RSS feed


Insightr joins the Yahoo! Web Analytics Consultant Network (YWACN)

Today Yahoo! announced the creation of the Yahoo! Web Analytics Consultants Network (YWACN) via an announcement on Dennis Mortensen’s blog. The global network is made up of 48 agencies all of whom have specialist experience in web analytics and a commitment to helping clients make better decisions from their data.

Insightr is proud to be part of this new network, and as a result will be offering Yahoo! Web Analytics services to all of our clients. As a member of the Yahoo! Web Analytics Consultants Network (YWACN) Insightr will be able to offer clients access to the Yahoo! Web Analytics tools for free as part of our consulting engagements with clients.

Insightr will be offering this service to clients across the Asia Pacific region, from our office in Singapore starting today. Until today the Yahoo! Web Analytics tool has only been available to clients who were already licensing Indextools and for Yahoo! advertisers who requested access via their account manager.

We have been using Yahoo! Web Analytics here at Insightr since the Indextools acquisition and have been very impressed with the capabilities of the tool for organisations who do not have the budgets required to license a product such as Omniture Site Catalyst and for those businesses who have been using Google Analytics but would like to see more advanced functionality.

The team at Yahoo! led by my good friend Dennis Mortensen are very forward thinking in their approach to web analytics, this leadership team has been very supportive of bringing the Insightr team up to speed in the latest developments of the Yahoo! Web Analytics tool

Yahoo! Web Analytics is free like Google Analytics but some of its key differences make it much more similar to products like Omniture Site Catalyst.

Here are some of the features we’ve found to be most helpful in making decisions from the data:

  • Customisation of data collection. Unlike Google Analytics, Yahoo Web Analytics provides a very rich customisation platform including up to 15 custom action tags (think ‘goals’ for Google or ‘events’ for Omniture) in addition to custom content and commerce variables.
  • Bulk campaign upload service. In a similar way to how Omniture allows uploads of campaign tables via SAINT, Yahoo allows your campaign data to be uploaded to the system.
  • Custom Segmentation. Users are able to segment data at an individual level and can apply segmentations to all reports.
  • Alerts and Emails. The Yahoo analytics tool allows analysts to setup automated email report delivery (as with Google and Omniture) as well as creating rules based alerts based on data (currently offered by Omniture but not by Google)

As the leading digital analytics agency in Asia, Insightr will continue to recommend the most suitable analytics tools for its clients, but the advantage offered by Insightr with new tools and partnerships such as this demonstrates our commitment to continued thought leadership for and on behalf of our clients.

If you’re interested in learning more about Yahoo! Web Analytics and would like to speak to us, please get in touch with us here. We’d love to talk to you.

Thanks to Yahoo! for accepting us into the network and we will look forward to working much more closely with the Yahoo! teams in Asia supporting clients.


Market segmentation and why you should listen to your analyst

An organic farmer found that he was producing so much milk that the local dairy couldn’t cope with his supply. The dairy kindly suggested the farmer might want to make some milk related products and sell them from a stall at the farm.

The farmer started thinking, what should I do - sell wholesome fresh milk, make cheese, yogurt, butter - he decided there were too many choices. The farmer decided to hire a marketing consultant to determine what the market wanted from organic milk products.

The consultant interviewed visitors to the region, local residents and business people. With the data the consultant collected a market segmentation was prepared. It was determined that because there was a cheese shop in the town that also specialised in butter there was no extra demand for these artesan products, so the remaining market demand was identified for variations of pure fresh organic milk. The market was roughly split into two very different customer types:

  • Customer Type One wanted Ice Cold milk that would be partnered with juicy fresh strawberry’s from the neighbouring farm to make milkshakes.
  • Customer Type Two wanted Hot, Steamed milk that they would add to their Espressos from the neighbouring coffee shop in the village to make delicious espresso macchiatos.

The farmer took the analysis that the consultant had prepared, and after some review decided that there would be a nice mid-point between the two customer types and that with no extra effort he could simply meet the demand by selling milk at warm temperature. This way he wouldn’t need to spend the time preparing the warm and cold milk, the customers would be happy with this he thought.

Three months into the new milk stall, the farmer sold none of his extra milk.

Lesson: listen to your analyst when you are presented with a market segmentation it might be make or break for your business.


Research from Atlas Institute proposes digital planners use TRPs as key planning metric

In a recent research paper by Microsoft’s Atlas Institute, Microsoft propose digital media planners leverage a TRP (Target Rating Point) formula to create a better proxy for traditional media clients to understand the relative performance of digital display advertising compared with traditional channels.

A compelling reason for agreeing with the research is presented in the executive summary:

Existing media planning tools may help identify sites that have the highest concentration of a target audience, but digital publishers write contracts in terms of ad impressions. So brand marketers are left to guess if the plan that nets out to an $18 CPM is better or worse than the $9 CPM plan.

How true!

The paper criticises the use of direct response metrics in brand advertising campaigns (as Comscore did with their research) arguing that a better formula is needed to help those responsible for ad budgets to make better decisions for media.

Once a planner uses a TRP metric, it can then be shared that - of course - using an interactive medium provides a multi-way dialogue to commence after the initial impressions. For this we think of the more direct response side of things - websites, order forms, shopping carts etc - but also the conversations that can commence in social media following the launch of a brand campaign.

Microsoft have published their formula for TRP’s as shown below:

Here’s a direct link to the pdf research paper.

What do you think about TRP’s? Do they make sense? Do you think they’ll help you get a larger piece of the advertising budget assigned to digital?


Research from Comscore and OPA demonstrates lift in branded enagement from exposure without clicks

A great research project completed last month between Comscore and the OPA (Online Publishers Association) has gone a long way to proving that digital display advertising has a strong brand engagement lift when analysed against non exposed control groups. The data shows a lift in basic web engagement metrics (time on site, pages per visit) as well as conversion metrics (spend per visitor). All the research was carried out without analysing any click-throughs.

It’s ground breaking in my mind as advertisers have been struggling to justify spend in display advertising by using click and conversion based metrics historically. The view-through conversion has been doubted and debated, particularly by search marketers who are fortunate enough to win credit for most web based conversion because of old-school last click attribution models.

Of course companies like Dynamic Logic have been sharing brand metrics through their market norms database for as long as I can remember - but those metrics largely focus on the more traditional brand trackers (aided / unaided awareness, purchase intent etc). This research by Comscore goes beyond that to help shed some light on the old view-through data that has been so heavily debated.

Research like this should generate further studies, but it’s likely to have a very positive impact on the digital advertising industry as it’s demonstrating that display media can be analysed, and proved effective (against the non-exposed control group) without looking at the click as a success metric.

This research goes a long way to supporting the arguments set forth by Young Bean Song from Microsoft Advertising in his blog article “Getting Back to Basics – Why Web Advertising Needs Traditional Media Metrics” from 6th July where he argues for the need for consistency in digital and traditional metrics and why reach, grp’s and frequency are important if the share of advertising pie is to grow:

Digital folks snicker when they hear advertisers make statements like “TV works”. Turns out, TV does work and there is plenty of quantitative proof that TV advertising drives sales. As much as digital marketers love to carry the ROI torch, what they don’t realize is that traditional marketers live and die by the same sword.

The presentation is available on Slideshare, and is embedded below:

Is the display ad efficient ? A study from the OPA



Guidelines for ethical and effective Word of Mouth Marketing from WOMMA

WOMMA have created a fun video that shares some useful best practice guidelines on how to create ethical and effective word of mouth marketing programmes. The background behind this video stems from the fact that the FTC in the US are planning to monitor blogs and the like for unethical marketing practices*, WOMMA recognised the importance of reacting to this and is offering a simple framework they call “Don’t Tell, Do Ask” to help marketers address the issues.

“If you walk into a department store, you know the (sales) clerk is a clerk. Online, if you think that somebody is providing you with independent advice and … they have an economic motive for what they’re saying, that’s information a consumer should know.”
Rich Cleland, Assistant Director in the FTC’s division of advertising practices.

Now, the FTC issues won’t apply to marketers outside of the US, nonetheless it’s very important to use ethical standards in one of the most trusted and effective forms of marketing, so to that end these guidelines should be considered universal.

A useful 3 minutes is spent watching this:

*basically the FTC plan on monitoring blogs for fake claims, payments etc (see article here)


How to guide for using Events in Advanced Segments for Google Analytics

I’m not sure if I missed the official announcement, but I discovered the ability to add Event based dimensions and metrics to the Advanced Segments features of Google Analytics.

The enabing of this feature by the Google product team finally makes using Event Tracking an integral part of your analytics.

Event tracking in Google AnalyticsIn the past there was a large problem with Event Tracking, that very few people talked about surprisingly. See here’s the problem, when event tracking was first enabled it was heralded as the solution to fix issues with (eg) pageless websites, flash content - you could ‘enable event tracking without inflating your page view count’. Good stuff? Well, in a way yes - the problem with this solution was that the data was held in it’s own walled garden disconnected from other data.

A classic problem was that, let’s say you tracked all of your video with event tracking.

Implementation was improved - the multi-dimensional capabilities made it very easy for developers to code in labels, categories and actions, and then we’d have a nice way of drilling into our data to see how the different dimensions interacted.

Analysis wasn’t improved - But what about conversion? Everyone in the industry voices strong opinions (rightly so) about how we need to keep an eye on goals, registrations, downloads, orders, leads etc - but with event tracking this simply wasn’t possible:

Event Label Reporting in Google Analytics

See, that’s a pretty report? But it doesn’t tell us very much of anything useful. In the example above I have set event labels for video starts and completes - I can take this data out to Excel and calculate a video completion rate but beyond that the data is not much more than some eye candy. Without full integration to other data it’s kind of useless.

This is why we frequently cause confusion to developers by asking them to track non-page view events as page views (eg pdf download link) so we can use ‘events’ as goals. To a developer who is following a design schema this requirement makes no sense, and neither should it. We were hacking Google Analytics because of limitations like this.

So what have Google done? It’s now possible to access the Advanced Segments tool and access event based dimensions and metrics as part of your filter creation. In the example below I’ve created a simple video completed segment:

Creating Event Segments in Google Analytics

Now that we’ve looked at some of the problems of event tracking in the past, and looked at the new advanced segments capabilties for introducing events to the process it’s time to take a look at what we can do with the segments.

In the example below, for a different client with the same type of segment, I’ve applied the video completion segment and the all visits (control) segment to a goal conversion report. This is telling me that we appear to more than double the form completion rate when visitors complete a video.

Analysing Event based Segments - Goal Conversion

Testing Statistical Significance for our SegmentA quick test for statistical significance (using a homegrown Excel based significance calculator) tells us that this data is significant at 99%.

This is a good thing. We can now hypothesise that making an investment in optimising our web based video will lead to increased visitor conversion. From here we can develop test hypotheses around how and where video is placed and positioned around the site so we can develop a multi-variate test to optimise the number of visitors viewing and completing the video. It’s probably too small to see - but we have over 1 million visitors, of which only 20,000 completed the video. This is a huge opportunity!

With this in mind, here are a few other ideas on how you could use events in advanced segments to make your analytics actionable:

  • Create a segment to benchmark event completion (eg video completions) with other segments to help appraise the value of content and features.
  • Create an event based segment to allow events to be used to in goal completion analysis
  • Analyse multiple goal types in segments (eg game play, video start, widget click, ajax error) to be cross tabbed to determine their contribution to overall success

I hope this was useful. Please let me know if you find new and interesting ways to make your events, segments and goals data actionable.


A framework for search engine marketing research measurement and optimisation

This is a revision to a presentation I uploaded to Slideshare a couple of years ago. It’s been an interesting couple of years for digital marketers who want to make use of free tools and technologies to assist their strategies.

The updated presentation shows how to make use of a number of free tools, including Quantcast, Google Adwords, Google Analytics and Google Website Optimiser to research, plan, measure and optimise your search engine marketing. It’s not meant to be a replacement for the best practices you have already developed, but rather a toolkit to help you make better use of data for research and management of search campaigns.

I’ve also been experimenting with Slideshare’s new content feature, so have also included a YouTube video from the Google Website Optimiser team on how to integrate Google Analytics and Website Optimiser in 60 seconds.

View more documents from Insightr Consulting.

A framework for SEM research measurement and optimisation

Note: I’m not advocating only using Google tools as there are plenty of other very good (and more powerful) tools in the market - such as those offered by Yahoo!, Omniture and Webtrends - but the key benefit for marketers in using Google tools is that they are free. This is incredible beneficial as you can be up and running with campaigns after just a couple of clicks - no contract negotiations.


Making sense of the Google Analytics Motion Charts feature

It’s very cool how Google continues to innovate their Google Analytics product, but one of the things I hear as I advocate use of the tool is that it’s not always clear exactly how to use a new feature. Today I want to run through a usage of the new motion charts feature available in most reports.

When Google launched this feature I had been lucky enough to be part of the private beta, so had been given plenty of opportunity to learn how the tool could be used to help with the data decision making process; unfortunately when Google launched the feature to all users I was a little disappointed to see some examples presented on the help website that while interesting to watch, probably weren’t the typical kinds of analyses that a marketer will be considering of the tool. For example this blog entry discusses how to visualise Browser and OS data - cool enough, but practicable in the real world? Perhaps not so..

I want to share today an example of how we used motion charts to simplify the presentation of campaign data to a client, and how that data was used to generate hypotheses that we later used in landing page tests and to optimize the media buy.

Client Background

This client was a CPG brand who have a major campaign push in the first half of each year when demand for products is highest. There is a limited window of opportunity for driving brand awareness, and driving digital intent to later translate into in-store sales over a 2 month period from January through March. Most of this brand’s annual sales are driven in this short window in a very competitive landscape.

Analysis Objective

Our role was to analyse the performance of the brand website at driving visitor registration and loyalty for the products, in doing so we were tasked with a multi-channel ‘beyond the click’ analysis that needed to address the following channels: display advertising, paid search, organic search, social media referrals, direct load (post impression visits) and DRTV (through marketing urls). With this in mind we were responsible for developing an agile analysis and optimisation process that meant we were to analyse data against the business goals, generate test hypotheses and rollout tests in a very short period of time.

The trend showed the campaign was performing well across all channelsThe media agency was responsible for paid media optimisation (primarily through their own analysis), but as our data showed there was a considerable amount of organic conversion that happened through direct visits, referrals and organic search. To assist with this we did time series analysis against this data and the media plan to assess paid media’s impact on organic traffic though a standard regression model.

Addressing multi channel analysis conflict

As we were tasked with being agile in our analysis, there was no point waiting for data to come in at the end of the campaign and performing a long drawn out analysis. Our initial plan was to develop a simple Excel based dashboard that took in weekly data from the media agency and to integrate with our Google Analytics data. This idea was well received by all, until the first presentation of the data.

The first dashboard was mocked up, data collected and presented at our first weekly status; unfortunately this meeting was lost in translation as the media agency felt a desire to defend a poorly performing banner campaign - despite the fact that our regression based data showed a (albeit weak) positive correlation between organic search and direct load visits and banner flighting. The dashboard was not discussed.

We realised that we needed a much more visual way to present and share the data - particularly as we were dealing with multiple dimensions used in our hypothesis generation - visits, bounce rate, segmented visits for existing registered users, segmented conversion rates etc were all used in the preparation of the dashboard - the intent being to simplify the data. Sounds simple enough?

Presenting multi-dimensional data

To assist with making sense of the data we turned to the Google Analytics motion charts - we felt that a visual representation of what was happening in the campaign across the (paid and organic) channels would get a better response.

For our consultants this turned a present-on-the-screen dashboard into a log-into-Google-Analytics-and-present-real-data process, very much like giving a pre-sales demo of an analytics tool to a client. Of course there was room for errors, what if the wrong data was presented? what if the client wasn’t interested in seeing the ‘raw numbers’? what if it raised more questions than it answered?

Silly questions - an analyst who knows their data is always going to be in a stronger position logging into a tool and presenting a real-time-story of the data to a client (provided that a conclusion is prepared and an action plan ready for discussion). The client loved the presentation of the motion chart, and as a result were were able to achieve the following actions from the presentation of just one chart:

  • Re-allocation of budget from display advertising to paid search agreed
  • Recommendations for landing page testing on a per channel basis agreed
  • DRTV recognised as best performing channel (despite reservations from client)
  • Organic sources (search, referral and direct load) recognised as key players in the mix for acquisition

A further analysis against a segment of existing registered customers, showed that the display advertising was doing a tremendous job at re-activating last year’s customers. Further analysis was to be carried out from the media agency on the ROI of re-activation.

Using Google Analytics motion charts:

Our interpretation of the data helped us to:

  • Compare bounce rate across channels alongside registration conversion.
  • Demonstrate that lower levels of (optimised) banner placements in the last 3 weeks of the analysis drove much higher conversion.
  • Identify organic media as channels driving low-bounce, high-visit and high-conversion.
  • Demonstrate the low-volume but high-conversion performance of DRTV.

Now, as mentioned earlier the use of motion charts was not the only analysis we provided in the campaign, but we used is as a catalyst to drive action from the data for media re-allocation, optimisation and testing.

How are you using motion charts in your analysis?


Mapping the Social Graph and Data Implications for Marketers

This is a quick follow up to my earlier blog article on Visualising Facebook Friends, I had some interesting conversations with people today asking more about the mapping of the Social Graph and what this might mean for a marketer who wants to start enriching their customer database (or even to build out a social media applications of available data).

This got me thinking, how can I represent my thoughts without spending an hour on my whiteboard, taking a photo, posting it here and then trying to annotate my thoughts! Luckily as I was exploring ways to leverage the Google Analytics API this afternoon I happened upon a new API Google have in their labs called the Social Graph API - a new data extraction API that allows developers to access data that Google has been indexing over the last few years and not necessarily using. There are two key sources of data that Google uses - XHTML Friends Network (XFN), Friend of a Friend (FOAF) markup - which many social network sites are already including in their source code.

So for example the source code of twitter shows me as a contact of @mashable with the following XFN markup:

<span class="vcard">
<a href="/mashable" class="url" hreflang="en" rel="contact" title="Pete Cashmore">
<img alt="Pete Cashmore" class="photo fn" height="24" src="..." width="24" />

Using the API a developer can then create an application to effectively map out the relationships between friends and their friends thereby creating something (that has always been possible within the walled garden of say Facebook via FBML) that is platform agnostic and part of the public web. Thus the spidering effect of mapping the social graph allows us to quickly build up a powerful (and relational) map of personas.

If you take away the documents, you’re left with the connections between people. Information about the public connections between people is really useful… There hasn’t been a good way to access this information. The Social Graph API now makes information about the public connections between people on the Web.
Google Social Graph API documentation

With this in mind there are clearly a number of very interesting social networking applications that could be built off this - for example a friend discovery solution that crawls friends to identify if any of my friends are part of that network. This would work very well for me - I have strong LinkedIn and Facebook profiles, but very few Xbox Live friends because it seems I’m unable to find like minded friends on Xbox!

So, for the marketer being able to collect data from customers, who for example share their blog url or their twitter account this can be a great way to start leveraging the social graph for better friend based conversations, and to use friend based data as an enagement metric when incorporated with (for example) advocacy explored through text analysis - how many of your customers are talking about your products in social networks?

This video from Google Engineer Brad Fitzpatrick gives a really useful introduction to the power of the API for developers. For marketers the opportunities for data collection are equally fantastic:


Visualising Facebook Friends

I’m always delighted when I find an interesting data visualisation, so when I chanced upon a new Facebook application this morning that allowed me to map my friends network I couldn’t resist. Using the Facbook social graph api and a Google Maps mashup the Friends Density application creates a fun way to look at where your friends are based (note it’s not 100% accurate as a lot of my New York friends were still showing up under their university location..).

I wasn’t surprised to see that my friends are clustered around the 4 cities around the world where I have been working over the last decade or so (I included San Francisco as a working city since I spent a lot of time on the west coast when working from New York, despite never having lived there).

A global network of Facebook friendsNow, although this application is fun for me at the moment, I think it helps us to understand something of the networked nature of our connected lives. There are some interesting measurement implications of this networked view of our touches:

  • International Audiences. As our networks grow around us, and as we become more influenced by those we have relationships with, there is an increased likelihood that your campaigns and digital content is increasingly likely to be consumed by a more international audience. When you look at your campaign data are you wondering about where your international visitors are coming from? Think about digital word of mouth as a started..
  • Networked Society. As organisations start to evaluate social media tactics, there needs to be an increased emphasis on measurement of the network influences. As you segment your customers, are you also finding ways to incorporate their influence as a measurement dimension? What data points exist in social media that can be used to extend and model data provided through purchase and loyalty? Can you prepare look-a-likes based on mapping existing customers behaviour in social media versus those who attribute similar behaviour? Can you use this insight to more accurately target those likely to respond?
  • Digitisation of Media. As mainstream media becomes increasingly digital the work of marketers needing to manage multiple channels, multiple conversations and multiple ways of connecting with customers and selling to customers we will need to ensure our analysis and strategies built off the huge quantities of data are both understandable and most importantly actionable. I’ve been speaking with many of the leaders of the analytics and campaign software solutions world in Asia over the past few weeks and one of the common conversational threads is about how can we streamline the analytics process, so that acting upon data is less of a chore (and cost centre) and more of a workflow process whereby standard actions and activities are formatted into the marketing business. We’re currently investigating ways to achieve this and to improve the efficiency of digital analytics through developing such workflows.

These are interesting times for those working in digital media, and as I see applications like this open up in Facebook it’s very exciting to imagine how the analytics technologires will improve their data visualisation techniques, helping marketers to act upon data with increasing confidence and automation.