Methodology
In general terms the project was to see if we could apply the same principles to web audiences at a local level that JICREG has applied to local print for the past 20 years. Those principles are:
- Jicreg only accepts titles with audited circulation data, thus providing an independent foundation for the data.
- Where readership research is undertaken we use this to create models for Readers-per-copy (RPC) by publication type
- The research does not have to be continuous as we can update the database using revised circulation data and the RPCs
- By collecting distribution breakdowns from the publishers we can allocate the readership at a postcode sector level
- And, finally, by using the models, we can estimate audiences for the un-researched titles thus providing a comprehensive, local planning database.
We needed a similar approach for web data. Audience data for local media websites would be based on: audited web traffic data; survey data from over 200 surveys (combination of face-to-face and new online surveys) with samples totalling over 100,000 adults; and statistical analysis and modelling.
There were six stages to the project:
- Establish the universe (build structure of Media Portfolio database)
- Populate the Media Portfolio database
- Collate readership and online survey data
- Analyse internet audiences
- Analyse internet/newspaper relationship and duplication
- Integrate with JICREG to provide planning tool
First we needed an equivalent to circulation, and we decided to use the industry norm of Unique Browsers. We chose Monthly Unique Browsers as our base since most local online campaigns are monthly.
As with print, we needed a base of independently audited web traffic data. It would have been cost-prohibitive for every site to be audited, particularly for larger groups with hundreds of sites. So it was agreed that:
- Any site that was to be included on JICREG as part of the Locally Connected project had to be registered with ABC Electronic or other approved independent auditor.
- Publishers would need to have their top 10% (or top 10) sites audited, along with the whole network, every 6-months as a minimum. The same web analytics software and settings would be used for all sites on the network.
- Any registered but un-audited site could be retrospectively audited at JICREG’s request.
Seven of the larger publishing groups signed up which covers around 70% of the local media market. More are expected to join.
The audience research surveys used in the project provide national coverage of the UK, with an estimated combined sample size of over 100,000 adults – believed to be the largest research on internet usage available in the UK. Data from over 100 face-to-face surveys was loaded onto a special automated system to help us understand the relationship between newspapers and their companion websites. But it was new internet audience research data from Survey Interactive’s surveys for Trinity Mirror, Johnston Press and Northcliffe which was critical to the project. Other groups such as Newsquest are conducting similar research.
This provided site specific data that comprised Adult audience with both demographic and geographic breakdown. In addition they measured the readership of the local newspapers so that the duplication with the web audiences could be ascertained.
To date we only have such data from Survey Interactive but we know from other countries that similar data is also now available from other research suppliers. In exactly the same way as we do for print research JICREG has established guidelines for internet audience research. Providing the research meets these guidelines it will be accepted for inclusion on JICREG and used in future modelling.
Five models have been developed to enable the full planning system. This is a summary of these models, which were based largely around regression analysis, and all of which have been independently and positively appraised by Ken Baker Associates.
The first model was to enable us to estimate Adult unduplicated audience from the monthly Unique Browser figure for the site.
The second model is used to estimate the size of the Adult Audience that falls within the Business Area of the web site. We define the Business Area as the distribution area for the regional newspaper or newspapers associated with the site. (It is worth noting here that the extent of out-of-area usage of the web site can be shown to be related to such things as the level of ex-pats (for Scotland in particular) and the presence of a Premiership soccer club: both of these factors increasing the out-of area usage of the website.)
The following graph shows how these first two models combine to provide an estimate of the adult audience in the business area. The model is compared to the data from the surveys.
The third model was to predict the demographic breakdown of the web audience. After many analyses it was found that relating the web audience to the newspaper audience was the best approach as that helped eliminate the fundamental differences in the demographics for different geographic areas.
You will note the high index for upmarket and middle aged demographic groups.
The next model is in two parts, and estimates the weekly and daily audiences from the monthly audience. By estimating these we can provide growth curves that enable planning systems to estimate reach over differing time periods.
The final model involves modelling the duplication between the web audience and the audience of the associated newspapers. Experience led us to decide to model the index of duplication compared to random duplication and as you can see from the following graph which compares the estimated and surveyed de-duplicated audience of the site and its associated newspaper, this worked very well.
