We developed NewsLynx within a short time frame and for specific research goals. If one were to continue building on its code base or design a similar system, these are new features and improvements we would recommend.
One difficulty we encountered was supporting the variety of external services that newsrooms use. Some of them even fell out of favor or came into vogue over the course of our research. Newsrooms have largely abandoned Google Alerts, for instance, first in favor of mention.net, which in turn was replaced by clip services such as Vocus and Meltwater. The latter is a paid service that returns higher-quality clip results and maintains a database of the circulations of those outlets—useful for determining the reach an article achieved when placed on a partner’s site.
Many of the current recipe sources are most useful for tracking influencers or specific groups of audience members. More external services would broaden the Approval River’s capabilities. For instance, one newsroom expressed an interest in monitoring Facebook groups it might create and tracking the “quality of discussion.” Phrases such as “let’s take this offline” or “do you have a suggestion?” could let an organization gauge whether its actions led to a self-sustaining, helpful community around an issue that its reporting highlighted. Integrating more offline-monitoring services would be useful, too, such as Sunlight Foundation’s Capitol Words API, which provides a queryable database of what is discussed on the floor of Congress.63 Participating newsrooms said that integrating more metadata around these citations would be helpful (e.g., aggregating Meltwater’s circulation figures for all pickups of a story). Many IEs currently export the data from Meltwater and do this aggregation in Excel.
Because new recipe sources will come in and out of fashion, an easy way to add and remove language-agnostic recipe modules would be the ideal.
One requested feature we were not able to implement in time was an impact event form submission page that staff could access without logging into the system. For example, a reporter who hears of a noteworthy event could fill out a form, which the Impact Editor could then approve or reject. CIR already uses a similar workflow with a form powered through Podio.64
Similarly, the more the platform could exist as part of an ecosystem and less of a portal you must log into, the better. This could look like automated emails, a mobile app to submit events, easily forwardable notes that create events, or push alerts to staff when a story surpasses some threshold and is “doing well.”
Currently, NewsLynx only supports articles published on a unique URL. It doesn’t distinguish between pages that contain text, video, or audio, for example. Being able to detect page types and assign specific metric collections would be a very interesting feature. Embed.ly currently provides analytics on video viewership behavior (e.g., how far into the video did most people watch).65viewership details into our existing comparative approach would be a useful addition. As analytics standards emerge in radio or for digital-broadcast hybrid organizations, these relationships will need to be formalized and operationalized.
NewsLynx users can currently group articles together with subject tags, but more complicated relationships exist in practice. For instance, subject tags don’t distinguish between topically associated articles and a specific package of articles that ran in a series. At broadcast organizations, you might have a digital story that ran as a companion to a TV or radio piece. Relating these different versions of articles to each other and then adding aggregate functions to combine the metrics of these two versions of the same story for reporting purposes would be a very useful feature.
Working with Google Analytics proved to be extremely frustrating. Each news organization often had a custom-defined property or view that represented its content. Often, it had multiple properties reflecting different subdomains that needed merging. Based on the advancements in the open source data pipelines discussed earlier, implementing something like Snowplow for quantitative analysis would make most sense going forward.