This one of three articles in this month’s issue that report on AMEC’s recent annual Global Measurement and Evaluation Summit:
AMEC’s annual Global Measurement and Evaluation Summit took place virtually last month. It was a real eye-opener, driving home the extent to which technological advances are changing the face of measurement. I can’t recall any PR summit—ever, over the course of my quarter century of attending several each year—that included as many technology-focused presentations. Speakers covered everything from new formulas for calculating engagement to how to turn deep fakes to your advantage. Here are summaries of five of the most impressive presentations:
1. Finally, an alternative to impressions, UTM, and site-level readership
Among the more intriguing technology presentations was from Maya Kaleva, head of research and insights at Commetric, and Rune Cleveland, CTO of Opoint Technology. They presented an interesting new approach to determine readership on an article level. (See their presentation “Estimating Article Readership.”)
We all know the problem with site-level metrics is that they generate large numbers representing the entire news site—and conveniently ignore factors that impact exactly how many people actually see your particular news item or article. Factors like:
- Where on the site an article appears. Is it on the front page or hidden in a subsection?
- How many posts does the outlet publish a month? An article in a focused publication that publishes 20 articles a month gets a lot more attention than one on a site like Bloomberg, which publishes 10,000 posts a month.
OPoint’s solution to this problem is a formula based on readily available data from SimilarWeb. SimilarWeb scrapes news websites for basic web analytics data to provide average visit duration, pages per visit, and bounce rate. Based on that data you can calculate the likelihood that someone actually saw a story. That calculation looks like this:
They make an interesting argument for the value of looking at reader time and engagement with an article as a measure of impact. Their premise is that the more engagement an article has, the longer it will stay on a page to drive more attention, therefore the more likely it is that it will be seen.
So for every story that OPoint captures you can see the calculated estimate of how many people have read it so far, the number of new readers per hour, and the final number before the post is no longer visible on that site. It also reveals the number of shares and likes, so you can see the amplification. Opoint’s system also reveals which reporters and editors are most likely to author the posts that drive that impact.
Commetric has been using these calculations with clients to report which stories are actually drawing reader attention and driving impact. According to Maya Koleva, the reception has been great:
“Most [clients] understand straight away that looking at SimilarWeb or similar stats on domain level is not helpful, e.g. in 9 out of 10 something akin to Yahoo! Finance news or MSN news would end up top by monthly visitors, just because those sites host news, search engine, are default homepages, email client, etc – but they just syndicate news from elsewhere. Then when I show the estimated article readership per article from such online outlets it is is blatantly obvious why the data point needs to be more accurate and article-centric rather than domain.”
2. Clustering analysis to figure out your brand’s voice print
Alexander Rose of Data Analytics discussed how he used a text mining system called Tidy Text to analyze coverage of brands to tease out their “voice,” as well as customer perceptions of that voice and their competitive differentiation. (See “Opinionated Text Analysis: TidyText for Social Listening, Analyzing Brand Voice, Understanding Sentiment, and More…”)
Rose looked at the discussion around financial services company Robinhood and the GameStop controversy. He analyzed the frequency and correlations of terms and sentiment to map the conversation and to create what he calls a “voice print.”
Rose’s voice print determines how frequently a brand is described using particular words. Voice prints allow comparisons between brands. Rose identified groups of words like “lawsuit,” “class action,” “stock price,” and “manipulation” that can be used to warn a company of potential crises. His conclusion is that one of the reasons why everybody was so mad at Robinhood was that it had violated its brand promise and upset its user base.
3. Share of search, the hot new PR metric
Among my personal favorites was the idea of using share of search as a proxy for awareness or interest. James Crawford, managing director of PR Agency One, and Stella Bayles, director of Coverage Book, showed how you can use Google Trends to find your share of search relative to the competition. (See his presentation “Beyond Share of Search – What Matters for Public Relations.”)
It’s far from perfect—especially if your brand is the fintech company Nutmeg and you’re turning up results for pumpkin spice recipes. But Google Trends is free, the calculation is pretty simple, and, at least theoretically, if people are searching for your brand, then they have to be aware of it. There is also some evidence that share of search correlates to a greater share of traffic and market share.
4. Deep fakes don’t have to be terrifying
One of the more intriguing presentations was from Daniel Fountenberry, a partner in Videologic. He showed how deep fakes can be used to transform communications strategies and generate characters that aren’t real, but are perceived to be “authentic.” (See his presentation “Beyond Deep Fakes.”)
5. omniearnedID: Yet another proprietary system to determine attribution
Omnicom’s Erin Lanuti and Ketchum’s Mary Elizabeth Germaine discussed an interesting new integrated system that combines multiple data points to create an “omniearnedID.” (See the presentation “New Frontier In Earned Attribution.”) It is essentially a way to identify the reading habits and characteristics of any client’s audience so you can match them to exactly the right influencer.
This presentation was among the most controversial, due to both privacy concerns and the proprietary nature of the metric. The case study claims an 8% lift in sales as a result of the system’s ability to reach people in less saturated media. Showing sales lift and $3.73 in sales per working dollar is a lot better than metrics like AVEs and impressions to measure influencers’ campaigns. We welcome this addition to the tool kit. ∞