The recurrence of a particular phrase at most digital media huddles was almost certain. One digital media “pro” would sip of his drink, holding his devoted audience with a levitated glance, and say at the first opportune moment, “I guess apps are the savior of digital journalism!”
So an agency would be hired at the cost of about a half-a-dollar per app download, and a big part of the contract would be delivered to countries that have nothing to do with the news brand. And very soon the gap between app “lifetime installation number” and the number of “active users” would start widening.
Managers who approved such campaigns would leave before CEOs became suspicious. They would choose greener pastures such as programmatic or Artificial Intelligence (AI), where far fewer people would voice their suspicions over the next few years. Emerging fields always offer the benefit of doubt on questionable performances and are dear to over-zealous recruiters.
The agencies typically explained to their clients how they use a data management platform (DMP) to get the interest-based audience information. They also explained how that information is enriched by data from third parties (which happens to be data mostly sold by mobile service providers, legality unexplored), and how that audience information would be cloned and re-cloned until the target download value was achieved.
All this, despite sounding so spectacularly high-tech. In effect, the results they delivered were akin to air-dropping bundles of Indian newspapers in Myanmar or Tanzania — where, besides the language, nothing else is on target.
So how do you identify the right audience for targeting?
If you care for your company’s advertising dollars, my first humble suggestion is this. Before buying behaviour-matched user data from questionable sources on DMPs or targeting clones on Facebook, look more into the audience data you generate — from your analytics tools, user engagement activities of the print brands, print circulation, commenting tools, and similar places.
We need to realise behaviour-matching and cloning on unknown audiences work well with non-media products and also with media products that have universal appeal and exceptional brand awareness, such as Netflix or Amazon Prime Video.
But legacy news media brands? No.
That, in my view, is because the adoption of legacy brands takes a different route. It’s more an induced or influenced habit than impulsive. To the evolved souls, it’s a matter of acquired taste — no less delicate than picking the right blend of Scotch whiskey or Darjeeling tea. It takes time but once that habit is formed, it sticks for a long time.
Also, cloning on the basis of “if a reader likes similar news brands ‘A’ and ‘B’, she would like ‘C’ as well” may not provide a solid, enduring interest base. Would you agree we don’t read The Washington Post or USA Today for the same reason we read The Times?
What the growing band of self-proclaimed data scientists (thanks to LinkedIn) need to realise is that, besides all obvious, popular, and objectively attributable factors — such as political leaning, length of news breaks, average article length, ratio of serious and entertainment coverages — used in their algorithms, there is this highly nuanced matter of style. A style of presentation, analysis, wit, disambiguation, subtlety, and, above all, the power of discourse.
Can data science clone audiences based on the above? Someone please educate me.
If I were an editor, I’d settle for no less. Each user who downloads my app is a promoter of my brand of journalism. Wouldn’t I use some discretion to select them? I would not target the flirts; they don’t make sense for a subscription revenue-based future of news brands. Unfortunately, current clone/match algorithms fetch more flirts than brand lovers.
Let’s try to figure out an approach to determine who to target.
I’ll use Google Analytics (GA) as the analytics platform, since many publishers use that. In GA, the user segmentation metrics based on “user value” leans more on e-commerce. Web sites with subscription plans would benefit more from that. But those who have spent money on Google Analytics 360 and connected a DoubleClick for Publishers (DFP) ad-server account to it can quickly build a great targetable pool of “first-party cookies” that can be targeted.
DFP’s own reporting structure doesn’t offer revenues/e-CPM data at page level. But it does when connected to GA. GA’s custom segment section even allows one to see revenue figures by users.
Is the data accurate? I’d suggest you take it with a grain of salt. Google’s improved technology technically associates the page/impression level publisher revenues data with users, but a part of the user metric is calculated on the fly. Still, it’s good enough when relative values are considered and considerably useful for targeting when audiences are built from it and exported to DFP.
So here’s how it works: First, create a segment from mobile Web site users combining “most engaging users” and “most revenue-creating users” — these definitions vary from publication to publication. With the segment created, build that into an audience. Now export the audience to both DFP and AdWords.
Why AdWords? DFP is your own supply-side platform. It will help show your app download ad for free to a user who meets your targeting criteria as long as she is on your products. To target her beyond your own products you need AdWords, which is a handy and simple demand-side platform, but this part is to be paid for.
This gives you a decent capability to target first-party Web site and mobile Web site users across the non-social media part of their digital lifetime and convert them into app users for better revenues and stickiness. You will see a better conversion rate. Hopefully the gap between app installation numbers and active users will also decrease.
This is one of the simplest methods. More accurate ones involve BigQuery and some custom coding, which I wanted to avoid in this article.
What about free GA users? There are some hacks, and it is a tad more complicated, but it may be worth exploring in the future.