Enter a search term such as “mobile analytics” or browse our content using the filters above.
Check your spelling or try broadening your search.
Sorry about this, there is a problem with our search at the moment.
Please try again later.
Those of an analytical/numeric persuasion in marketing can often be heard wittering on about 'RFM scores' and 'quintiles', and much of the time these ideas feel like something completely irrelevant to the “real world”…
…but a couple of weeks back I had a real OMG! moment, when I had a stark lesson in just why these ideas matter, and how powerful they can be.
A real, live, OMG moment
A treat for the loyal customers
We wanted to target the 'loyal' segment of one of our customers, people who came to the site more than most, to whom we could make an offer recognizing their loyalty, and that meant we needed to write a 'trigger'.
So, we were looking at some data from the customer’s site (a publisher), and looking at how often people returned to the site (visit frequency). We discovered that their visitors had an average visit frequency of 1.73.
Interestingly, we had another customer (a retailer) who’s visitors seemed to visit on average 1.87 times (almost exactly the same number) so we thought some comparisons would be interesting too.
So we loaded the data from the two sites into a spreadsheet, and drew graphs to show the makeup of these two customers’ audiences. These showed how two sites with very similar average visit frequency (number of visits per visitor) can have very distinct audience make-ups, and reminded us why averages are dangerous.
The graphs showed the number of people who have racked up a given number of visits but made the vertical axis non-linear so we could better see what is happening.
Publisher audience profile:
Retailer audience profile:
Some interesting data emerged. If you look at how many visitors have made 20 visits, then we can see that there were about 300 in the Retailer’s audience and about 2,000 for the Publisher.
Actually, more strikingly, there are 25,000 visitors to the publisher who have visited more than 19 times, but only 2,200 visitors in this category for the retailer, and when we look for people who have visited more than 50 times, the retailer has 80 visitors in this category, whereas the publisher has 6,000 (75 times as many people).
And when you consider that the publisher’s audience is only eight times bigger than that of the retailer you can see the effect of people returning to see new content each day (news). The power of content marketing?
But how would you target these people?
The original point of this analysis was to consider how we could target their most loyal followers. So we thought that targeting the 10% most frequent visitors would be a good plan.
The 10% most frequent?
Based on the above data we discovered that if we were going to define a trigger to do this for these two sites, we would need to write a trigger saying 'target people with four or more visits' for the retailer, and 'people with three or more visits' for the publisher.
This surprised us as it meant we would only be excluding the one-time and two-time visitors on the publisher from our target segment, so this would hardly be an exclusive offer!
The top 10% of visit frequencies?
So perhaps we would do better if instead we decided to target the people who had the top 10% of visit counts. We discovered that this meant on Retailer’s site we would target people with 72 or more visits (10 people in total) and on the publisher's site we would be targeting people with more than 189 visits (four people in total).
…this was taking “exclusivity” a little too far!
We needed another approach!
These were not campaigns that were going to work very well (a target audience of four people is a bit too precise for most situations!) but I bet these would be exactly the strategies most of you would have intuitively chosen too, right?
… and of course, worse than that, if we took either of these approaches, then the numbers we would be writing into the triggers would need to continually change over time, as more data was accumulated and as the client's audience matured.
Enter the quintile
This is why the world invented quintiles…
If you read up on quintiles (Wikipedia is good) you will see that the quntiles break an audience down into five segments based on their Recency, Frequency and Monetary (RFM) scores (definitions here).
The Fifth Quintile (we know as 'Frequent' in the case of frequency scores) would contain the most frequent visitors.
So instead of targeting people who have 'more than xxx visits' we can target people 'in the firfth frequency quintile' or 'the Frequent Visitors'… when we did this the numbers got much better…
We discovered that our publisher has 26,000 people in this segment (0.6% of audience) and the retailer has 3,000. Now we were cooking, we could run a campaign which would target a significant set of our visitors without spamming almost everyone, or targeting an vanishingly small population…
Given that most current-generation tools provide this kind of analysis and triggering options we can target our campaigns without resorting to SQL, Excel or the IT team… Yee Ha, targeted marketing lives.