This question is a great one that confuses many – even the Wikipedia page it references could be made a little clearer…. maybe in my spare time I’ll update it
I understand what it is saying but why doesn’t it apply to First Time Unique visitors which gives you the same total for the month in comparison to adding up the value for every individual day?
It will explain how Prior Unique Visitors is different depending on the date range.
Any further thoughts?”
This is a common question and basically it comes down to the measure of “uniqueness”.
Prior unique visitors seems to be a “return unique visitor” while the Ffrst time unique visitor is an absolute unique visitor.
Your total unique visitor count should be a combination of Prior Unique Visitors and First Time unique visitors.
If you look at the table below;
Day
1
2
3
4
5
6
7
8
9
10
Visitor A
1
1
1
1
1
Visitor B
1
1
1
Visitor C
1
1
1
Then you get this result,
FTUV
1
1
1
0
0
0
0
0
0
0
PUV
0
0
1
0
2
1
1
0
1
2
TUV
1
1
2
0
2
1
1
0
1
2
When you look over different time period and add these visits together you see a lot of variance that seems contradictory.
1 Day granularity
10 Day Granularity
3
3
8
0
11
3
FTUVs will never change with granularity because a visitor can only be new to your site once (unless they delete a cookie, and then they are a new visitor).
I recently got asked a good question about calculating unique visitors to a particular page without actually having the UV data.
“Can I determine the number of visitors to this page in the following way?
Total Unique Visitors Jan = 100
Total Page Views Jan = 1000
Avg Page View per visitor = 100
Notebook page Jan = 600 page views / 10
Therefore 60 unique visitors visited this page in Jan?
60% conversion.”
It isn’t uncommon to have to extrapolate some things from data that doesn’t quite answer what we want to know, and we can do this by using proxies from what we are measuring, or using relevant trends and applying it to the data we have.
Unfortunately, you can’t really make this assumption – unless you want to be a bit fast and loose with the truth.
The only real way to know this is to actually measure it, and without knowing the specifics of the tool, I’m not sure what it’s capabilities are. In Omniture, you can just turn on a unique visitor correlation to pages. In Google analytics, you can just drill down to the content page and it will tell you the unique views and other juicy info.
I would suggest that if you can’t get this level of detail, you should implement Google analytics as well as what you are currently using, or totally migrate.
As for logic in the question, you can’t say that “Therefore 60 unique visitors visited this page in Jan” because you just don’t know that for a fact. It would be largely dependent on your site structure. For example, if the notebook page was 4 pages deep, then to construct the average of 10 page views per person, you would make up those 10 pages of something like 5 at one page deep 3 at two pages deep, 1 at three pages deep and 1 at 4 pages deep (5 + 3 + 1 + 1 = 10). If that was the makeup of the average visit and the page resided at 4 pages deep, then 600 page views would correlate to 600 unique visitors.
Why are offer impressions a better metric than a catalogue page view?
Offer Impressions are a better metric than a catalogue page view for many reasons that can best be appreciated when looking at the short history of the web. Initially, the metric that was used as common currency for doing apples to apples comparisons of web real estate was the “hit” – the number of client requests to a web server. This was quickly outdated as pages became more sophisticated and “hits” were being registered when every asynchronous element refreshed (widgets etc), tracking servers were called and a myriad of other activities resulting in “hits” took place.
Then, to replace this came the Page View metric. Marketers decided to disregard how many “hits” there were since it was no longer one page = one “hit” and it was more valid to see how many page views there were. However, this metric soon became outdated as well when AJAX (asynchronous JavaScript and XML) became the standard for usability in the Web 2.0 world. Now, more than half of a page’s content could change without a new “Page View” ever being registered since the URL remained the same.
So, the impression was born to be the apples to apples comparison metric for marketers. Regardless of what all of the information on a page was doing, how many servers where “talking” to produce a complete page of information, the one thing that would remain constant and relevant to marketers across all mediums was – how many times did my offer appear in front on the eyeballs of a consumer.
Catalogue Page view’s suffer the same drawbacks as page views with regard to performing accurate analysis for a marketers spend – despite being the common currency in the analogue world where “offer impression” granularity is hard to come by since offers per page vary from page to page. However, ad space inside catalogues is still sold at the offer level, indicating that more granular metric is the more relevant one to marketers.
We can track catalogue page views, yet, if this is all that is looked at, then the offer impressions that appear in search, and not on a catalogue page view is never counted, despite this being the more relevant and targeted advertisement. Some may think you could just measure both and report this, but now you are dividing what would otherwise be a consistent metric, making it harder to understand and do accurate analysis on. What is the point of looking at a report that may say
100 Catalogue Page Views
13 Search Listings
813 Offer Impressions
Rather than
813 Offer Impressions
13 Search listings
Where catalogue pages are evident inside the more granular metric?
The result of this approach is that when comparing to other media and you are looking at
Print: Circulation
Print: Readership
Print/ catalogues: Distribution
Viewership: TV
Listenership: Radio
Visits: Web
You have one universal metric to apply to your ROI calculations.
NB: This will become much more important when we look forward to a more convergent Internet as described here by Kevin Kelly.
Part 2. Using a universal metric when calculating ROI? The offer Impression
Reducing offline media to a universal metric – the offer impression, is fairly intuitive, but has some intricacies in finding the correct estimates for certain variables.
Offer Impressions = (Circulation x (wastage factor)) x (probability of reaching ad page depth) x Offers per page
Offer Impressions = (Readership x (wastage factor)) x (probability of reaching ad page depth) x Offers per page
Offer Impressions = (Distribution x (wastage factor)) x (offer impressions per person)
Offer Impressions = (Viewership x (wastage factor)) x (offer impressions in TVC)
Offer Impressions = (Listenership x (wastage factor)) x (offers in radio commercial)
Offer Impressions = Offer Impressions
Now you have an apples to apples metric to begin calculating ROI (In fact, you could even work out offer impressions per unique visitor – however, this is a different yardstick compared to the other media that are largely absent of any tracking).
The next thing required is to establish the quality of an offer impression in each medium with respect to sales. This is the responsibility of the marketer, not the ad network. The job of the ad network is to generate the highest quality offer impressions in the greatest quantity possible for the lowest possible price.
The general formula then for establishing media effectiveness would look something like
Media Effectiveness = Sales per offer impression = Sales / Offer Impressions
I.e. Sales = Offer Impressions x Media Effectiveness
This can now be ascertained per media channel.
ROI = Sales/ Dollars Invested
Hence, your ROI can be predicted and measured by
ROI = (Media Effectiveness x Offer Impressions)/ Dollars Invested
Now, to get a relevant statistic for this you really need to run a control set of offers, probably simultaneously across different media that is unique to each state – eg Radio in Melbourne, Search in NSW, TV in Queensland, and then assess what the performance was in each medium so that you can ascertain what the Media Effectiveness was for each channel. If you have eCommerce enabled, you can get an idea of the effectiveness of online channels by seeing the quality of different traffic streams on online sales. However, this totally ignores web to store conversion which research has indicated counts for up to 95% of online traffic on the eCommerce enabled sites.
