DMA Report - Accuracy

[chris_sitaram]

I'm trying to understand how Nielsen defines certain markets like Philadelphia, Wash DC, Baltimore, and Richmond.  Numbers in the DMA report for Visits, Page Views, and Unique Visitors are significantly bloated compared to the total traffic we get. For example, Visits for the above stated 4 markets in the DMA report for June is 546,469 while total Visits to the website was 817,947.  We think the 546,469 (almost 33% of total) is way too high.

Does the fact that there are lot more ISPs in these markets, make our #s high?

Appreciate your insights.

Chris

[A_Wil]

Those numbers do seem strange.  DMA is populated on reverse IP lookups, so it's possible that it's an ISP thing, but even that doesn't seem to make sense.  I'm not sure why those DMA's would be so overpopulated in your case, but when I run into strange things like this, I usually run a Data Warehouse report.  Sometimes it leads to some surprising insights.

In this case, I would probably use these settings:

Reporting Date:  any single day

Breakdowns (in this order):  GeoSegmentation Countries, GeoSegmentation Regions/States, GeoSegmentation Cities, GeoSegmentation Demographic Areas

Metrics:  Visits, Page Views

If you run this, I'd be curious to know what you discover.

[chris_sitaram]

Thanks Aaron!  As suggested, I ran a DW report for a single day and notice no big changes from the #s obtained the first time.  Specifically, my single day Visits count is averaging at the same high count I had obtained the first time.  I also ran a monthly (for good measure), and notice the same high count obtained earlier. I'm attaching the DW reports (both, single day and the monthly) for your reference/verification as this is my first time running one of these and hoping I get to learn something about the DMA report.

I'm expected to produce a DMA report for these markets for Jan 2013 thru June 2014, monthly; so even if the single day were to have produced an accurate report, I probably wouldn't have used it.  Nevertheless, it was a good exercise to go thru and I appreciate your suggestion.

Thanks again for your time!

Chris

[A_Wil · Accepted solution]

Yeah, that's really interesting.  It looks like 2/3 of your traffic is legitimately mapping back to just a few DMAs.

• philadelphia (504)
• washington (511)
• baltimore (512)
• richmond-petersburg (556)

As far as reverse IP lookup accuracy, it's a mixed bag.  For normal web traffic, you're getting the location of the ISP, which while not always an accurate representation of the user's city, is still usually local to the user.  However, with mobile traffic, you usually get a backhaul address that can be much farther away from the user's actual location - sometimes even in the next state.

It might or might not be worth more digging, but as a follow-up step you might break this out by mobile/non-mobile traffic and see if that changes anything.

Either way, the ballpark is probably correct.  It legitimately appears that your traffic is highly concentrated around DC (give or take 100 miles).  Because that is not what you intuitively expected, this might be a valuable insight.

Edit:  Just a thought, but it might also be valuable to compare this data against other location data you might have (zip codes, billing addresses, etc.)