a survey of 16,819 banner-ads, from 42,300 websites and apps, over 365 days

There's a multitude of ad-spy tools out there: AdGooroo, SpyFu, AdBeat, WhatRunsWhere, MixRank, KeywordSpy, KeywordCompetitor, iSpionage, SEMRush, Follow.net, etc. I tried all of them. And they all suffer from the same downfall from my perspective: there's no differentiation between direct-response & branding ads.

So I set up my own system, (and you can too using 80legs.com or scrapinghub.com)

The collected data was scraped from 42,300 websites and apps, over 365 days. I juggled the data as part of my Exploratory Data Science course with Udacity.

I sliced & diced the data with R, RStudio and StatAce.

The Data Set

Here are some examples of the 16,819 banners that were scraped, from 42,300 in 15 countries

Example: 160 by 600 pixels
160x600px banner ad
Example: 486 by 60 pixels
Example: 728 by 90 pixels
Example: 300 by 250 pixels

Direct-Response, or Not Direct-Response?

Coming from a purely direct-response, direct-mail background (does anyone remember Fax-blasts?), over the years I've developed my own internal checklist to weed out the brands. Big Brands have "burn budgets" and when they turn on 'the firehose', it is very easy to get misled into thinking a particular type of ad should be replicated. Or that there is something 'new' that now works in advertising, when in actual fact it doesn't perform.

So what are the criteria by which I decide whether or not an ad is "direct-response"?

is it a Big Brand?
This is quite easy to judge. Is it Coca-Cola, GM, Pfizer, ...? Chances are there is no monetization strategy, and they are going for the "branding effect".
no CTA, no attempt at monetization, or lead-capture?
This can be quite a lot more subjective, and relies on my experience/idiosyncracies as a marketer. Let me know if you disagree with any of my choices.
is it a Charity?
Charities can get $10,000/m free AdWords advertising anyway, and usually have very poorly crafted ads. Although there are some remarkably workable campaigns off-line, in the online world however, I didn't see any in the scraped results.


Banner sizes scraped (width/height):

This graph shows which banner ad size is more popular, per country.

A total of 16824 banners were scraped, in 15 countries.
The Websites scraped are relevant to each country, although scraping errors are bound to occur. See notes below.
Time Frame
365 days prior to 17 May 2014

468x60 is clearly the most prolific banner on the web


This Pie Chart shows the sampling sizes per country

For the 15 countries sites' scraped, each country received equal attention. Although some countries have less info readily available.
The Percentages show the relative size of the dataframe each country received
Country banner-ads scraped
Australia 1022
Germany 1226
France 1206
Italy 1159
Netherlands 950
Turkey 972
UK 1003
US 1229
Spain 1238
Israel 960
Mexico 1064
Brazil 1182
Russia 1142
Sweden 1278

Ads per Capita

Displays the quotient of population divided by amount of banner-ads found.

The outlier is the USA, which shows for one thing that each banner-ad found is targeting a much larger audience, which comes as no surprise. Although it could also indicate missed oportunities in smaller niches, since the volume of people qualified to be marketed to, is so much larger.
Missed Opportunity
The population-size is much larger, however the amount of banner-ads and variety is not proportionately larger, showing room for additional marketing in this channel.
comments powered by Disqus