A/Z Testing: Crowdsourcing the ad creation process

rather than having your ad agency come up with only a handful of ad options, why not crowdsource?

You want to tell the world about your new service – an on-demand nail clipping service. Ads seem like a straightforward way to educate consumers. But what should it include? Should you place a picture of a nail clipper? Maybe untrimmed toenails? Actually, maybe a short clip is the key to lowering your customer acquisition costs The classic way to resolve this issue is to work with an ad agency. Surely they can leverage their expertise and generate the ad that will get that much-coveted users’ attention. But how good are they, really, at understanding MBA students’ nerdy jokes? Or art enthusiasts’ refined taste?


Maybe it’s time to let the crowds take a stab at it?


How might this look like? First, a company wishing to generate ads would post a brief on the ad generation crowdsourcing firm’s platform (which we will call A/Z Testing, borrowing from the term “A/B testing”): what’s the product/service? What are the key messages they think should be communicated? Who do they think is the target customer? Any specific guidelines that need to be followed? Second, A/Z Testing would communicate the brief to relevant ad generators (e.g., based on price point, interest, past performance etc.). Third, each ad generator would generate an ad, and test it with consumers. The metric that needs to be tracked, most likely, is customer acquisition costs per given size of population. For example, identifying which ad could capture ten thousand customers the cheapest. One ad generator might test a simple ad – just an image with a clear tagline. Another might think that the product is too innovative to be communicated in a still image, and thus creates a short clip. Both ad generators run Facebook campaigns, and track conversion – ideally of actual engagement (e.g., purchases or sign-ups), but at the very least cost-per-click. The video will cost more to push to consumers, but might get more, relevant clicks per expression. Importantly, each ad generator would need to specify the exact advertising mechanism (e.g., “target college students, in the Northeast, via Facebook”). Ultimately, the winners would emerge. Fourth, the company would pay the winners, and run a larger-scale campaign at (ideally) a low customer acquisition cost.


The process outlined above is just one of many options. One can imagine many more. For example, the company could simply pay a fixed figure per purchase/sign-up to the users, and leave it to the ad generators to bring the traffic. This would be similar to the (already existing) affiliate marketing platforms, and would leave the company with less control.


There are, of course, challenges with this approach. A critical question to answer would be: when would this make sense? For example, with so many people working on marketing messages, will the required compensation be more than offset by the savings on acquisition costs? It is not uncommon to two ad concepts generating conversion rates different by a good multiple, so perhaps testing many ideas would lead to orders of magnitude? Another challenge is whether companies would be comfortable giving up control of their branding and marketing messages – if an ad generator uses messages not aligned with the company’s existing brand positioning, would the company still use the ad? In addition, there are technical challenges to overcome. One of them is tracking performance. Each company would need to track who referred the newly-acquired user (e.g., the short clip or the simple ad), and then automatically feed that back to A/Z Testing.


Assuming the above challenges are dealt with, the value created to the companies using the service is is two-fold –lowering acquisition costs is an obvious benefit, but perhaps more importantly, this would also let them better understand their customers. The value to ad generators is obviously the compensation, but also the ability to learn and to build a reputation, possibly allowing them to make the transition into the more traditional ad agencies should they choose to. A/Z Testing could capture value simply by charging a fee for using the service. Maintaining a barrier to competition wouldn’t be easy, with both companies’ and ad generators’ ability to multi-home on multiple platforms, should those emerge. There might be some advantage, however, to collecting data on ad generators, and better predicting which ad generator should be assigned to which project.


The idea of crowdsourcing marketing might make sense in a world where marketing expertise just isn’t as valuable as having a diverse set of ideas to test, with a low cost of experimentation. Time will tell whether such a model will emerge.


Fiverr – get anything done with $5


Nextdrop: sourcing from the crowd when the crowd doesn’t have internet access

Student comments on A/Z Testing: Crowdsourcing the ad creation process

  1. This is such an engineer approach to marketing! A/B testing is a fantastic feature to test ideas, on the other hand – exceptional ideas come from understanding your customers, your competition, the context you operate in, and a brainy gut that tells your what direction to explore. I believe that the crowd would struggle coming up with quality ideas because the winning idea is the one that is marginally better that the second best and, given the relatively low chances of winning since many will participate to the same auction. I would argue that ‘marginally better than second best’ is often ‘not good enough’.

    On the other hand, if you are curious about optimising digital ad campaigns, check out yieldex (http://www.appnexus.com/en/publishers/yieldex-analytics – oops, turns out Appnexus bought them when I wasn’t looking). They start from a creative idea and optimise the heck out of it!

  2. This is a very nice model. A few thoughts on that:
    I assume that the same way you would like to have customer data coming from the ads, in order to make the ads better you also need prior data. For example, how well your previous ad campaign did. A type of data I’m not sure the company would like to share with the ads generators.
    I totally agree that there is a big risk allowing someone that could be working for your competitor having access to your new product ad. There’s a risk letting him/her access to make a bad ad that will hurt your business.
    I would also think about the IP rights issue. How to certify the company doesn’t uses an ad generator idea, slightly change it, and avoid paying the ad generator.

  3. Thank you for your post Asaf. I agree that crowdsourcing advertising research offers interesting opportunities when designing very targeted ad campaigns on the Internet. However, companies can simplify and deepen the effort for broad-based campaigns – they can crowdsource the ad creation process altogether.

    Doritos was a pioneer in this effort. In fall 2006, the company launched the “Crash the Super Bowl” contest, whereby it solicited full-length television ads on its website and asked visitors to vote on the best ones. It committed to airing the ad with the highest vote-total during the Super Bowl. The result was an absolute coup for the company – Doritos became the first company to air a consumer-generated Super Bowl ad (it actually aired two of them during the 2007 Super Bowl – http://www.prnewswire.com/news-releases/doritos-surprises-super-bowl-television-audience-airing-two-consumer-created-commercials-live-the-flavor-and-check-out-girl-54272512.html) and received very positive feedback. In fact, one of the two commercials it aired was rated among the top 5 Super Bowl commercials that year. This was all achieved at minimal cost – Doritos simply offered the top 5 entries a cash prize of $10,000 and a trip for 2 to the Super Bowl. While the size of the grand prize has dramatically increased since 2006 (it’s now $1 million – http://ktla.com/2015/02/01/usc-graduate-wins-doritos-super-bowl-commercial-contest-1-m/), the company has been able to save millions of dollars in ad agency expenses by using crowdsourcing.

    The extensibility of Doritos’s strategy to other advertising venues is, of course, questionable. The Super Bowl occupies a unique place in American pop culture and the prestige of potentially having your ad aired during the game is a reward in itself. It would be far tougher to generate as high-quality productions from the crowd for ads that will be aired during a typical weekday television program. However, the viability of a crowdsourced advertisement should establish a price ceiling that companies are willing to pay ad agencies. After all, in a worst-case scenario, a company of sufficient scale could run a similar contest just among its employees and likely generate content that is nearly as creative as that of an ad agency. In other words, as companies tap “the crowd” to generate creative advertising content, the “creativity premium” should decrease.

Leave a comment