How prevalent is click fraud? Chrysalis stopped 39% of one bitcoin ad campaigns budget from going to waste

It’s been an interesting few months in Dublin, as the team at Advanced Marcomm Services let our v0.1 of the Chrysalis Marketing Engine run for a few months. Born of the games industry, we’ve been testing a wide variety of things with the help of some of our agency clients. There is nothing better than having the opportunity to test a theory in the wild, and

So my first Chrysalis Case study has been a very satisfying experience. In my time in games middleware, publishing and development, I got hardened very young to the realities of the promises of sales people, and the results their networks delivered when it came to digital advertising. Quality, Cost and Volume are what’s important to most people, but may have a different order.

Repeated visits to Cologne’s Gamescom showed me a clear pattern when it came to those working the exhibition circuit, namely friendly(ish) but aggressive salespeople looking for large, up-front payments to run campaigns on their networks, which often would give minimal information in their statistics on what they were delivering. Just impressions, clicks, action, and a big invoice with net 14 terms, or if you were a new client running a “test campaign”, you’d have to deposit this upfront.

It was generally a one-way thing. You lodge money, clicks or impressions chip away at it, you get traffic, and maybe some actions. But then you’re confronted with the alert asking you to top up funds, to keep the campaign going. However, a very much under-utilised clause that is standard with networks that I work with is that they have a chargeback and cancellations period. While some reading this will have used it, I think a lot will have never invoked these clauses. And managing a marketing budget means you’re supposed to call the shots, right?

Most advertising networks work along the same tried and tested principles. They have an ad server. They have to recruit publishers, people running single or multiple websites or apps, generally in different themes, and get them to put their ad code into their site. A person (or an increasing proportion of non-human traffic with adblock) visits the page, and the banner is served.

Ad networks love this, they serve an ad, you as a marketing manager pay per 1000 impressions, and they’ll take a cut of x%, passing the rest onto the publisher. Alternatively you’re paying for a click, which is the networks second preference, or an action for something, which they’ll take, but their ad server will put at the bottom of the pile, unless it’s clickthrough rate, performance and price is competitive with the CPC and CPM campaigns running.

So unless you’re serving your own ads, you can’t really challenge this, and for the purpose of this case study, I’ll not be talking about CPA actions, which is for another time and another story.

As a marketing manager, it’s high risk, low return, nearly impossible to get a directly measurable ROI out of CPM, and you’ll usually take a hit for “brand awareness”, or if you did what I did in games, you’d use it to leverage some added value stuff, content, or competitions as part of the deal.

But for clicks and actions, there’s a lot more recourse, if you have the information available to you, and you act within the chargeback and cancellations period.

So with one of our clients in the Bitcoin space, we’ve been executing a marketing strategy using our Chrysalis Marketing Engine to measure what’s happening during and after clicks we’re putting paid advertising, or equally paid effort into, and it’s thrown up some very interesting things.

For social media measurement, it’s allowed us to see the lifetime of tweets, compared to forum posts, and other activities such as passive links back to their site. But two weeks ago I got asked to start running a couple of ad campaigns. Adwords was a given, and we’ve been testing successfully there since May in the transport sector, but for this, it was like launching a game all over again.

There’s a niche, with an increasing number of websites (and copycat websites) in it, and a couple of aspiring bitcoin advertising networks that have sprung up to bridge the gap between companies and fans.

Unlike those I worked with in games, the campaigns are technically are a lot easier to get going. Client sends be bitcoin for a number of campaigns over the next couple of months, I divide this into small chunks, and split the budget across several markets and networks. And because of the way we’ve designed Chrysalis, I don’t have to worry about the time consuming elements of setting up links for each network. So you can get ads going within about half an hour, instead of days.

I tested with a number of these self-serve networks, each which had different approaches, and promises. One in particular looked pretty good, based in Central Europe, I spoke to them on the phone, did a bit of research, and committed some funds to their CPC campaign for banners and contextual ads. We started it running for a few days, and on my first review of the data, I saw a couple of familiar fraud patterns emerge from my time in games.

Now when confronted with this, they acted in the way the better networks had. They explained that they have a 7 day chargeback period built into their publisher terms & condition. So with my being able to give them an itemised report within the cancellation period on my side, it was easy to get these clicks credited back onto my funds.

Now, armed with Chrysalis, we could put a deposit on, let all the good traffic through, get credits back for the traffic, and effectively double the length of the campaign, for the same money.

In the few weeks it’s been running, we’ve seen at least three different distinct fraud patterns, some which we can see in real-time, others that emerge over time.

And in this case, the proportions were pretty in line with what I’d expected from my own experience spending hundreds of thousands of euro a month in video games, but may shock others.

Keep in mind this is just one single campaign, among a number we’re monitoring.

Number of publishers running ads (1 product x 4 markets) – 115
Impressions: ~500,000
Clicks measured in period – 851

Number of publishers detected with fraud patterns – 17
Number of clicks from these publishers – 331

% of detected fraud publishers of total ~15%
% of traffic from overall campaign ~39%

Funds recycled and credited back on account 100% ;p

So this has helped us validate one of the (many) functions of Chrysalis. And with my Agency hat on, our client is happy too that we’re looking after the money they are entrusting with us. We’re giving them better information on the traffic we’re spending their money on than the ad network can provide and in some ways have more information on this traffic than the network’s staff.

Which adds up to actionable information that can save money, stretch out budget, and let us build a working relationship with the network over time, because they’re doing the right thing.

So if you’re a marketing manager sitting there reading this, and have pretty much sweet F.A. information behind all of those clicks you’re seeing for in your supplier’s statistics panels, and are paying hand over fist for the privilege, why not tweet to @DublinJammers or contact jamie [at] advancedmarcomm [dot] com or skype jamiemccormick to have a chat.