There's a lot of standard practice when it comes to avoiding fraud in digital marketing. Some of it is crucial, but some of it is woefully out of date. Rhiannon Collings from Machine Advertising cuts through the nonsense and provides some essential guidance for marketers looking to rule out app install fraud.
Fraud is a complicated topic, and some of the industry accepted wisdom isn’t actually in the best interests of advertisers. These can waste your time and energy, without putting much of a dent in your real fraud problem. So, whether you’re currently working with Machine or not, here are a few tips for how you can better beat app install fraud.
Consider an alternative to rebrokering bans
Rebrokering is a common scapegoat, and it’s certainly a contributing factor to fraud, but the truth of the matter is that it’s practically inevitable. Our industry is built on rebrokering, and that’s hard to change.
As an advertiser, rebrokering does mean a potential loss of control for your campaign – but ultimately, it is very hard to prove and something almost every single supplier will do. Even for those networks who have their own inventory, if a month’s demand exceeds supply, then the obvious solution is outsourcing.
Rebrokering rules are certainly worth enforcing, but a complete ban might not be the right approach if you want an honest and open dialogue with your suppliers. One solution could be to provide a blacklist of networks with whom your campaigns absolutely cannot run. This method does still entail a level of trust – even if you take a zero tolerance approach, it can be hard to catch suppliers if they do pass campaigns to blacklisted networks – but it’s a starting point.
Don’t limit yourself with publisher names
Another commonly-touted solution that doesn’t actually have much impact in terms of reducing app install fraud is publisher name transparency.
This is because publisher names can be easily fabricated or simply not verified. We frequently see one sub-publisher name or ID that cannot possibly represent the single application that it claims to, just from the sheer level of clicks it’s delivering in a given minute. The likelihood is that most are actually an umbrella for multiple publishers or even a whole rebrokered source.
So it’s not clear how much visibility this approach actually offers. Given that demanding publisher names can limit your options – many CPI suppliers currently can’t or won’t offer this – isn’t it better to utilise more of the market, rather than relying on transparency methods which can be falsified?
Go beyond compliance rules
A popular approach taken by advertisers to catch fraudulent players is enforcing broad, publisher-based compliance rules across their campaigns. For example, if a publisher delivers a suspiciously low – or high – click-to-install rate, some or all of the installs they deliver will be disputed and the publisher blacklisted.
This may seem like a good idea on the surface, but in practice there’s no consistency in the thresholds used by different advertisers – and that’s just the first issue with this approach.
As mentioned earlier, publisher IDs will commonly contain a range of different publishers grouped together within them, some genuine and some potentially fraudulent. When this happens, publishers can include two or more methods of fraud that have opposing signals, like unauthorised incent (providing install volume) and click stuffing (providing what appears to be strong performance). This results in a good blend of volume and apparent conversion performance, meaning the publisher doesn’t break any of the compliance rules, and is actually likely to receive up-weighted budgets.
On the other hand, grouping installs together in this way can also mean that a single breach of compliance rules can lead to all installs being disputed, including those which are genuine and valuable.
In our opinion, the best way to detect and prevent fraudulent app installs isn’t via these broad probabilistic rules, but to utilise a deterministic approach. This means analysing every single click and install individually, picking out the fraudulent installs and leaving the genuine inventory that should be paid for and optimised against.
This isn’t just a better way of catching fraud, and thus saving marketing budgets, it also improves efficiency. By removing the hassle of broad and retrospective analysis, deterministic methods can save a matter of days each month, leaving you to get on with the important stuff.
Have the confidence to call networks’ bluff
Something we often hear from advertisers is that, when fraud is detected, they experience pushback from their suppliers. Ad networks offer various arguments as to why their inventory couldn’t be fraudulent – arguments that, when examined closely, don’t hold water.
One common example is that networks claiming that anything bought on a CPA (Cost Per Acquisition) or CPE (Cost Per Engagement) basis can’t possibly be fraudulent, because the advertiser is paying for conversions from a real user, as opposed to just buying an install. This is not true.
It’s not just installs that can be faked, and fraudulent installs can deliver seemingly great in-app performance. For example, attribution fraud – which makes up 51 per cent of fraudulent installs detected by our platform – is one of the hardest types of fraud to catch, precisely because the installs it produces will perform. This is because the method works by stealing attribution from organic installs, meaning that while the user is real, they weren’t attracted by the inventory you’re paying for.
Other arguments we hear from networks are that suspicious click volumes are a technical error – 99 per cent of the time, this isn’t the case – or that the publisher in question is one of their top performers and drives great results across other campaigns.
These kinds of answers aren’t acceptable, and it’s vital that you have the confidence to tell networks so. If you’re working with Machine, we’re here to fully back you in these disputes. As well as supplying the granular data you need to prove installs are fraudulent, through our Helix platform, we offer the DNA service, which provides professional support for clients for every part of the process, including disputes.
Rhiannon Collings is client services director at Machine Advertising