【SpiderAF Case Study】Doing Fraud Investigations In ¼ The Time With Data Visualization!



i-mobile Affiliate

I-mobile is a pay per performance ad service provider. From June 2018, i-mobile started using SpiderAF for fraud prevention.
Although fraud can not be found out with just log data alone,
it was possible to detect these clever fraud upon introduction; greatly reducing the amount of man-hours and time spent on fraud investigation. Today we’ll be talking with i-mobile’s Miss. Nishimura and Mr. Matsumoto.


Purpose

・Protect transparency in ad networks and ensure a sound ad delivery

・Reduce the amount of man-hours and time spent on fraud investigation

Problem

・When ad fraud occurs, it’s difficult to stop and continue projects urgently as well as accept new projects

・Since it can not be detected and confirmed on the admin screen, sound ad delivery has become tough

・An enormous amount of time for fraud investigation is spent for fraud that can not be detected from log data alone

・We could not answer calls from both the demand and supply side about more scrutiny and adjustment 

Result

・Man-hours and time spent on fraud investigations were greatly reduced

・Was able to detect fraud previously unnoticed before

・Boosted detection of illegal mediums and able to investigate mediums in a more timely manner 

・Became an advantage when making proposals to clients



Previously could not detect ad fraud with log data alone



– Please tell us about when you felt the need for fraud detection before introducing SpiderAF.


Miss. Nishimura

We started feeling that this was necessary about a year ago when we were having some ad fraud running rampant in July to September of last year and had to subtract millions of units because of it. Even when claims were outside of our scope,  we didn’t have the data to prove it wasn’t so we couldn’t withdraw publications – so then it peaked.
In addition, we couldn’t clearly answer to our clients when talking about fraud prevention and ultimately lead to the lost of one of our deals to a competitor.


Mr. Matsumoto

We couldn’t prevent anything before it happened and we were getting feedback from our clients – just identifying that something was ad fraud the first time took an enormous amount of time every day. There are a lot of clever methods for ad fraud out there now – there were even times when we didn’t know what kind of fraud we were looking at.


Amount of time for fraud investigations was reduced to ¼ the time through visualization


Mrs. Nishimura

After we introduced SpiderAF, we started to able to see a large pattern of fraud just by looking at our admin screens where the data was graphed and scored out. So up until now, I’ve dropped the data and assembled a function in Excel… and then what would have usually taken 1 hour got shortened to 15 minutes.

After we started using SpiderAF, what would have been at most a several million yen deductible every month became almost non-existent in some months. Not only did it reduce the number of deductible cases and costs – but we got many contracts in a single month and were getting more orders in while also decreasing the number of loss transactions.


Catching up on system development and ad fraud for improving business


Mrs. Nishimura

Since one sales staff is attached, we can ask about any unclear points right away and get a response. When we say “I want this kind of function!” since there is going to be an answer coming that’s one merit that’s not on the admin screen. Recently we’ve had a function implemented that would import the memo function in the report.


– We can not respond to everything, but we will respond to your requests as much as possible

Mr. Matsumoto

After introducing SpiderAF, in the first 2-3 months Mr. Miyamoto (SpiderAF Product Manager) would occasionally come to our offices and explain to us – in a logical way – what kind of fraud has occured. It’s because of that that our own knowledge on ad fraud has increased and made a huge difference.


– What are some of the most useful and frequently used features on SpiderAF?

Miss. Nishimura

We look at the waveform a lot.

Mr. Matsumoto

Because fraud forms altogether come out as impacts, we look at the types of frauds very often. Also having the batch telling us if the device is from overseas or old is extremely helpful.

Fig.:Screen capture of wavefroms made from periodic clicking 

90% satisfaction with SpiderAF!


– Overall, please tell us about your future business/service plans using SpiderAF and the reasons!


Mrs. Nishimura

If we include the expected value we’ll get from the shared blacklist, 90%! 
In the future, we would like to build a fraud detection system that detects issues before they occur by linking up API with SpiderAF’s shared blacklist so that it could detect whether it is fraud at the time of clicking and not jump onto an advertisement.



Mr. Matsumoto

By accumulating the data of clever ad frauds, we’ll be able to operate more suitable ads and empower the soundness of the platform on the publishers side. Not only just by the number of times installed but I would like to also see ROAS grow beyond into a product that will be ordered continuously