Last week, we talked about a growing issue within the auto insurance lead-generation market. The crisis as we describe is the influx of fraudulent leads into the system, but not just a small number. These fraudulent leads are like a denial of service attack on those who sell leads to agents directly.
A group of initially well-intentioned middle men began sending along already passed over leads, tweaking them slightly and frequently trying to trick the filters of those aggregators who have agent bases. Like many attacks, this happens in real time, and the systems of the aggregators are for the most part dumb. They will look at anything sent their way, reject those that don’t match the criteria, accept those that look right.
Complex systems but dumb as to their ability to pick up on certainly unnatural changes, e.g., a huge spike in leads being reviewed, the similarity between leads reviewed, rejected, and accepted. These aren’t neural net machines, and so there will be a natural limitation in their ability to make predictive decisions.
We left off with a simple question. What will happen now that more and more companies are realizing so many of the leads they bought were fraud? The simple answer would be to find a way to turn off those sources, to do a better job of uncovering which leads coming in are fraud and, as a result, cleaning up the system.
By all intents and purposes, that is exactly what should happen. But, that isn’t what is really happening. The question is why? To answer, we must first look at exactly what the rise in fraudulent leads has done and how it began.
As one of these aggregators with an agent base shared, it took a while to realize that the problem was fraud. At first, the only empirical evidence they had that something didn’t look right came from the return rate. Agents return leads for a variety of reasons, and companies will more often than not push back on agents for some of those reasons. In this company’s case, the return rate started to creep up, and in the span of a few months time was double what it was before.
More disconcerting, the number of agents no longer buying leads started to creep up. Their agent base started to shrink rather than grow. When they asked the agents returning leads for the reason, they noticed an increasing trend.
The agents said the leads did not match their buying criteria. This actually happens to some degree quite regularly. Agents will say I wanted drivers with this filter, and this lead didn’t have it. What made this situation different was the percentage of returned leads with that reason.
Usually, when an agent returns a lead because of incorrect filters, the biggest source for incorrect filters is in lead mapping. For those who have worked with any lead buyer directly in a host/post scenario, mapping will sound familiar. The words we see on the form as a user are not how the buyer necessarily calls them. We see “First Name,” but when the seller must post that data, they will have to translate that into “FNAME” for example. If the field has values, such as Credit Rating, Good might be 3 not “Good.”
Do that with enough fields, and there are bound to be mistakes. Each person buying leads from the exchange has different ways of handling these errors, many end up sending them off to buyers they shouldn’t. They just don’t know that they do.
This, though, was not the source of the returns. The only conclusion after looking at the data, is that the leads themselves were changed. That made leads which were originally rejected all of a sudden worth purchasing.
The right thing to do is not buy fraudulent leads or to sell leads to companies who are known to modify data. Similarly, in an efficient marketplace, the economics of bad leads would force a change. Of course, the right thing is not always what is done, nor is the most efficient thing. The numbers tell why.
Fraud leads add more “revenue” into the companies. The ones with agent bases may have seen a rise in returns, for example 10,000 instead of 5,000 leads, but they saw an increase of leads sold by more than the 5,000 they returned.
If an agent base company wanted to, they could buy a lead they know to be bad. They would pay a much lower rate, because no one else really wants it. They might pay $2.50, whereas before they paid $8 and made $10. Now, they have a 75% margin not a 20% margin. In the past, they saw more than 20% returns, so they lost money.
Now, as long as less than 75% return they can make money off these bad leads. They are basically taking advantage of the fact that not all buyers will get in touch to know the leads aren’t good.
If enough people knowingly buy fraud leads and price them to make money, the auto insurance lead industry will enter into its own race to the bottom, continually adjusting price to stay ahead of returns and diminishing agent base. The hardest part now comes from forcing oneself to accept lower, real revenue numbers.
If companies don’t, you’ll still have some good short-term gain, but we’ll wind up with a lead bubble that when it pops takes an industry down. And, unlike education or mortgage, it will have been done internally and not externally.