State of Fraud: Lessons Learned From 2022 and Insights Into 2023

0:01
Hello everyone and welcome to our webinar today, state of fraud, lessons
0:06
learned from 2022 and insights into 2023,
0:11
brought to you by SIFT. I’m John Stewart, editor of Digital Transactions.
0:18
The economic outlook for the year is looking increasingly challenging as
0:22
consumers deal with increased inflation and recession expectations.
0:27
Traditionally, this type of environment is a breeding ground for fraud.
0:31
Merchants that aren’t prepared for it can find themselves dealing with a
0:35
significant spike in fraud and chargebacks. In this webinar, we’ll go
0:40
in depth to explain the challenges merchants should expect to face in
0:44
tougher economic times and provide clear actionable solutions for adjustments
0:49
they can make to minimize the impact on their business.
0:53
Let me introduce our speaker today. Brittany Allen is trust and safety
0:59
architect at Syft. She has more than a decade of experience combating uh
1:03
e-commerce marketplace fraud at companies such as Etsy, Airbnb,
1:09
First Dibs, and LetGo. Her expertise in fraud mitigation, policy leadership, and
1:15
dispute management has led her to speak at numerous industry conferences and
1:20
join Sift as trust and safety architect. So Britney, I’ll turn the mic over to
1:27
you and uh we’ll get started.
1:31
>> Thank you so much, John. Uh so hi everybody, welcome and thank you for
1:36
taking the time out of your day to join us for this webinar. Uh before we dive
1:41
into these lessons learned and talk through the recession and other
1:45
challenges that are in front of merchants, I want to just briefly give
1:48
an example or talk a bit about what I do at Syft and what we do there. So is a
1:54
fraudrevention platform and a leader in digital trust and safety. And we help
1:58
companies including those that you see here and many more with payment fraud
2:02
challenges uh to stop account takeover with reviewing user generated content
2:07
and also with chargeback dispute management. So as John explained the
2:11
company that I’ve worked for in the past, you can probably see how my
2:14
experience as a merchant fits right in with that. And I am very happy to be
2:19
able to do fraud research and to be able to speak at events like this to share my
2:25
knowledge and you know the knowledge of others as far as I can. We’re going to
2:31
touch on three different topics today starting with fraud in a recession. Then
2:36
uh focusing in a little more narrow on firstparty misuse which you might know
2:40
better as friendly fraud and how that has changed over the past few years. And
2:45
then we’ll end with some takeaways on convincing management for you know
2:49
changes you might need, improvements you might need with the data that you have
2:53
available. So let’s start with fraud in a
2:57
recession. Now looking at this graph here which is from the NBER which here
3:04
in the US is the National Bureau of Economic Research and actually the
3:08
authority that is officially in charge of determining whether or not we’re in a
3:12
recession. You see that we’ve got uh quite a few patterns here of both spikes
3:17
in unemployment, that’s the blue line, lining up with recessions, and those are
3:21
the vertical gray bars. Now, even if you haven’t personally worked at a company
3:27
that has dealt with the recession, maybe you’re newer to the job market, your
3:30
company has very likely weathered these before and maybe even multiple. At the
3:35
very least, most tech companies remember the Great Recession and its impact. You
3:40
might have been laid off. You might have watched colleagues being laid off. We’ll
3:44
touch a bit on the impact of that in a few more slides. But you also then
3:50
recall how unemployment or high unemployment rates lead generally to
3:55
less discretionary spending or what we love in e-commerce when consumers are
4:00
happy to buy gifts and uh to spend a little bit more especially on our
4:05
websites. So then if we’re looking at the very right edge of this graph and
4:10
that blue line that’s trending downward showing a decrease in unemployment,
4:16
we should be feeling pretty good about there not being an impending recession.
4:21
And yet for the past year, year and a half, for what feels like an
4:26
interminable amount of time, that’s the only buzz we’ve been hearing. So are we
4:30
in a recession? Are we approaching a recession? Let’s look at what some of
4:35
those key indicators are and break them down so that you can have as a takeaway
4:40
from today a better understanding of how to gauge whether or not we’re actually
4:45
approaching a global recession. So I chose a scale here to represent uh sort
4:51
of both sides of the good and the bad sides or at least the indicators that
4:55
lead to economic expansion and then others that would indicate recession.
5:00
And I think we should start on the left here with the bad side. So with
5:05
continued tech layoffs, that should, you know, technically lead to an increase in
5:10
unemployment. And I think the news from yesterday about Meta laying off another
5:14
10,000 employees uh goes right hand with how that’s a current concern. Second on
5:19
the list, I’ve got cost of basic goods is increasing. Uh right now we’re at a
5:24
point where consumer goods cost roughly 6% more than they did last year. Those
5:30
were the numbers from February. And then lastly, the Fed uh raising interest
5:36
rates. That’s something that’s been constantly happening on a you know
5:39
quarterly basis here in the US. And just this morning, it made the news in the EU
5:44
when the European Central Bank raised the interest rates by half a percentage
5:50
point to combat inflation on their end. in response to the credit squeeze news.
5:54
So all of that being sort of negative indicators that say we’re going towards
5:58
a recession, the reason it’s so difficult uh especially for those of us
6:02
who aren’t economists to discern whether that’s the direction we’re heading or
6:06
not is because we’ve got some equally good signals balancing out each of these
6:12
bad signals. To start with the continued tech layoffs, we’re actually here in the
6:17
US looking at a falling unemployment rate. It’s at 3.6% 6% right now. That is
6:22
the lowest in nearly 50 years, longer than practically any of our careers on
6:28
this call, unless you’re someone who’s very, very close to retirement uh with
6:31
us and still working. Then to counter the cost of basic goods increasing,
6:38
we’re actually looking at falling inflation. It’s been moderating since
6:42
last summer’s peak. And for that uh increase of consumer spending, January
6:47
saw Americans with surprisingly high spending amounts. So even though the
6:52
goods are more expensive, it seems like people are still buying. And then
6:56
lastly, on the opposite side of the Fed’s raising interest rates, the US GDP
7:01
was actually up 2.7% in Q4 of 2022. And that organization that I mentioned
7:07
earlier, the NBER, they determine a recession based on two
7:13
consecutive quarters of declining GDP. So that indicator right there will cause
7:18
them to not, you know, declare a recession. So this is really difficult
7:23
for us to predict. I’ve done my best to summarize it here based on my limited
7:28
experience. I want to tell you that in my past life, I was actually a history
7:31
teacher and I taught civics and economics for multiple years. So this is
7:35
kind of the best I can do from that perspective. Uh but what we can look at
7:40
to try to get away from focusing on predicting whether we will end up in a
7:44
recession or we’re going to make a soft landing is to just focus on what sort of
7:49
matters to us within the world of e-commerce and to look a little bit
7:52
closer at what the consumer is doing. So let’s look at these two charts here with
7:58
consumer behavior. Now, on the left, I’ve got a graph from Adobe on the Black
8:04
Friday numbers starting in 2014 and going up to last year in 2022. I do want
8:10
to call out here that even though Black Friday numbers hit a record high for
8:14
consumer spending, we’ve kind of got this flattening of the steady increase,
8:19
which we all know is due to many different causes, the pandemic, supply
8:22
chain issues, but it does still at least show that those numbers haven’t fallen.
8:28
So, could that uh record high along with that flattening be a result of people
8:32
getting stimulus checks, people spending more because they have to due to
8:36
inflation increasing prices, or maybe due to limits within the supply chain
8:41
that make people more willing to spend full price for an item instead of
8:44
waiting for it to go on sale because they know there aren’t as many of those
8:47
items available. I don’t have the exact takeaways from that, but we can compare
8:53
what’s happening right now or at least as of the last full month we’ve got data
8:57
on February 2023 by looking at some of the inflation breakdown in that chart on
9:03
the right. This takes some of the top consumer goods that have experienced
9:08
massive inflation or at least massive in the case of eggs. That is the dark blue
9:12
bars that are trending towards the right. And then looking at goods in the
9:16
light blue bars trending towards the left that have actually seen a decrease
9:21
in inflation or a decrease overall in their prices. Now you could say because
9:26
we’ve got televisions down there and some other electronics like major
9:30
appliances as being ones that you know weren’t as impacted by inflation that
9:35
might be things that people spend on on Black Friday. therefore continuing to
9:40
drive those Black Friday um increases in spend. Whereas those consumer goods that
9:45
make up most at the top of the ranking aren’t going to be ones that really
9:49
trend towards the Black Friday spending behavior, but of course are core goods
9:54
that individuals must continue to keep buying. Even up there with the motor
9:59
vehicle insurance, people need to have their car to get to work. All of those
10:03
being strong needs for consumers should be something that is an indicator of how
10:08
that behavior could continue to shift through the year. But now, how does
10:12
fraud fit in? This webinar was called or is called the state of fraud. And that
10:18
is represented best by looking at cyclical fraud trends. So if you break
10:25
fraud trends down into four categories, you could look at evergreen fraud
10:30
trends, which is the kind of fraud that doesn’t change very much and is with us,
10:35
you know, through perpetuity, through the decades. Uh a good example of that
10:39
would be, you know, card not present credit card fraud. You could then also
10:43
look at emerging fraud trends, those that are not yet fully defined and maybe
10:48
taking advantage of new types of alternate payments, new ways of
10:52
interacting online as the internet of things expands and you can make a
10:55
purchase from your refrigerator. Then looking at cyclical fraud trends and
10:59
just to mention the fourth category that last one I always have an edge case
11:02
category for the weirdest and wildest cases but for cyclical fraud trends that
11:07
will be the eb and flow of particular types of fraud based on how well the
11:12
economy is doing. So I’ve got an example here of what could happen or what does
11:17
commonly happen when the economy is expanding and that example is tax return
11:22
fraud. So when there’s a very low unemployment rate and companies have,
11:27
you know, very high headcount and maybe they are a little bit more overwhelmed
11:32
when it comes time to to tax season, we’ll see those attempts to get W2s to
11:37
get tax information from companies and file those tax returns on behalf of the
11:43
employees stealing their returns. And that is a pretty big spike when
11:48
unemployment is actually quite high and people are maybe thinking, “Oh, I can’t
11:54
wait until April to get my refund. I’m going to file my returns January 1st.
11:58
I’m not currently employed. I have all the information I need.” Then there’s
12:01
less of those tax returns for fraudsters to get their hands on. On the flip side,
12:06
when they’re in a recession, we see a big increase in job listing scams. Now,
12:14
this is because fraudsters are they’re fully defined by the opportunities that
12:19
they exploit and their success is going to spring from the behavior of the
12:24
victims. So, let’s look at what some of that behavior is just so we better
12:30
understand. Uh, and now I do want to call out even if you don’t work for a
12:35
platform that has job listings, this is still something that’s going to impact
12:39
you because of the end goal of fraudsters who run these job listing
12:44
scams. So, what they’re doing in a sense is they are advertising jobs that don’t
12:50
exist. Those are either with madeup companies or they could be pretending to
12:54
be your brand on various job listing platforms or social media platforms,
13:00
which is another potential impact with brand damage. But they’re taking the
13:04
information that the individuals provide in order to get a job, which is their
13:08
IDs, their banking info, a ton of PII, potentially even a payment via a credit
13:15
card for what they think is a background check. And then they’re using that
13:18
information to commit identity theft. It can lead to a, you know, increased uh
13:24
creation of synthetic identities, all kinds of resources that fraudsters would
13:29
then use to target your platform and potentially commit payment fraud. But
13:34
let’s loop back then to the behavior of UN, you know, the victims and and what
13:39
makes it so attractive during a recession to do these job listing scams,
13:44
which the FTC recently reported that they received 70,000 complaints about
13:49
job listing scams over the first three quarters of 2022. So, we know there is a
13:54
high volume of these scams out there. So first reason that this is successful
14:01
in the recession unfortunately goes back to desperation. People who are
14:05
unemployed, have recently been laid off, are more willing to ignore the red flags
14:11
that could otherwise indicate to them that this is not a real job or that this
14:14
is risky. Unfortunately, we also see some people who become willing mules.
14:20
They understand that what they’re participating in is potentially
14:23
fraudulent. Those might be ones where they don’t give up their PII, but they
14:27
go willingly make bank deposits or they willingly do uh some freight forwarding
14:32
fraud or other types of fraud for the criminal individuals who have quote
14:36
unquote hired them because they’re, you know, not exactly approaching it from a
14:40
moral standpoint. And then of course there are the new fraudsters who come
14:43
out at this time period, they themselves needing to earn money so that they can
14:48
take care of their basic needs and they’re then more willing potentially to
14:52
commit fraud. I make these three distinctions here because I’ve done some
14:56
work before talking to job listing platforms about how to differentiate
15:00
between these three categories uh and determine whether or not somebody who
15:04
has fallen victim for a scam is a true victim who you might want to invite back
15:08
as a good customer or is somebody who you should you know never allow on your
15:13
platform again. But it’s important to look at those three.
15:16
Then we’ve got resourced trust and safety teams with all those tech layoffs
15:21
we mentioned. Unfortunately, that does impact the fraud prevention and payment
15:25
side of an organization, leaving those organizations with less uh c headcount
15:30
with less resources in order to do review. And fraudsters know then that
15:34
there may be less scrutiny on orders. It’s sort of the same uh the same
15:39
approach that they would take to making a fraudulent order on Black Friday or
15:43
Cyber Monday and hoping to hide in that traffic.
15:47
Then you might be working for an organization where the rejections or
15:51
the, you know, dismissed orders are under more scrutiny because you are
15:56
supposed to be preserving or protecting as much revenue as possible, which can
15:59
then unfortunately allow more of these job scans to go through, especially
16:03
those in the gray area where a merchant might not be as sure as to whether or
16:07
not they’re actually risky. And then lastly, demographic swings. I love to
16:11
call this out because I think so many people in e-commerce were thrown off uh
16:16
at the beginning of the pandemic when all of a sudden you would see velocity
16:20
spikes in orders, but they were for items that weren’t really risky such as
16:24
toilet paper. But still, if a fraud prevention system was not, you know,
16:28
tuned properly or or set up in a right way, then it could maybe identify those
16:33
as potential fraud just due to velocity swings. You’ll see some of those
16:37
demographic swings the same way with job listing scams as people are all laid off
16:42
within one industry or a particular industry that does well during the
16:47
recession or even during this, you know, um, post-pandemic still work from home
16:52
time period. You’ll see swings that way. I talked to one job listing platform
16:56
recently that said about a 100% of their scans were all set as remote job
17:02
opportunities. And so that would be a demographic swing there. All right,
17:07
we’ve barreled through talking about the recession, thinking about how it impacts
17:12
consumers and fraud. Let’s turn to a specific example that we can work
17:18
through over the course of, you know, about two or three years worth of data,
17:21
and that is firstparty misuse. I do want to say we are going to be pausing during
17:28
the course of the webinar for questions, but if anything you’ve got uh springs to
17:32
mind as far as a question throughout this presentation, please do submit that
17:36
and we’ll have plenty of time at the end to address your questions.
17:41
All right, so I know this is 2022 and 2023 for the title of the webinar, but I
17:46
couldn’t resist going back to 2021 and just giving a snapshot of return and
17:52
refund fraud from that year. Uh so in 2021 we saw inflation begin to rise and
17:58
it was largely attributed due to the imbalance of supply and demand with the
18:04
supply chain issues that were happening. Everyone remembers their favorite giant
18:07
boat the ever better that got stuck in the Suez Canal. And it was later then
18:12
fueled by uh an imbalance in labor, a strong demand for labor and not a return
18:18
necessarily from the supply side of labor. And you’ve got the article there
18:23
from the Wall Street Journal in the middle highlighting return scams jumping
18:27
as fraudsters exploit the e-commerce boom. That was something that had
18:31
already been occurring. Return and refund fraud dates to years earlier, but
18:35
it certainly was something that had grown into what we call fraud as a
18:40
service or fraudsters who don’t just focus on selling stolen data or
18:45
committing fraud on their own, but guiding and assisting others with
18:49
committing frauds either by actually teaching them how to do it or just sort
18:54
of providing a service and a guarantee that will allow them to commit return
18:58
and refund fraud. The screenshots that you see either on the left or the right
19:02
of the Wall Street Journal article are from refund fraud forums from this time.
19:06
Um, the one on the left has my favorite name, the fabulous refunder. I do
19:11
believe that one has gone out of business or shut down subsequently, but
19:15
was still a good resource to show exactly how easy they made it for
19:20
individuals to place their orders and to then be guided through the process of
19:25
committing policy abuse and getting a refund back from the merchant. In
19:30
instances where they weren’t successful in getting that refund back, then we
19:34
would see a lot of chargebacks pop up, whether they were for non-dely, whether
19:38
they were for fraud, whatever the reason codes were. This was the state we were
19:41
looking in at looking at in 2021. So what did that lead us to do in 2022?
19:48
Well, the most common thing that I heard mentioned last year as a potential uh
19:55
attempt to work against refund fraud was to make it easy for your legitimate
20:01
consumers, the ones that actually do have a valid reason to return to
20:05
complete that return without having to resort to chargeback disputes or
20:10
otherwise, you know, create a headache for you, the merchant. So, here are two
20:13
examples from last year’s holiday season. We’ve got uh TJ Maxx and we’ve
20:18
got Saks Fifth Avenue, both of whom expanded their return window
20:24
significantly. Uh I think it says here roughly that
20:28
they allowed for returns from early October, in some cases through beginning
20:34
of February and in other cases through a date in January. But expanding that uh
20:40
window for potential return allowed them to try to keep the customers happy, not
20:48
also have a deluge of returns in their return facility all within a narrow
20:52
window, but spread that out so that maybe they could sort through those
20:55
packages, actually address them and examine the ones that were potentially
20:59
fraudulent returns uh more quickly. But these customer-friendly approaches were
21:04
very common. Now, in the same vein, the way that the refund fraudsters work is
21:11
that they take advantage of a merchants’s policy and they abuse that
21:15
policy. So, of course, the wider that window for returns possibly is, then the
21:20
more, you know, opportunity it gives that refund fraudster to commit their
21:24
their fraud. So, this kind of cuts both ways then. But I do think last year I
21:31
was also hearing a lot of discussion about a shift to cracking down on return
21:35
and refund fraud, whether that was via policy constraints or whether that was
21:40
via actual product changes uh that a merchant made, which we’ll get into some
21:44
examples in a couple of slides. But I did hear about that quite often last
21:49
year. So let’s move on to some numbers from 2022 then. So, we saw the policies,
21:54
we saw what happened, but what has actually been reported from 2022.
21:59
So, the first screenshot there on the left is a prediction. This is from
22:03
Insider Intelligence, and it’s dated December 2nd, 2022, so before the
22:08
holiday season officially concluded. And they projected that $279 billion would
22:15
be uh wrapped up in online returns in 2022. It’s a really significant number
22:22
and it is one that was almost completely accurate. So, drum roll please. There’s
22:28
the reveal from January 10th, 2023. The same source, so pulling from the same
22:33
numbers and hopefully having the same methodology towards their reporting
22:36
actually found that the returns totaled less than what they were worried about
22:41
at $23 billion. That is a decrease of 2.5% but it is still a significant
22:48
amount of items that were returned. There could again be a lot of theories
22:52
into why that was different um why it went slightly down even though we know
22:57
that it was holiday spending went up. Not really here to debate the numbers
23:01
there, but we at least did not see the numbers come in significantly higher
23:06
than what merchants had been bracing for based on those predictions. So, I also
23:12
want to call out the example screenshot that I’ve got here on the right from
23:15
another refund fraud forum. This one, I want you to look at the text towards the
23:20
very bottom because I want to say that it’s not just that refund fraudsters
23:25
have figured out how to abuse a merchant’s policy in order to falsely
23:29
report that an item was never delivered or was delivered not as described. But
23:34
they are providing tools as significant as the FTI ones that you see there. That
23:39
means a fake tracking ID. So they can generate that based on whichever shipper
23:44
it needs to match and also do that fake tracking ID with a weight match. So
23:50
they’ve realized and they’ve adapted to the changes that merchants have been
23:54
making in an attempt to slow down or stop return and refund fraud. And they
23:59
said, for example, merchants have picked up on the fact that we’ll take a return
24:03
label and just slap it on an empty paper envelope so that it gets shipped back to
24:09
the merchants’s warehouse or whatever the facility is that collects returned
24:13
items and shows successful delivery, which will force a refund or at the very
24:19
least, you know, give me the best chance of getting that refund via chargeback if
24:22
I find that I have to fall back on that. So, they’re now saying, “Look, we’ll
24:27
give you a weight match as well.” Whether that is a box of rocks, whether
24:32
that is a box of other unrelated items. I was at the merchant risk council
24:36
conference uh last week and I heard one merchant mention a box of salamis being
24:43
returned. This is an electronics merchant. So whatever they can put in
24:47
that equals the weight that’s nowhere near the value that is something that
24:50
they’ve learned merchants track and they have adapted to it. So now that we’re
24:55
here with return still being a bigger challenge, what strategies can we
25:00
project forward into 2023? Uh now I hesitate to do projections earlier this
25:05
year because we don’t know what’s going to happen between now and the holidays,
25:08
but these are still things that you can work on throughout the year. If you
25:13
would like to learn more about uh combating return and refund fraud, let’s
25:18
look at about three different ideas that we can touch on. The first one being
25:24
figuring out who your biggest challenge are challenges are within your returns
25:28
and refunds. So, I’ve got three examples here. Wardrovers, resellers, and new
25:33
customers. Are any of these your biggest challenge? If your biggest challenge is
25:38
wardrobers, those would be people who buy clothing from your e-commerce
25:42
platform, wear it in real life as if it’s their own, you know, wardrobe, but
25:46
then return it to get a refund and buy something new. So, they’re using you
25:50
like a closet. Are they resellers? Those would be individuals who buy massive
25:55
quantities of goods in lots of different sizes, try to sell them on third-party
26:00
marketplaces, and then return whatever they can’t sell. That will of course be
26:04
a different pattern of behavior but still cost your company money and can be
26:08
abusive and again lead to other fraudulent activity can go hand inand
26:12
there. There’s always a little bit of a blurring between abuse and fraud that
26:15
you need to be aware of. Or is it just new customers where they are not
26:19
familiar with your brand? You don’t have really great sizing info written out
26:23
online and they don’t actually know what will fit them. So they need to get
26:26
multiple sizes sent to them in order to figure out what fits best and return
26:31
what doesn’t fit. Now, of course, beyond that, you’ve got fraudsters who are
26:35
doing their professional refunding. And so, are you monitoring what their
26:40
current offerings are? So, looking at that first bullet point and thinking
26:44
that most of those people fall into abuse, maybe resellers, not that’s
26:48
something that could be up for debate, especially if there are any physical
26:50
goods merchants here uh who would like to have that debate a little later. But
26:54
the professional refunders are absolutely setting out from moment one
26:58
to defraud your platform and to not pay for any items that they receive. So, are
27:04
you monitoring their current offerings? There are tons of refund fraud guidance
27:09
on uh the deep web, on the dark web. There are literal sort of bibles, they
27:15
call them fraud bibles, written out per merchant that will say what items you
27:20
can and can’t purchase, the dollar value that you’re allowed to purchase up to,
27:25
and to give other steps within that process. And if you’re aware of what the
27:29
fraudsters already know about your brand and how they’re targeting your brand,
27:33
then at the very least, you could step in and, you know, take control there
27:38
over that data that you do have access to and you are aware of to try to get a
27:43
start on seeing where your refund fraud problem is and then taking steps within
27:48
that once you’ve identified those orders and those transactions to try to
27:52
mitigate that refund fraud. And there’s another thing to consider that I’ve
27:56
heard come up a lot in merchant discussions over the past few months,
28:00
and that is considering whether or not there’s any kind of I use the word
28:03
consequence. Maybe I shouldn’t use that. Maybe that’s the uh the old high school
28:06
teacher within me that used a stronger word. Maybe it’s just a there’s a better
28:11
way to describe it, but should there be something that you add to the return
28:14
process that makes it a bit more difficult to return? And a big example
28:18
of that being a restocking fee. Now, maybe if we go back to that top level,
28:23
if you’ve identified someone who might just be a new customer and is ordering,
28:27
you know, a couple of of options of sizes because they don’t know what fits
28:31
and you’re going to take steps, of course, then to make sizing clearer and
28:34
to help them understand uh better which size they should pick from the
28:37
beginning, maybe they don’t get a restocking fee applied to them. But
28:41
maybe that reseller, maybe that person who bought, you know, 200 units over
28:45
multiple sizes and honestly is probably targeting your newest drops, your most
28:49
popular items, maybe they get shown at checkouts the potential that they’re
28:54
going to have to pay a restocking fee on anything that they return. And maybe
28:58
that makes them find you a less attractive target for return or refund
29:03
fraud. That’s just something to consider. It’s going to be different for
29:06
every vertical and every company. Um, but here’s three potential strategies to
29:10
be worked through with the goal in the end of still providing a great customer
29:16
experience for your legitimate, you know, happy customers and deterring
29:21
those opportunistic and malicious first-party uh, fraudsters or misusers.
29:27
So, let’s look at a couple more examples here because I know that prior slide
29:30
went really heavy into physical goods uh retail and I wanted to provide something
29:33
that’s a little bit of a broader view when you’re talking about friendly fraud
29:37
or firstparty misuse. So, I’ve got two suggestions here for dealing with that
29:42
opportunistic side of it. One would be if you can automate your return process
29:46
to optimize for data collection and customer education. That can be
29:50
extremely valuable. I heard of some merchants who for example uh make sure
29:56
that they are monitoring not just the number of orders that are returned by an
30:01
individual but the number of units within orders that are returned because
30:04
that provides a better view to their data of what the actual impact is on
30:09
their organization. Then I’ve got another bullet point here of add
30:13
customer outreach to your chargeback representment process. I love this
30:16
because I find that so many consumers feel like a merchant will never hear
30:21
about a charge back, that they’ve just filed it with their credit card company
30:25
and then all of a sudden they get their money back and it’s as if nothing ever
30:29
happened. But actually completing that loop and reaching out to them and
30:33
sending an email acknowledging that you received the dispute can either help you
30:37
get through a dispute in the case where it might have been an accident or we’ll
30:41
at least let that consumer know, hey, we do see this behavior. This should not
30:45
turn into a regular abusive uh process and that gives you a great piece of
30:50
documentation for your representment. Then under malicious, I’ve got two
30:54
bullets. One being learn from and claw back fraudulent refunds. I know it’s not
31:00
easy to talk about clawing back refunds, but maybe when we’re talking about a
31:04
dynamic pathway for certain customers trying to prevent this type of abuse. If
31:09
somebody has become abusive, perhaps they don’t actually qualify for a direct
31:14
refund. They qualify for store credit or some kind of credit within their account
31:18
that then gives you more of an ability to revoke that or pull that back when
31:24
your team is able to examine their refund pattern and has determined that
31:27
it is either abusive or fraudulent. And then lastly, stay informed of emerging
31:32
patterns of chargeback abuse. As we as merchants crack down more on return and
31:38
refund fraud, it leaves those individuals with no other choice than to
31:42
file a chargeback dispute. And I’ve seen some really interesting ones coming
31:46
through. Uh I’ve seen a lot more use of the non-fraud reason codes. I think that
31:50
could be because there is some difficulty of uh for some individuals to
31:55
continue to use that fraud reason code. And that ties in with the bullet point
31:59
directly to the left about adding customer outreach to your chargeback
32:03
representment process. The credit card issuers are able to see the entire
32:08
activity on a consumer’s credit card. And therefore, if they see, you know,
32:12
multiple refunds and returns across multiple different merchants, being able
32:16
to call out that you reached out to the individual, you didn’t have
32:19
communication, showing uh evidence that a empty envelope was returned or a box
32:24
of salamis, being able to provide that information back to the issuing bank can
32:28
uh in fact help you increase your chargeback dispute win rate. Maybe not
32:32
immediately, but especially over time as they start to see that pattern of abuse
32:36
within their card holder. So, we’ve talked then about uh
32:42
first-party misuse. We’ve talked a little bit about the recession. Let’s
32:46
talk about taking all of that data within our system and convincing
32:51
management of whatever we need of advocating for ourselves as fraud
32:56
prevention and payments uh professionals which is I know something that can be
33:01
sort of easier said than done in many cases but I still hope that you get some
33:05
takeaways that you can use within your organization. So let’s start with the
33:11
idea of reporting. I’m sure especially if you are a leader of a fraudrevention
33:16
team, you report on some if not all of the metrics that are listed out here in
33:22
bullet points. Whether those directly tie to financial loss like chargeback
33:27
rates or refund rates or if they tie to account activity, like maybe you’re an
33:32
organization that really prioritizes new signups or monthly active users, you
33:37
might look at those atto rates or customer insults. anything that you’re
33:41
looking at within there. Being able to easily access that data and pull it and
33:46
report on it is something that unfortunately a lot of organizations
33:52
don’t have. And when you do get that data and you are able to report on it,
33:57
you might be seen as uh the bringer of bad news. uh or certainly at least not
34:02
seen as revenue driving which is uh you know a category that I found myself in
34:07
quite often as being seen as the person who just talks about the doom and the
34:11
gloom but it’s still necessary information to report on. So, let’s
34:17
focus on these quote unquote pieces of bad news here and then the argument of
34:23
getting more funding for your trust and safety team with the goal of bettering
34:28
your accuracy or bettering your ability to both stop fraud and to allow good
34:35
legitimate customers to have a friction-free experience. So we’re at an
34:40
organization that is focusing on growth and you are then reporting on these
34:47
numbers with the goal of bettering your accuracy. One thing that you can use as
34:51
far as a tool is to not just report on the fraud loss that you experience as
34:56
far as a dollar value of what you missed within manual review or what your
35:01
automated systems missed, but actually look at what that true cost of fraud is.
35:05
Now I pulled these numbers from the recent uh MRC or merchant risk council
35:11
global fraud and payment survey from 2023 which they just released this week.
35:16
It also is a comes from the insource of Lexus Nexus that’s well known for uh
35:20
releasing these true cost of fraud numbers every year. It may differ for
35:24
your organization, but this is looking at not just the impact of that dollar
35:29
value, but everything behind it from the operational costs that are spent
35:32
fighting fraud to other, you know, fees that are built into payments and that as
35:37
a whole. And they find then that for every dollar of fraud loss, that could
35:41
cost merchants $375. So that could be then a significant uh
35:47
impact that is beyond just reporting what fraud you’re losing. that could
35:51
then show how that truly affects your overall organization. But now maybe in
35:57
this instance, instead of just saying, “Hey, look, fraud is actually worse than
36:00
just the dollar value that we lose,” you can tie that back to metrics that other
36:05
stakeholders within your organization outside of the fraud team are very aware
36:09
of, especially those that might be on that growth or revenue driven side. And
36:14
you can also then focus on false positives. I know I gave that away in
36:18
the header there a little bit, but if you look at this formula right here,
36:21
it’s one that I’ve used uh quite often with merchants. You can use it in whole
36:25
or in part, but look at the number of false positives within your system. And
36:30
then multiply that by the average user CAC and LTV. CAC being customer
36:36
acquisition cost and LTV of course being lifetime value. So let’s say then that
36:42
we’re arguing for more budget or for more impact to improve the accuracy of
36:47
our fraud prevention process. We are able to represent our true cost of
36:51
fraud. And then also with this number, we’re able to represent once we’ve
36:54
determined false positives, which if anyone has questions about that, please
36:57
let me know. That’s something that we can touch on for another time. We’re
37:00
also looking at what kind of money we have quote unquote wasted by bringing in
37:06
a customer and then turning them away just as soon as they arrived. Uh, so I
37:11
found a lot of varying numbers online for customer acquisition costs for
37:15
consumer goods. Anywhere from $20 a customer up to $80 a customer. But in
37:21
general, that ratio should be about one to three of your customer acquisition
37:25
cost to the customer’s lifetime value. So if you spend $20 to bring in a
37:29
customer, then you uh hope that they spend at least 60. If you spend $80 to
37:35
bring in a customer, you hope they spend at least 240.
37:38
But if that one customer that you’ve then spent, let’s say it’s on the $80
37:42
end, makes a $100 order and it turns out to be fraud, and then that $100 order
37:49
actually cost you $375 worth of fraud loss. You can already see
37:54
here how the bigger picture is uh sort of exponential as far as what you’re
38:00
able to represent with your if you’re focused in on this type of reporting and
38:06
speaking the language to other stakeholders within your organization
38:10
outside of fraud. This is just one example. This is just one potential
38:15
approach. There’s a lot of other things that can be done um with different kinds
38:19
of risk assessments and road mapping for your company, but it’s just one piece of
38:24
a puzzle that I want to make sure we get to look at here and call out and that is
38:29
ensuring that you cover all of the steps that are necessary for your company to
38:33
be able to fight for and to win at scale and reporting. So understanding your
38:39
data is just the first step within this process. After that, you’ve got to make
38:44
sure that your analysts, that your teams have both the tools they need to
38:48
identify fraud and the feedback to let them know how accurately they did. So,
38:54
you know, whether or not they’ve met your specific SLAs’s and KPIs for your
38:58
organization. Then you’ve got to look at your workflows and rules within your
39:04
fraud prevention system, seeing how best you can update those and keep them fresh
39:08
and making sure they scale with your organization. Whether that’s because
39:11
you’re increasing your order value or expanding to new geographies, then the
39:16
ability to use machine learning. That’s really where that scale component comes
39:20
in strong because you need to be able to automate things that are a higher volume
39:26
of data than any human could handle. And lastly, scanning within your
39:31
organization what high volume digestible data you’ve got in order to ingest that
39:36
into your fraud prevention system. What different signals are there that are
39:42
strong indicators of potential fraud? I take you all the way back to the
39:46
beginning of this presentation when we talked about job listing scams and how
39:49
that one platform found that 100% of the scam listings had the job uh type set as
39:55
remote. That’s a great high volume digestible data signal right there. But
40:01
tying these all together to have that complete package of fraud prevention.
40:06
Now, my last takeaway for you will be to call back for resources and being able
40:11
to find ways to be assisted and work with other merchants towards these
40:15
goals. I mentioned the uh merchant risk council report which is a great resource
40:20
but I also would get in big trouble with our marketing team and you know feel bad
40:24
about my the effort that I and my team have helped put into it if I did not
40:28
mention that released a report just today our Q1203 digital trust and safety
40:33
index which focuses on payment fraud data and insights from our network. top
40:39
three uh takeaways that I can call out from this report that you can read on in
40:43
more detail is we did a consumer survey asking people about their spending
40:47
habits and actually got 16% of people to admit to having committed payment fraud
40:52
on their own. There is your first party misusefriendly fraud category. We found
40:58
that buy now pay later fraud surged 211% between 2021 and 2022. makes sense that
41:05
people would use buy now pay later when they’re attempting to stretch their
41:08
dollars further and be able to afford larger purchases during an economic
41:13
downturn. But the fraudsters have also figured out very key vulnerabilities
41:17
within that payment type. And then lastly, we found that 17% of consumers
41:22
encountered online offers to commit fraud. That ties into what I briefly
41:26
touched on with fraud as a service with people sort of having that entrylevel
41:30
step into fraud where they themselves aren’t the ones stealing consumer data
41:35
or stealing credit card numbers, but they find a fraudster who is willing to
41:40
help them along the way to get an item for free. And as they continue to see
41:44
these advertisements on social media and any other platforms therein, they may
41:49
unfortunately become more likely to participate in those scams.
41:54
So, all of that said and done, it is time for questions. And John, uh
42:02
you can let me take a drink of water now because I definitely need one.
42:08
>> Well, that uh that was quite a uh quite a presentation, Britney, and thank you
42:14
so much for that. Uh very informative and full of good information and advice.
42:19
And I uh and I think that indicative of that is that we’ve been getting some
42:24
good questions uh very good questions from the audience. Uh so if you’ve uh if
42:29
you’ve had that water uh and uh what did you whistle we can get going with these
42:35
um and uh the first one from the audience
42:39
these are all from the audience uh is asking um what are good strategies
42:44
to combat product not received. Uh the um questioner asks, “We already require
42:51
signature upon delivery, but we still have fraudsters open such chargebacks
42:56
and win them too, even if they do not make any attempts to contact us to
43:01
appear legitimate.”
43:04
>> Yeah, that’s something that I absolutely dealt with when I was a physical goods
43:08
merchant. Um and also when I was at First IPS and we had luxury items, we
43:12
would very very often require signature upon delivery. There are some strategies
43:18
there that you would be able to implement. I’d like to say unfortunately
43:22
though, u I’d want to see your entire chargeback representment document to
43:26
make sure that it’s formatted in a way that sets you up for the best possible
43:31
win. Um, for example, is it where an agent who only has 30 seconds to a
43:36
minute to review your response, can tell from the very first page what the
43:40
summary of your argument is, knows where they can find your signed proof of
43:44
delivery, and has already seen that you’ve called out that that proof of
43:48
delivery matches the card holder name and is to the verified billing address.
43:53
So, shipping equals billing. Are you calling that all out? And I would say in
43:57
the first top half of your representment document. Now, if you’ve set up your
44:02
document so that it is optimized so that that evidence is clearly then visible to
44:06
the issuing bank when they’re reviewing the dispute, you may then need to change
44:11
your strategy for item not received or to anticipate item not received claims
44:15
for particular items. I have heard from one major big box retailer how they’ve
44:20
actually um started putting technology into packages that can send a message
44:25
back to the retailer that confirms a box has been opened. And so let’s say if it
44:30
was delivered at 3:54 p.m. on Tuesday and the box was then opened at, you
44:36
know, 3 um well, I shouldn’t have said 354, but at 4 p.m. they can show that
44:41
the opening of the box was very, you know, close in time to that proof of
44:44
delivery, which sort of strengthens their case that it was delivered. If the
44:47
box was never opened or the box was opened hours later, again, that could be
44:52
legitimate behavior, but they’re building up their own data set to start
44:55
analyzing if that’s helpful. So, that could be something where if you’ve
44:59
already got some of those ideas about item not received based on your own
45:04
data, you could maybe potentially find your patterns there. Um, and then
45:08
another key indicator could be, do you have your delivery service take photos
45:13
of the item being delivered? Um, are you able to use a full proof of delivery
45:17
which shows the actual customer signature or are you only able to use
45:20
the smaller version that doesn’t show the customer signature with the full
45:23
delivery address? There’s different ways I would want to break that down. Um, if
45:27
you have any questions, you can feel free to reach out and we can go into
45:30
more deep uh more depth on your particular representment strategy, but
45:35
there are some highle answers there and I hope some of those were helpful.
45:41
>> Yes, absolutely. Uh, thank you for that Britney. Um, we have time for a few more
45:47
questions. Uh and uh one has come in asking uh broker marketplaces especially
45:54
virtual ticketing are highly vulnerable to firstparty fraud. Have any tactics
45:59
proven effective in this market vertical?
46:03
>> Yes. So I have spoken to some ticketing platforms and I think one of the biggest
46:08
takeaways that they shared with me is that their patterns uh for velocity or
46:13
for targeted events change daily because it depends on what event is happening.
46:19
Is it a giant Britney Spears concert or is there nothing like that happening on
46:24
this particular day? And then of course it is increasing as the start time of
46:30
that event approaches because you’ll have people who become sort of more and
46:34
more desperate to sell off the tickets with their resellers and increasing
46:38
their activity there. So, as far as what has been successful, uh, I do think
46:45
categorizing your events in a way that represents risk, either by them being
46:50
the most attractive events or potentially events if you’re seeing any
46:54
kind of money laundering. I’m not sure if you’re a marketplace where people
46:58
have to go through a high threshold of friction in order to register or if you
47:02
allow like a P2P sale where someone just needs to create an account with their
47:07
email address and a password to sell. So depending on what your what your
47:10
friction there is for your sellers, there could be more that you’re looking
47:13
at on the supply side. Um, but I would say most tactics that I’ve heard of of
47:17
being successful will take those events, we’ll break them down into sort of that
47:21
riskiest set of categories and then we’ll focus in on what the different
47:26
levels or different patterns of fraud are as the event date approaches. and
47:31
then especially focusing in on the first few hours of sale for those big events
47:36
and then the last few hours before the event actually happens.
47:44
>> All right. And um we have one that has come in recently
47:51
asking um what are your thoughts about uh excuse me what are your thoughts on
47:56
using 3D secure to get protection against friendly fraud.
48:02
So, I’m not sure if that’s going to be as useful since the whole purpose of 3D
48:07
Secure is to prevent true fraud to authorize that the uh card holder has
48:12
actually participated in and completed the transaction. Yes, some friendly
48:17
fraud does come through with a fraud reason code or would come through with a
48:21
fraud reason code and you could be protected in those cases. But we are
48:25
seeing uh this feedback from merchants of an increase in people abusing
48:31
nonfraud reason code categories to commit their first party misuse of
48:36
friendly fraud. An example that I like to highlight of that is I was speaking
48:40
with a cryptocurrency exchange. So this was a situation where somebody used fiat
48:45
USD to buy Bitcoin and they then did not like the fact that the price of Bitcoin
48:52
fell very far and they felt like they lost money on that transaction. So they
48:56
went to file a chargeback dispute. They filed the dispute as nondely
49:02
and claimed that they had purchased a pallet of wood.
49:07
Does not fit the merchants’s type of business whatsoever. does not fit what
49:12
the good was that they actually purchased but still got through the
49:15
system and then made a very difficult uh approach for the merchant who then had
49:21
to figure out how to describe you know I am actually a cryptocurrency exchange
49:24
this was actually a purchase of digital currency it had nothing to do with a
49:28
physical good and so now that the credit card company is looking through this
49:32
representment document and trying to find a signed pod for a physical good
49:38
there’s a big risk there of the merchant losing that chargeback dispute. And
49:42
we’ve seen more and more fraudsters learning the chargeback system and
49:47
betting on how they can then commit first-party misuse via these non-fraud
49:52
quote unquote non-fraud channels. So that said, yes, 3DS will help you with a
49:58
certain subset of firstparty or friendly fraud that isn’t as sophisticated. Um,
50:05
but you definitely need to keep looking forward to seeing how each type of fraud
50:10
evolves because you can’t just assume that 3DS will get rid of all friendly
50:14
fraud.
50:21
>> All right. Yes. Well, anybody who u is surprised these days by the volatility
50:25
of cryptocurrency hasn’t been following the market very closely. Um,
50:32
and let’s see. We have, um, excuse me just a second while I sort
50:41
these. Uh, a bunch have just come in. And um
50:51
the question is beyond money saved, how do most merchants present their fraud
50:57
risk team as valuable as as valuable? Let me reread this. Beyond money saved,
51:05
how do most merchants present their fraud risk team as um a valuable uh the
51:11
question reads a piece to their business? I think they mean a valuable
51:14
asset to their business. Oh yeah, definitely. I mean, I’ve worked
51:20
uh where I’ve managed fraud teams and there’s been the request from upper
51:24
management to outsource where I knew that that would not be the correct
51:27
decision. Although, of course, it would save money uh in the short term.
51:32
Doesn’t mean that outsourcing isn’t the right decision. In all cases, it can be,
51:35
but uh when you’re trying to argue or you’re trying to demonstrate the value
51:40
of your team, um I think one really good example that I can point back to is not
51:46
just the amount of fraud that they prevent because you’ll often get the
51:49
argument that oh, some of that’s false positives. That’s not really all fraud.
51:53
You can’t be conclusive. You can’t be certain. but to maybe point to the
51:57
feedback that you get or the interactions that your team has with
52:01
your customers that are positive. Uh, one example I want to call out is a
52:06
major company I know where they were able to take their NPS scores or net
52:11
promoter scores which you know come from consumer surveys. I think we all know
52:14
about that. And they were able to take their NPS scores for individuals who had
52:19
experienced an account takeover, confirmed atto and then went through the
52:24
process of having their account restored. And that NPS score became the
52:30
highest and most positive NPS score of the entire organization. It beat out
52:35
people who just, you know, wrote in asking to have their password reset or
52:39
wrote in asking for a tracking number. for very general inquiries. It was a you
52:45
know difficult process where people feel like their account has been sort of
52:49
invaded and maybe items have been purchased that they didn’t purchase or
52:53
you know they think that their payment method has been compromised. It took
52:56
that very damaging impact on a brand of atto and through their process of open
53:04
communication with the customer and with making sure that everything that a
53:08
fraudster could change could easily be reverted and any of those indicators of
53:14
the ATO would be removed before the card or before the account holder logged back
53:17
in. And then of course being able to have any sort of post uh support with
53:21
them to answer their concerns. they turned that into their highest NPS organ
53:25
uh NPS response of the entire organization and that just showed how
53:30
vital that trust and safety team was. So maybe thinking beyond just dollars of
53:35
fraud loss stopped but to what are the goals of our organization? Are we
53:40
focused on customer experience? Are we focused on customer retention? What are
53:44
we actually trying to hit with those numbers and what does my team do to
53:48
impact them? And then flipping the way you speak that. Uh, another one from my
53:53
experience was at one point I was working uh at LetGo and we had a lot of
53:58
B2B sales with signing up used car dealers to sell on our platform. We of
54:03
course got fraudulent leads and signups via that portal and it wasted our sales
54:08
reps time because they would follow through, engage with these fraudsters
54:12
and then of course not get any commission and not actually bring any
54:15
value to the platform. And so I framed the work that my trust and safety team
54:20
did on screening those signups in an automated way and actually working to
54:25
prevent the fraudulent uh leads from ever getting to a sales rep. I framed
54:29
that as helping the company be as efficient as possible with those sales
54:34
reps time and got the buyin from the sales team because then they didn’t have
54:38
as many frustrated uh sales account executives who never got their
54:42
commission at the end of the month because they signed up a fraudster. So
54:45
thinking in that way are two examples that I would throw out.
54:50
>> Okay. And I think we have time for just one more question. And uh this is um
54:56
asking if you are a digital good merchant uh can you allow
55:02
All right. Yeah. If if if the uh uh seller here is a digital good merchant.
55:10
>> Um can you uh Britney elaborate what you recommend for data collection and
55:16
customer education? for example, a survey via email communication,
55:20
informative email communications, etc. Um, any advice there?
55:27
>> Yeah, so we got two parts of that question, data collection and then
55:32
customer education. Uh, the customer education part might be me throwing out
55:37
that word over and over. And by that I mean, you know, making your policies and
55:41
terms clear and easily surfaceable to a consumer as they’re proceeding through
55:45
checkout or making a purchase. Um but also then just anticipating
55:49
any potential issues and sending outreach as necessary. I mentioned
55:54
crypto earlier that is technically you falls in that digital goods bucket and a
55:58
lot of crypto exchanges and wallets did great proactive work over the past year
56:03
of sharing warnings about scams that their customers could fall for. And that
56:08
would then be a good example of that customer education via via email. But
56:13
when we flip it back to data collection, I’m not sure what the goal would be. So
56:18
what the goal of collecting data would kind of then determine what data you
56:22
collect. Uh but let’s see. So if you’re looking for data collection for
56:27
chargeback representment, let’s put it that way, and you’re looking at digital
56:30
goods. I think we had someone else who was asking about, you know, visa uh
56:35
compelling evidence 3.0 changes and whether or not that’s something that’s
56:40
actually to be beneficial. I know it’s really tough for digital goods merchants
56:43
who then don’t have a shipping address. But we do think that changes will be
56:47
coming to 3.0 uh as it rolls out and over the course of the year where then
56:51
maybe a data element that becomes allowed as a proxy for a shipping
56:55
address is a proof of download is part of an activity log is the email address
57:00
that access something. So basically telling the story if we’re talking about
57:05
data collection for chargeback representment telling the story of how
57:09
the card holder uh did actually you know participate in a transaction or it did
57:14
actually receive the good as described is what I would focus on first. You can
57:18
look through the user journey. You can just look through whatever that pathway
57:22
is for disputes and that’s what should inform data collection. Beyond that, if
57:27
we’re not talking about chargebacks, um I could throw out some tips about, you
57:30
know, the volume of data that you’d be collecting, making sure you stay within
57:34
different privacy regulations, and then therefore have a process in place when
57:38
you get data deletion requests such as those that comply with GDPR. There’s a
57:42
lot I could say there. Um, if whoever asked that question wants to follow up
57:46
later with a little bit more specifics, I’d be happy to take it because I do now
57:49
realize that we are at time.
57:53
>> Yes, indeed we are. uh Britney, but thank you so much uh to you for that uh
57:58
extremely uh interesting and enlightening u presentation. I’m sure
58:01
the audience gained a great deal uh from it. Thank you to our audience and uh
58:06
thank you to Sift uh for uh for today’s uh uh presentation and um it’s uh time
58:14
to say goodbye. So, thanks to everyone uh who attended and spoke and uh we’ll
58:20
look out uh for uh letting you know soon about the next
58:25
presentation. But thanks for now and take care.