Dropship Dollars: Scaling E-commerce on Native Ads | AWeurope 2016
Articles,  Blog

Dropship Dollars: Scaling E-commerce on Native Ads | AWeurope 2016


Alright, so I’m here to talk briefly
about how to scale ecommerce on native ads. I don’t have a whole lot of time, so
I’m gonna do a very brief case study and then talk about the general scaling
process that you would take on many of the top native ad platforms. So
super short bio here, I’ve been an affiliate marketer for about five years
now. I’ve gone through many of the big affiliate waves, from dating to app
installs, binary, nutra and as many of you have probably faced, you end up
having to start over quite a lot when you’re running some of these niches and
frankly, I started getting tired of it and I said okay, let’s try to build
something a little bit more sustainable. I’m in China, pretty much everything the
world is made in China, so let’s take a stab at ecommerce. So
the biggest question when you’re starting with ecommerce though, is
product selection. How do you choose products that will work especially on
native because native tends to be quite expensive. So the number one thing
that we’re looking for here is broad appeal. Many of you, if you’ve run
native in the past you’ll know that you have very limited targeting options. So
you need a product that has extremely broad appeal, these are the kinds of
things that if you were going to do a Facebook campaign you would target all
men and women over 35, or all men and women over 30 or something huge
scale. You need higher payouts due to the higher click cost that you’re dealing
with, I generally would not try a product anywhere less than $20. We usually go
more to the $40 to $60 range before we do our up-sells and cross-sells. If you can
get any moat possibilities built-in, meaning a moat around your sales funnel.
Either through the advertiser, the owner, the actual owner of the product if
you’re doing affiliate, gives you an exclusive on that traffic source, which
has happened in the past or you have your supply chain tight enough that
someone can’t just go on eBay and do your entire sales funnel right away. You
need to have either cap or inventory, able to push up to five figures, because
you really don’t want to be in a situation where you’re paying for all
this media and then suddenly you have no way to fulfill any of these
orders because your supplier only has a hundred of this product. That, you never
want to have, you’ll waste a ton of cash on that. And it’s important to
test many, many products. When you’re first trying out, you might have one that
seems like it has a little bit of a success but you don’t really have a
baseline yet. So it’s good to test or put aside a set budget and test many
products before choosing which ones to scale as big as possible and that way
you can also have the potential of having several products that you’re able
to scale up. Alright, so I have a brief example of how you would take a product
and then test it and then optimise it and then keep scaling it long-term. So
here’s just an example product that we’ve tested in the past, it’s a
Bluetooth personal tracker, they cost to consumers about $65. We
started off with the advertisers landing page, later on we made some custom
variations. We’re trying on Taboola but any large major traffic source
would be fine, and we tried a bunch of different angles
right and this is of course very important. You want to cast as wide of a
net as possible, so if you lose your keys frequently or your wallet or
your phone, this thing could help you find them. We also had an
angle where you can use it for tracking your kids even. There’s the forgetful people
angle, look at this breakthrough startup in this cool technology that they have.
We tested a bunch here. So it’s important to make sure that your angles will
follow through you’re creative as well. So this is just showing you how we
implemented these angles in the creative. It’s probably pretty obvious that these
are native because you end up having like big arrows that necessarily don’t
look good but they just get higher CTR so these tricks are also important to
try out. Alright, so how do we set this one live? How did we do this? Well we
definitely wanted to focus on english-speaking countries first, because
that’s where the advertiser told us most of the sales were coming from and we
added in Germany because that was a good one as well from them.
We started off direct link because we just wanted to get a baseline, we wanted
to figure out which device types will work the best. Mobile versus desktop and
tablet as well and which countries we should focus on first. So we’re gonna
keep any country that lost less than half of the test budget. So our initial
results from that test, we had to focus on UK and Canada and the goal was to
scale to the US. Now with native traffic, the US tends to be quite expensive, if you
have a product that’s only able to be sent to the US or you’re doing something
for example with the e-packet shipping method, which is quite good for China to
US then you can dive right in but for this particular funnel, we decided to
start in less expensive geo’s then optimise the funnel and then port it
over to the US. So how do we optimise it? Of course we start with the largest
factors first. You want to get the maximum bang for our buck, so we wanted to focus
on getting cheap clicks so that we can then use the cheap clicks to run all
sorts of A/B tests on the funnel. So we nailed down the angle, we nailed down the
device type and we nailed down the country. Then within the angle, we’re
studying our high CTR ads. So not necessarily which ads have the
highest CTR but what is the psychology driving the CTR between those ads. Maybe
you can take the super high CTR from one angle that didn’t necessarily convert
the best and move it over to one of your other angles just by following the exact
same psychology from the creative. It’s also very important to use dynamic
tokens, this is pretty standard if you’re running any sort of native
campaign but you don’t have much to optimise on. So be sure to get those
dynamic tokens tracking right into volume or into whatever tracking system
you’re using. They’re pretty priceless for the creative and site ID testing and
when you create your blacklist as well, which I’ll talk about in a minute.
Intermediate conversion metrics are very useful, I told you here that you should
be running relatively more expensive products compared to like
an iPhone lead gen or something like that and so an intermediate conversion metric
can be a great way to get a lot of data early on without having to wait until
you have 40/50/60 actual purchases. So I’ll talk more about how to
do that and where we did that shortly. Another way you can do it, another good
way to optimise is once you see which categories are starting to work, so Revcontent does this, Taboola, a bunch of others offer the campaign category or at
least the category of sites that the campaign is being shown in. Sometimes there
will be certain categories that just have really, really bad CTRs and
other ones that are just performing great for you and so it’s the
the low ones that can actually really bring down the rest and the algorithm
will punish you for this and so once you have an idea of which categories seem to
be working well, it’s good to break them out into their own campaigns. Okay,
another thing that definitely needs to be talked about is bots and bot traffic,
if you’re doing some of the higher-end native networks for example, Outbrain
Taboola, Revcontent, branded stuff, you don’t need to worry too much about this
but if you’re anywhere else then it’s definitely a factor. So what we do is we
set up a system of rules and then we have a report that runs on all the
traffic and if you pass this test, this test, this test. Okay, it’s not a bot. If it
fails say two out of our five tests okay, it’s highly likely to be a bot and so
the thing with this is, what do you do once you know that you’re getting bot
traffic. Well, the main thing we do is we use it as a earlier metric for possibly
black listing a site earlier than we would normally. That is to say we don’t
want to blacklist a site if it has lots of bots but it’s still profitable, it’s
fine. We usually just leave it in that case but it’s good to at least know what
you’re dealing with with the traffic you’re buying. So for LP testing, you want to
test widely as possible just like we’re doing with the angles. We want to make
sure we’re testing an advertorial style, a video sales page, we want to test
direct to the product itself and also we want to spy a little bit.
Now, the spying I’m talking about here would be just going
into your volume refers after you’ve been running traffic for a little bit
and you have the converting site IDs and then check out the refers, match up the
site IDs, by like going into the source and you just check wherever the Revcontent widget is or the Taboola widget and it’ll match up to the site ID
that it sends to you, and this is a really good way to get ideas for more
advertorials because you know the exact type of content that you’re your highest
converting sites are giving to the readers and so you can do something more
in the style of them, if you’re catering to for example on a network
like Taboola where 90% of the traffic comes from MSN. It’s a very
useful way to come up with ideas. Intermediate conversion metrics,
we’re back to this, so typically we set it up on either the
add to cart or on the click through on the advertorial itself and so by moving
it forward, you actually get a magnitude more of data early on
and this is very, very useful, especially if you’re using the
the conversion pixels given by the traffic sources themselves. Another thing
very important for all your LP testing, I do this kind of a standard practice now
and my partners actually laugh at me for it sometimes, is FK readability this is
the Flesch-Kinkade scale for readability. People have probably talked about this
in the past, but essentially what it is, is it tells you what reading level your
ad copy is, what grade level it would be and this is actually a function built
right into word, but there are also other apps that can do this really well. The
one I usually use and I think real copywriters will laugh at this, but
there’s one called a Hemingway app and for me it just does the job really well.
I’ll write up the ad copy, I’ll put it into Hemingway and it’ll turn sentences red if they’re like
really confusing, it’ll turn them yellow if they’re a little bit confusing and so
usually I can bring down the grade level from around like a 10th grade reading
level. If I can bring it down to under 5th grade then I know that of this
broad traffic that’s coming in, they’re going to immediately understand at least
the offer that I’m providing because you only get a few seconds to engage these
people, especially when you’re giving them really strong click-bait headlines
and so it’s good to bring the the reading level as low as possible to make
sure people are understanding what you’re saying. Okay
creative testing obviously this is the same stuff you can do for
non-ecommerce but the thing is you can’t steal directly because the chances of
someone having the same product as you when you’re doing ecommerce is pretty
much zero. So yeah you can read BuzzFeed and get some template ideas, you can read
copy bloggers and understand or the psychology behind different headline
styles. If you’re gonna spy using spy tools, copy but tweak it a bit right
because the chances are that the stuff you’ve seen there has been used for
something, somewhat like what you’re doing but not exactly the same, it
could already be burned out on the network. It’s just worth it to at
least split test their version versus your own version of what they did. And
then again, this is a kind of a standard native tip, you want to Photoshop all
your images to make them pop a little bit more. I believe that there
was a guide for those of you that are on STM about how to do this in detail
with all the different Photoshop tricks but the general, safe one is just go to
vibrance and jack up the vibrance quite a bit and that’ll help a lot. So
how do you scale up with ecommerce? Well the good thing about ecommerce is that
because you’re doing pretty much everything above board you can go with
the top traffic, you don’t really need to be fighting against people on the
the cheapest traffic possible on Adblade or something like that. The good
thing about that is you can get the branded campaigns from for example, Revcontent, you can get on Outbrain and you can get on all this good traffic.
To optimise it, still stick with a standard kind of blacklist of the rundown
network and then moving over all your good stuff into white-list campaigns. If
you want to get very granular about it or if you’re on Revcontent,
they have a really good API so you could do this programmatically,
where you just sort things into basically different tranches, by like how much
profit they’re giving you. So if you can put higher bids to those sites that are
really, really converting well and then basically make almost per site bids.
Conversion pixels, I put a star next to it just because some people are very
uncomfortable with sending back data to these traffic sources. When I use
intermediate conversion metrics and I have built my moats, so I know it’s very
hard for them to get my supplier, I have a very complex funnel,
I usually don’t care because the conversion pixels at least in my
experience have done a lot of benefit and if you’re on a platform like Yahoo
Gemini, there’s not much you can even do to optimise, so you pretty much have to
use their conversion pixel. So this is at the point where you’ve pretty much gone
through all the the basic steps of optimisation, this is the point where you
could go back and try scaling to the US or try scaling to any of your other
geo’s. Try building out the mobile part even though the mobile part did more
poorly in the beginning. And this is one small tip for most of the native
platforms that I’ve worked on in the past, is you need to maintain a
surprisingly high account balance because the algorithm will
punish you if it believes that it might have to go negative to deliver
your traffic. So generally I’ve heard anywhere between $2,000 to $3,000
for say Revcontent and Taboola I would probably do even more just make sure you
don’t come anywhere near to empty and won’t the algorithm will not punish you.
Okay now on the offer product side, obviously the clearest way to scale up
is to roll your own offer. So check out Aliexpress for dropshipping, eBay, Global
Sources these are all places where you can find tons and tons of products. If
you see for example something on Facebook right now, where people are
doing tons of these free plus shipping things, you can probably just
search the direct term into global sources or Aliexpress and find
exactly the guy they’re buying it from.
So when you do find a good supplier of product definitely, definitely
rate them on communication. I had a situation just a few weeks ago where we
were scaling up a particular product and we started having an issue because we
needed some feedback on what happened with a particular set of
orders and it was essentially impossible to reach the Chinese supplier. And this
is even with me being in China and using all possible back roads to contact them
and so needless to say we found one who was I think it was like 10 cents more
expensive per unit and now we just do all our business with them because they
actually reply to our emails and we know we’re not screwed if something bad
happens. Definitely push for better terms then you’ll get off eBay or Global
Sources but only do it once you actually have real sales. This is kind of common
sense, you need to show them that you’re a real customer, a real client
before you can get them to push for example, free shipping or a locked in
much lower rate on all product. Now if you are doing some work directly with a
client that owns a product or you’re doing an affiliate-related sale, then
it’s a probably a smart idea to pass angle IDs, device type IDs, traffic source
IDs. Now I definitely say IDs right, I’m not telling you, say oh yeah I’m doing
this angle, it’s working great, here’s my traffic source. Don’t give your
campaign away, obviously, use an ID say like oh traffic source 1, traffic
source 2, traffic source 3 and that way not only can they tell you when it’s not
working but they could say hey, traffic source 2 is doing great, let’s
like pump up the volume on that and I’ll boost up your per product sale
by 5 bucks. If you’re not sending your data in a granular fashion like that,
it’s impossible for them and for you to see where the real money is in the
campaign. Okay, this is very, very crucial. It’s not necessarily easy to do, but do
whatever you can to get an account manager, once you have
a campaign that’s working at a source that you like, get an Account Manager
ASAP. They give you constant feedback, they can tell you for example if you get
your ads rejected for some weird reason, they’ll tell you what the real reason is
and they’ll push your stuff to the front of the queue. It makes life infinitely
easier, for those the bot reports that I was talking about before, if you have a
good rep you can send the bot report over to the rep and get a refund on all
your bot traffic. That’s awesome. Faster ad approvals like I said, let’s
say you want to take this campaign that you’re running and you have it now in
the US, UK and Canada. Okay, you want to test out some European countries, if you
have a good rep they’ll give you a list of sites, say hey like this is like a
standard blacklist for Switzerland just like block these off the bat and you
won’t have to waste your budget. You can get really, really good tips from Account
Managers. So do whatever you can to get one. Okay and then just ongoing, this is
something you’ll be doing throughout the entire life of the ecommerce store or
the product, if you’re just pushing individual products, generally for
ecommerce you’re on native and on other platforms, you’re not going to be using
volume or attractive systems like that so much. It tends to be more either mixed
panel or even just Google Analytics. Google Analytics has really, really
strong ecommerce components and a lot of our experiments we just do right in
there. So what we do is we focus on getting as many people clicking through
on the advertorial and then we worry about the rest then, once we have people
clicking through then we can worry about how exactly we’re monetising to get the
maximum revenue per user. But first we definitely start at the advertorial
itself, once you have that tweaked then you can worry about your shopping cart
up-sells. We like to do the the shopping cart up-sells mostly after the CC has
been entered, after they’ve already submitted the credit card and this is
really, really powerful because you can test pretty much anything and yes, it
could not do as well as something else but it will never jeopardise the sale
because the sale has already been made. And so, even if they just exit
your website at that point, they’ve already submitted their
credit card for the first product and so it’s one-click
up-sells after the sale. That has been huge for us, that’s helped a lot. Then you
can also experiment with buy one get one, percentage discounts, all this stuff has
been very good for our revenue per visitor. If you’re doing mobile, which you
definitely should be checking out because a lot of native ads inventory is
strictly mobile, definitely do some speed optimisation. This is kind of standard
operating procedure for mobile, definitely do that. If you’re running on
a system like click-funnels for example, which is great for building out funnels.
Once you have a tweaked funnel and if it’s mobile focused and you have a
tweaked funnel already, I would even suggest getting that funnel done in
actual code because you’re gonna get a huge boost just by having it in
optimised code instead of this WYSIWYG funnel. Once you’re at that point, then
you can worry about the re-targeting of your customers and the long-term
value there, through email marketing and abandoned carts and building out your
newsletter and all this kind of stuff. Okay so here’s just a couple tool
recommendations, there’s a lot of stuff in the space, so this is just stuff that
we’ve used that we found works really well. You can try it out for your own
system and see how it works. Yeah not too much else to say. I guess that’s all I’ve
got. So thank you.

One Comment

Leave a Reply

Your email address will not be published. Required fields are marked *