This post is part of the series eCommerce Café
Other posts in this series:
Last month we attended the SMX Advanced conference in Seattle…meaning we also had the pleasure of digging into the state of Product Ads and eCommerce with Adwords evangelist and Optmyzr CEO, Frederick Vallaeys. Check out the full interview below.
Direct Youtube link.
To stay tuned with Frederick, follow him on Twitter, @SiliconVallaeys.
ANDREAS: Fred, great to have you here. Could you explain what Optimizer does, and also what your role at Optimizer is?
FREDERICK: Yea, so, I’m one of the co-founders of Optimizer and CEO, and we have a tool to make agency life much simpler and much more efficient. So, if you’re an agency or a PPC expert, there’s certain things you want to do in PPC, that are very time-consuming – like building app shopping campaigns, something you guys know a lot about. And so rather than taking days to do it manually, we have tools and specialty workflows to make it a 5-minute process, so you can actually do testing and things they want to do, but usually don’t have time to do.
ANDREAS: And what’s your personal role?
FREDERICK: So, as a CEO, but what I actually do at the company? So, obviously I was an evangelist for Google, so I had a lot of connections in the industry. So, I still do a lot of marketing, speaking, blogging, just kind of to make sure the customers know about who we are, what we do and then my team handles a lot of the engineering and operations to make sure that, once the customer comes in, they’re happy. But I set the strategy and the direction.
ANDREAS: OK. We just saw that Google shared its product road map on Google Marketing Next, and there are things like custom in-market audiences and data-driven attribution, so we see that AI is taking over. What do you think are the implications on our industry in the long run?
FREDERICK: Yeah, I think it’s this continuous shift that we’re seeing and even when I was working on quality score at Google, that was already machine learning, at a very early stage, and it’s getting much more sophisticated. But I think the realities as account managers, we’ve always played this funny role where we had this great advertising platform on the one side, and we had all these customers on the other side that want to buy our stuff. And, basically, as account managers, we’ve been that middle man who connected those two systems together. Now, a lot of that work is being taken over by computers, and a lot of it is very mathematical and data-driven. So it makes sense to have computers do it. And, so, I think, for us, as ‘’marketers’’, we probably actually need to become marketers again more so than data analysts, because that data, that’s going to be handled by the computers and, really, what we need to figure out is, well, what kind of a message resonates with a consumer, right? And the computer will figure out, ‘This is the message for Amy. This is the message for Fred.’, but we still have to feed the machine the different options and that’s what we’ll do much more of. And, at some level, I think it’s also kind of the self-driving cars that you see, when we see you’re still required to put a human in the driver seat to oversee and make sure you don’t run over somebody. And to some degree, I think we’re going to have that same role, to oversee the system and make sure that we steer them in the right direction when they don’t fully visualize what’s going on with the accounts.
ANDREAS: Doesn’t Google, to some extent, disintermediate and is there need for agencies or tool providers on the long run and how would you position, on the one hand, an agency and, on the other hand, a tool provider?
FREDERICK: That’s the thing, right? So, Google is very much a company about scale, about technology and humans are this thing that you need to put in that flow at some point, but the less they have to, I think, they would almost prefer it at some level. And it makes it very interesting for what does it mean for our careers as account managers, tool vendors, agencies. But I think it will evolve and I think there are still things, even like Jerry Dischler said yesterday at his keynote, that humans will simply do better than machines. So, it’s going to be a shifting of roles, more towards the creative, more towards figuring out what are the insights and then what are the strategies that we take as a result of these insights, but leave the eventual number crunching and very specific… So, basically, if you think about bid management, the bid management used to be super simple, right? You basically set the CPC and that was it, but now you have all these bid modifiers and they all stack on top of each other, so even if you’re only managing five different bid modifiers, while as a human, when you have a million key words, there’s so many combinations now, you can’t stay on top of that any more. So, that’s what we should back away from, because we’re humans, and just let the machines figure it out. But the machines are often really good when there’s large data sets and for many advertisers the reality is they may not have that much data. And, so, for them to be able to see if the machine’s doing a good job there, because what’s happening is relatively similar to other advertisers or to things that Google and AI can figure out. You know, let Google handle them. The stuff where Google’s dropping the ball, you should be on top of it, you should see it and you should take back that control. And, so, I think that’s something we’re going to be doing for many years to come.
ANDREAS: Can you mention a specific example where you see that Google won’t get involved, where’s Google dropping the ball?
FREDERICK: Honestly, bid management, flexible bid strategies, a lot of people are very happy in the beginning and then there always comes a day, this is almost in 100% of cases, there always comes a day when, for some reason, the machine doesn’t figure it out and the results just tank. And that could be seasonality that Google is not picking up on quickly enough. In those situations we have to take back that manual control and the scary thing for me is whenever you have a black box that doesn’t tell you what it’s doing, we don’t know why it’s failing all of a sudden, right? And if you as a human have stepped back so far that you have no clue anymore what’s happening, what do you do on that day? You basically get to turn it off or on. That’s it. And if you turn it off, now what you do? Now you have to actually go and figure it out. Listen, let’s understand, even if the machine’s doing the bidding, it a day of weak performance a big indicator of likely conversion rates? So, if the machine starts to fail, I at least know what I can recreate to leverage some of those insights that we’ve had from the machine over the past couple of months or years. The other great example is Google had this project to automatically write ad texts and this was for very small advertisers, certain verticals, so you could imagine a plumbing company, if it’s the plumber himself, he’s not that good at writing ad texts, hadn’t thought about it, okay, so why doesn’t the Google do that? Well, the problem was that Google would make so many, and all the ads would start looking the same. And machine learning and Artificial Intelligence actually needs variety to learn. If everything becomes the same, then it’s hard for it to learn. That’s where the human element still comes into play. And you go to these sessions, I don’t know, I actually don’t believe this necessarily works, but people pull on emotional triggers in ads, where they try to get really fancy with the ad, it’s supposed to be really relevant and giving information, they try to be more gimmicky. Some people claim this works, some people think it doesn’t work, but at least this is the variety that allows us to figure out A/B testing and what we should do more of and what we should do less of.
ANDREAS: If you take a look at the key growth driver of Google in e-commerce, they are exactly cutting out this last remaining variable, which you could influence before.
FREDERICK: Which is the ad, right?
FREDERICK: But at least you still have, I suppose, the title and the description, but it’s mostly the title that will show up.
ANDREAS: We still like experiments when they cut off the title and just show the brand.
FREDERICK: Obviously PLA is a thing, Mary Meeker was saying 52% of all clicks on Google were e-commerce, so, it’s scary but it’s a thing, and I’m sure you see this in your business quite a bit, too – a lot of advertisers who are not equipped with good tools just get very few bids across shopping apps. So again, to trust Google to do that bid management for them, it doesn’t always go right. It doesn’t always update, right? And you may know something about your business, like say that you’re selling shoes. Maybe size 10 and a half sounds way better than a size 11 online. Maybe because your stores mostly don’t carry the half sizes. Or they don’t carry the odd sizes, let’s say, size 15 for men. I’ve got to go store, and exactly find a lot of choice there, but online it’s a great sell, it’s a great return on ad spend, because it’s really hard to find it in the physical store. Right? And that’s an insight you could have as a big shoe retailer, big sporting company. And maybe the machine doesn’t pick up on that.
ANDREAS: You work with retailers. What KPIs do you recommend them to use? And how does this compare to the reality of daily business of ecommerce KPIs?
FREDERICK: When it comes to KPIs, return on ad spend, that’s the obvious starting point. The tricky thing is always where do we set the return on ad spend? Do we pick out for margins, do we just have the same case, ROI starting for all of our products? Even if you have it same for all your products, what is your presumed profitability? So, I think people make a lot of assumptions, ‘Hey, we need to get 500% return on ad spend.’
ANDREAS: Where does it come from? Just fell from the sky? Why isn’t it 450? Or 550?
FREDERICK: Yeah. And that’s the thing, that’s the disconnect. I think a lot of account managers just take direction from the executives, and the executives will not necessarily know how to make the right numbers. So, yeah, you get into these funny situations where people are probably not getting the right value. So I think that some of the announcements that Google made at Marketing Next, around data attribution and data-driven models, that’s really very interesting to me, because it allows you as an advertiser to really build an attribution model based on the reality of what’s happening in the real world. Too many models, too many classical attributions, that’s based on a lot of assumptions, a lot of incorrect assumptions, so it’s undervaluing things here, it’s overvaluing other things, so I’m really excited about data-driven. But the one component that data-driven still misses is what is the margin on my product, right? You made a really interesting point, with your session at SMX, which is sometimes the bid is your lever to get more success, but often times you could have had a tiny decrease in the price of the product, which is smaller than bid increase would’ve done, and that would have actually given you more traffic, because price is such a big factor in ranking, impressions, all of that.
ANDREAS: We run our campaigns. We track performance and we kind of want to believe that this is the true value of our ad campaigns. But when we actually don’t really know what the incrementaility of these ads is, when it comes to retargeting – that’s even more complicated, because your target customers, which have visited your website, and part of them, of course, have made their purchase decision already, and they use your ads as shortcuts. So, what are we actually able to measure, or are we rather living an illusion where we make decisions on the gut feeling and we justify them with data?
FREDERICK: Yeah, I think there is, definitely, a lot of illusion there. But, honestly, we’re doing the best we can with the information we have. So, we try to get the clear picture. We do have to make those assumptions, and that’s why it’s so exciting for tool vendors and agencies like us.
ANDREAS: Let me quickly challenge this, because there is a situation where sometimes all the different participants have incentives to not raise questions about things. And because, if you have a retargeting campaign and you’re delivering great numbers, if it’s an agency, they ask, ‘Should we question these numbers?’ Their marketing executive is selling these numbers to the CEO, he would also not question them, so eventually we all know it may not be true, but no one wants to really dig deeper into it.
FREDERICK: I completely agree, and so, I’ve literally seen agencies when they come up with these crazy metrics, where they’re basically counting everything twice, like consistent conversions and remarketing… So you’ve already paid in some way for that traffic, and now you’re putting a huge value on that – that’s not the correct way to look at it. But, you’re right, good numbers that you can show to your executives makes them happy, but also I think agencies that will survive are the ones who actually do ask some of these questions, like you’re proposing, which is, ‘Let’s dig a little bit deeper.’ And maybe you’re presenting these numbers to the executives to make them happy, but at the same time you’re doing some experimentation to see what is the true value that we’re getting. Because at some point, somebody will get to that next level, measure it more directly, and then they will have the actual numbers that make sense, and all of these businesses looking at the wrong numbers, they might start to hurt, their margins will start to fall. And one thing that’s really interesting, too. In the shopping spaces, I don’t know your opinion on this, but small players, small vendors don’t have the efficiencies of big players like jet.com, or Amazon.
ANDREAS: Less data?
FREDERICK: They have less data, but their business is not as efficient, right? So they may have to pay more for shipping because they have to negotiate it additionally.
ANDREAS: Don’t have the same buying power and so on.
FREDERICK: Exactly. And so, at what level is it not about – the key words – at what point is it not about your landing pages or your bids, but at what point is it about, ‘Listen, you really just need to go back to your business and make that more efficient, because that’s where you’re going to pick up more results.’ But, again, that’s a question nobody likes to ask, because it’s a much harder question to answer, whereas changing the bid – boom. It’s done. Figuring out how to negotiate better with the USPS for delivery fees, that’s hard. Right?
ANDREAS: When you take on clients, what are the lowest hanging fruits usually, what’s your starting point?
FREDERICK: So, we do a lot of one click optimizations, a lot of keyword campaigns. So, looking at your query data, often times we see lots of nuggets in there, nuggets of gold. Converting queries that you haven’t added to your account, your keyword list. And so, some of the less sophisticated agencies will say, ‘Well, why should I even do that?’ Because, until you put it in as a keyword, you can’t actually control the bid or be creative, right? So, add stuff to your account like positive keywords, negative keywords. We also see a lot of negative keywords opportunities, so a lot of people still use broad match or phrase match, which will still show up on different queries. And you have these spikes and random traffic around certain events that happen, and it’s completely irrelevant. And you often don’t pick up on them very quickly unless you do a very semantic analysis of the keywords. And we can do that very quickly, so we can do it in 30 seconds, and can tell you, ‘These are the keywords that you probably should add as negatives.’ A/B testing for ads – I think that’s becoming less and less relevant as Google gets better and better at automating it. But that said, I see a lot of accounts where people just literally never remove underperforming ads. So, again, it’s an illusion. They see, ‘Yeah, we have 4 ads, one in every ad group.’ But, yeah, Google has already optimized it, and 98% of your impressions go to this one, and these other 3, Google doesn’t care about, right? So you’re not really testing, you just thought you were testing them. Stuff like that, we see a lot of.
ANDREAS: Great. Thanks a lot for your time, very insightful.
FREDERICK: Thanks for having me. Thanks.
ANDREAS: It was really good, good content.
Interest piqued? Frederick and Andreas will be discussing Andreas’ tests in Google Shopping reverse engineering in an upcoming webinar on July 27, 1:00 PM EST. Sign up today!