Alan Kent

21. Google Spot, Machine Learning, Search Engines, Google Merchant Center [Part 2]

Episode 21

Play episode

Alan Kent is an eCommerce Developer Advocate at Google. He is been with Google for the past 2 years and his previous roles include VP of Architecture at Magento and VP of Search Engine Engineering at eBay.

This is Part 2 of the episode where I sat down with him for a Q&A. Check out our Part 1 to understand more about what Google Merchant Center is all about.

Show Notes

  • [1:04] What is your role at Google as an eCommerce developer advocate?
  • [3:19] Google Spot looks like an interesting concept, why was it launched in India? Can you elaborate?
  • [8:07] How machine learning [ML] and natural language processing [NLP] is helping improve the UX in eCommerce?
  • [14:28] Does Google penalize if we do not use the Google Merchant Center?
  • [16:07] Product feed, how does it work?
  • [17:45] Does Google recommends any extensions for the eCommerce platforms?
  • [20:35] Will Google web crawl along with the feed?
  • [24:44] Best practices to optimize the feed and improve the ranking.
  • [27:20] How has eCommerce search has improved over the past 10 years.

Show Links and References

Shiva 00:02
Hi, you're listening to driven ecommerce at work, a podcast that brings you conversations with the e commerce experts on their processes and the lessons learned. And I'm your host, Shiva Kumaar, head of digital strategy at BC cap. This is actually part two of fireside chat with Alan Kent, ecommerce developer advocate at Google. He was with Magento, commerce and eBay prior to that, if you haven't listened to the part one where he talked about Google Merchant Center, please go ahead and listen to it and just come back over here. Let's get on to the episode. So I'll start with your role, just to give some background to the listeners. And you know, I don't even have any idea about that. So can you tell us a little bit about? So what is your role at Google as an e commerce Developer Advocate?

Alan 01:21
Yes, so it was interesting when I joined Google, so I come from Magento. And they weren't quite sure exactly where to fit me in. So Developer Advocate was a little bit arbitrary in terms of the name. But the general goal of a Developer Advocate is to get Google knowledge, Google experiences Google information out to the community. So that is one aspect of it. And the other aspect of it is to bring that information back into the organization. So by talking to customers by talking to, in this case, merchants, talk to platforms, we bring that information back in to the internal teams. So I spend as much time talking to various internal teams, making sure they understand how the community works, understanding of the role of agencies, the role of SEO. It's interesting that many of the teams within Google often have great knowledge of their particular areas, but don't actually understand how the old whole ecommerce space works. And so they don't always understand the difference between a technology partner, a, an agencies such as yourselves, well, yourself, you do both in some ways. And so part of it is making sure that people under internal inside Google understand the community better. And so can direct what they build, and solving real problems for real people out in the outside world. So it's a little bit of going both ways, within Google both collecting information. So one of the other projects I've got going on at the moment, for example, is we're revising the search documentation for organic search. And there is actually a group that's been set up that anybody can join, who wants to provide feedback on the documentation and the direction it's going. And we're using that then to get external feedback on the documentation before we publish it to make sure we're publishing the most useful documentation for developers.

Shiva 03:19
So I've been using Google pay, I'm not sure if it's, there occurs other countries as well. So I'm talking about India. So I believe Google part was launched around the end of last year, I guess. So it looks like an interesting concept. Can you please elaborate a little bit about you know, about that, and why it was actually launched in India? And how was it successful? How is it?

Alan 03:44
Yeah, I probably can't talk about the successful side of things just because I'm actually not in the Google pay team. And so I've got to be wary of directory, our developer advocates for Google pay specifically. And they may be able to share more data. I can't say it's doing well. Yeah. Okay. That's not particularly useful answer. The part of the reason is historical. So there's actually a really strong Google pay team in India, used to be part of the test platform. And so a number of these different payment platforms have been brought underneath the one umbrella. And so it's also a strong and emerging market. I mean, India is exploding at the moment. And so it's a completely logical place to try some of these new experiences. One, you've got local expertise to it is a massive and growing market. I would say that e commerce market in the US, yes, it's big, but it's pretty stable. There's not a lot of change going on there. And so with a new emerging markets, it's also easy to get adoption with new techniques. And so to me, the Google spot platform is a particularly interesting one. from multiple perspectives. There is the payment aspect to it, but it's also To me, an experiment in the miniapp satisfies the ability to build small custom experiences for people, which you can tie into your own store. So you can have a physical store, have a spot there, scan the code, bring up these in store experiences, it could be data product information, it could be payments information, if you just want to pay directly. So it's, it's an interesting experiment to see how it works. And there are different regulations in payments per country. So it is a matter of you have to take these products per country one by one, make sure you're satisfying all the local regulations. And so it's not like, wow, we've got a working does roll it out globally. No, you have to worry about all the local laws in each country one by one. And that's why Google pays available in search of certain countries and spot is available in less. It's just an incremental rollout, making sure that you cross all the T's and dot all the i's in the process. I think it is a smaller one, I think, you know, it's no surprise, not everything that Google does works, we try things, and it's part of the culture of Google is to explore and trying to push new boundaries. And if it doesn't work well, we don't just hold on to it for the sake of holding on to things. So Google spot is one I particularly like, I think it's got very interesting possibilities. And that crossover between the physical world into the online space. I also liked it, it's easy to do app development, you can do all sorts of interesting kiosk style experiences. And it's really up to you how to use it, and a fun one. I mean, certainly, we're not the only people doing QR code sort of strategy. So it'd be not suggesting that for a moment. And some of the stories I've heard is like, you can walk into a restaurant or a cafe, you scan the code on the table, it brings up a menu, you order your menu from your own phone app, you hit the button, like QR code also included your table number, so the waitress or the waiter brings the food and you are what you ordered directly to your table, you didn't have to order, you don't have to wait for them to come to actually submit your order. You just told it from your phone, you pay from your phone, you and when you're done, you just walk out leave. And so it can make it more efficient. And it was possible because the QR code on that particular table had the table number built in. And so you can actually explore with different sorts of e commerce experiences using these sorts of platforms in new and novel ways that weren't possible before. So it's interesting. That's why I actually enjoy e commerce. It's just because of the variety people have.

Shiva 07:42
Yeah, and especially in countries like India, I mean, with the absence of Apple Pay. So we've been saying that for the past couple of months, I would say, I think merchants are evolving at the same time, you know, offline stores are evolving as well, and trying to go online. So I believe I strongly believe that, you know, Google pay was helpful, and I'm sure that Google's part is going to be helpful for the merchants as well. So can you explain, you know how? So this is more on the technical side? So can you explain a little bit about, you know, how ml and NLP is helping, you know, improve the user experience in e commerce? Particularly, you know, on managing and searching huge catalog of products? You know, I think you got written a blog on the same I guess,

Alan 08:27
I'm sorry, I've missed which one too? Yeah.

Shiva 08:31
Yeah, ml and NLP?

Alan 08:34
Oh, machine learning and natural language processing? Yeah. So, again, it's sort of interesting. And so what I'd written a blog on Google is got all sorts of different technologies. And so one of them, for example, is vision recognition. And if you've got a large catalog of products, it's just interesting, for example, to say things like color. And you might say, color simple, isn't it? You've just got simple colors. Well, product people like creating interesting product names at times. And I one that really stuck in my memory was baby powder, or something that was a color name. So like, what the heck is baby powder, and it turned out to be light blue. And so if you'd use that product information directly from the supplier, and you fed that into your catalog, and somebody searched for blue, it wouldn't find that product. And so where it became interesting was the idea of using vision search against all of the products in your catalog. And it would actually automatically come up with a number of aspects and recognize out of those products. So for example, it would say it could recognize it. Oh, I think this is a party dress. I think the primary color or the main colors are listed colors out and so it would come up the number of product attributes which didn't realize On the original product data. And what you may decide is, well, vision recognition is great, not always perfect. Maybe I don't show it on the site, because I'll just use what the product owner I've got, that's your call. But you can still use it on the behind the scenes on site search so that when somebody searches for blue, it will bring up that product. Showing even though the product description, the field color is actually saying baby baby powder, or whatever it was, it will still match searches based on blue, because you're not conveying any information you don't trust you haven't reviewed. But you're giving, you're bringing up products that you might not have brought up before. And so that these vision recognition systems, it's an interesting way of getting, augmenting your product data. So you're able to provide better experiences to users, machine learning in general, this is so many things you can do with it. Like just understanding what searches add then feed into now, for somebody running a Magento site, this may actually be more interest for the extension developers rather than the merchants directly because it does take effort to set up. Or if you're a larger merchant, you may be worthwhile doing it directly. But you can start doing things like saying, Well, I know that when people searched on this term, they clicked on this item, maybe I should be giving that a slight ranking boost for my on site experience, or personalization experiences. If I know something about the user. And I know when they query for fossils, they're more interested in rocks rather than fossil handbags. Maybe when they come back to my site on my homepage, I will show them more mountain boots, rather than ladies accessories. And so when they come back, you may be showing me your featured product of the day on the homepage. Well, if you know what they're interested in, if you can start categorizing and start learning from what they did on your site previously, oh, they're probably interested in the outdoors. And so you can define your your segments and perform automatic segmentation based on what the visited, then you can direct them at different product placements and recommendation experiences. And these techniques are known to improve conversion rates for people. So using these technologies is getting easier and easier. It's all there are the TensorFlow libraries are all supplied. I was just playing around with the library yesterday, which is not really for a cup match it's it's is actually record usable for e commerce, which is it can recognize your face and put a mesh over your face real time through a web browser via JavaScript. Now, the e commerce application is if you want to try glasses on your face, or you want to try makeup on your face, if if it can recognize what the position of your face, then it can superimpose that imagery over the top. And you can do it with JavaScript in a web browser, which to me is just pretty darn amazing that you can get the performance that good. And so a lot of these experiences like you can create more engaging user experiences to fully understand your products better. And do you have to do all these experiences? No. But there's more and more competition out there. And so if your claim to fame is I've got a list of products, and I've got a catalog, and that's it. You know, products clicking buy? Well, you're competing with a whole lot of other people pretty well, anybody can do that. And so how do you engage your community better? How do you create new experiences where you can offer something that your competitors can't, so that people come back to you, because you've got the challenge of getting people to come to, you've got the people challenge of coming back a second time, your acquisition versus retention challenges. And so the more you can use machine learnings to actually get people to return, your conversion rate goes up. But you need to do something different to your competitors. And so it's always a matter of trying to work out where you're going to invest. And I think the machine learning technologies are getting easier and easier. I won't say trivial. But the more you can learn from the characteristics of your users on your site, the better experience you can provide to them when they come back to your site.

Shiva 14:24
So I'm going to pick a couple of questions from the attendees as well. So here's one question. Does Google penalize if we do not use the Google Merchant Center?

Alan 14:34
It's not a matter of penalize. So it's also a question of what experience you're in. So organic search will not penalize you if you're not in. Like it's not a negative ranking signal. Google Merchant Center is not a ranking signal to my knowledge at all. It is more about if you remember on the homepage, earlier pages, different search experiences and so it would say Here's a review for the side and my found five star rating reviews. Google collects ratings from different sources, including your own site. And you can you feed this all this information info products. And so we can only show that rating information if we find a page and we know it's a product page. And so it's not a matter of penalizing, it's more a matter of you missing out and experiences. If you don't tell Google richer information. If we don't know that this web page is a product page, we're not going to include that product page in experiences. Now, that doesn't mean you're automatically going to get in one if you got that data, because we can only display so many products within it. And there's so ranking still does come into it. There's a number of different factors. But it's not a matter of being penalized. It's more a matter of not being eligible for experiences. We can't bring you up on a map search. If you don't tell us where your inventory is. That's another example. You've got to tell us information for us to be able to use that information in these different sorts of search experiences. Okay.

Shiva 16:04
And here's a question regarding product feed. If I were to send product availability and price data in my product feed, and if I have too structured data also configured on my website, so which data would take preference?

Alan 16:21
Yeah, so there are different rules in terms of that. And so one of the features I talked about is there's actually a setting that you can turn on, so normally will use the feed. But it may flag them to sort of say, because if your website gets updated more rapidly, it's become so common of an issue that there is actually a setting now in Google Merchant Center that you can tell it to take the website and override the feed data on the assumption that if you've got a website, and like a driven by something like Magento, and you go into the admin panel and the update, the price will be instantly on the website reflecting that price. And so you can actually tell there's a setting in Google Merchant Center to say, look, if you see a discrepancy, update my data with the website data because it is more, more fresh, it'll be up to date more rapidly. And so if you don't set that setting will use feed data, but potentially flagged that product is sort of saying, hey, this looks strange, what's going on here? Is everything good. And I can't tell you the exact rules for those sorts of checks. But there is a setting, as I said, to say, the website, I've detected the difference, just use the website data on the assumption that's pressure. I can't remember the name of the setting off the top of my head, sorry, but there is a setting in Google Merchant Center to tell it to automatically update the feed with the website data. Okay,

Shiva 17:44
so this is regarding the e commerce. So since most merchants these days use platforms like Magento, Shopify, bigcommerce, eBay, etc. So does Google recommend any top extensions for these platforms? extensions that may be you know, kind of like Google approved.

Alan 18:03
We're always looking at different extensions. But it's very hard for Google to recommend one different reasons that there's no, if it took one point of view, we don't operate e commerce sites at the end of the day. And so we can look at the technical evaluation, we talked to many of these platforms, if they ask us questions, we're happy to sort of talk to them and give them recommendations and advice on how to do it in the best way possible. But we don't operate a website, we don't know what the support is like. And so we tend to not make recommendations. There may be exceptions from time to time. Just because we don't necessarily have the best input. And so getting feedback from agencies getting feedback from other merchants, I'd actually trust those reviews more than one from Google, we do work with these platforms a lot better. As I said, it's really to make sure they're doing the best job they can and make sure they're aware of the options and how to do the best thing. So we tend not to recommend specific ones. We do recommend ones that we build. So sometimes, Google does build extensions. And so there's a surprise, we recommend the ones we build. And the benefits are they generally kept up to date in sync with Google technologies better, but it doesn't mean they're the best extension and may actually be a better extension for your particular names not built by Google. That's absolutely fine. It really is sort of a mix. We often provide one to make sure there is a good option people can if they don't want to think too much just use our one. But it's not like a no no, no, we're trying to take over the extension market. No, we're just trying to make sure there is a good option for people who just want a default option. But there's so many platforms. We can't afford to And we're not experts in all these platforms. And so often an extension developer will do a better job, or we simply pay an extension developer to go off and build something. So it's a Yeah, but recommendations, that's something we generally steer away from. One, I wouldn't trust them, even if we did end to it, there are other sources that I would trust more such as know, agencies, such as yourself, who actually used them on multiple projects, you'll have a much better idea and be able to provide better advice than Google will be.

Shiva 20:32
Okay. So here's another question regarding the feed. So I had a conversation with merchants who owns an online store, they have this sort of like beard request of generating different data feed than their website, like different title, and then the different description of the product. So how Google actually handles this, you know, I, I'm not sure if you answered this. But let's say Google compare that crawled along with the feed.

Alan 20:59
Yes, we do compare them. Because what we're trying to do is to work out information about the product. Well, we don't want to, we want to present one product. And we want to represent that same product. And we need to if you're in the Google Shopping type experience, we're going to bring up the product information there. If you on the organic search, we want to bring it up there, we don't want sort of like three variants, was a little bit tricky, because it's the sort of like the idea of having variants of products. But then as the concept of saying here is one product and really only got one product, I really only want one product listed on the store. How do I get that result in there? Well, we've got one view of the product that we're trying to build up. And it's frequently based on like the GTIN number or similar some global unique identifier. And we will merge these results. And we'll also at times detect variants between them. And I can't describe the exact algorithms. One, I don't know the exact algorithms and two, I'm not going to tell them for the did. Because part of it is for issues like we want to make sure that people are trying to mislead shoppers. And so my general advice is there's exceptions to all rules. But my general advice is what you're trying to do is to have Google to understand what that product is. That's what the feeds are for. That's what the structured data is for the needs to be a reason why you're trying to present the same product in different ways. And generally, I would lean away from that, I'd be saying trying to get the best experience. But there are specific cases. And I'm not going to say that it'll never work. And you should never do it. But my general guidance is to avoid it. Because we do compare the data because we're trying to make sure people aren't doing malicious things. I'm not saying they're trying to be malicious in this case, but some people do. They try and put up feed one lot of information via the feed. And then they have a completely different quality product at the end of the day. And what they're trying to do is to get such traffic to their product. And so it looks wonderful in Google Shopping tab, and you click on the product, go to the users website, and it's something completely different. And it's a terrible experience for the user, or it's a lower quality product, or the price is quite different. And it's a bad user experience. And so it's the people trying to game the system. They're the ones who are trying to block. But how do you do it? Well, you check the fields, you compare the website, you could check, compare the data. And so if the more different it is, the more it may get flagged and these things to sort of say, hey, something, Something's fishy going on here. And so I'm sort of wary of those sorts of strategies. Google will change algorithms used to detect fraud from time to time internally, no, we don't publicize that information. So there's always the risk, then, that you're gonna get triggering these experiences by accident? Even if it was, there's a good reason for it. Why you want to provide that different information. So my general advice is I'd steer away from it, because I'm worried about you triggering one of these frauds, to check experiences. But feel free to reach out to me. I mean, I'm on Twitter, a Kent 99. Always drop me a message or trying to find me other mechanisms like LinkedIn, and so forth. So it's interesting to understand the needs for it, because that's the sort of information I bring back to the internal teams is, oh, why don't we support this case? And that would be the sort of information I'm quite happy to bring back internally. But my general advice is no, don't try and do that sort of thing. You risk getting your products flagged and so forth.

Shiva 24:41
Perfect. Perfect. So here's one final question from the session. So he talked about Google Shopping tap. Can you help us with some best practices to optimize the feeds I mean, to improve the ranking like two or three best practices you want merchants to learn or implement

Alan 24:59
it So to when you get into the ranking and how your position goes up, that really gets into the SEO field. And so there's nothing to do really with Google Merchant Center. As I said, Google Merchant Center is more about the experiences. Now, having said that, things change all the time internally as people try out new experiences. So I'm not going to guarantee that it's never relevant. But at the moment, I found no indication that makes any difference to your ranking. So what makes a difference to the ranking, the ranking is more about making sure your content is interesting to people doing searches. So if you want it to appear higher, it is the normal SEO strategies, make sure you've got good descriptive words that line up to what people are searching, the richer the information you can provide to Google, the better we're able to understand it, the more we can bring that to people who might not have found that otherwise. If a product has been searched by lots of people think about different terms. So rather than going for the top search terms, think about other different sorts of search terms that people might be using, that your competitors are not. And so, you know, like, should you use pants versus trousers. You might think about different sorts of, like, if I'm appealing to a particular community, and they tend to use different terminology, I'm not hip. So I don't really know what the terminology might be. But if you can work out terms that other people aren't finding, this is where the SEO comes into it is looking through trends of searches to work out what are people searching, there is a little bit different, and how can I tailor for those needs? And so what some people do is they create multiple pages for content, and then tweak the different pages, and then sort of like, there's pros and cons of these sorts of techniques. And there's not a single answer. And that's why and what I can't say like if everybody picked different terms, well, there wouldn't be any main one anymore. So it's there's no single universal way. But it's really the content is what I'd be focusing on making sure you're satisfying some group of users needs so that when they enter a search, they're more likely to find your product. Okay.

Shiva 27:19
So, or towards the end of the session. So here's one personal question that I actually wanted to ask you. So you, you have a long association with the search engines, right? So you worked with Cassini search engine, we wrote the project for eBay. And today, you're in Google, right? So the mother of all search engines. So how has ecommerce search improved? In the past 10 years or 20 years?

Alan 27:43
Has it improved? Absolutely. And I was involved in search before I came to eBay. So back in Australia, I did my PhD in search indexing techniques, which is, and then we built, we did a research project that got spun out into commercial product, it was a whole fun journey, that I didn't just arrive at eBay, and run the search engine team for no reason. my whole career has actually been in search. But if I looked at Google, for example, there was a pure and back in the early days, it was pure algorithmic ranking. How do you try and be purely recall and precision and trying to get your the most accurate products towards the top and search terms was key there. These days, if you look at a Google search, as I said, I just did a query for backpacks. And there's a whole series of different search experiences that come up. And I think a lot of that now is where the more advanced search capabilities are coming in, is trying to understand the user's intent, trying to build up experiences trying to guess what they really after, based on maybe on previous careers as well to sort of say, well, they keep asking about backpacks and pricing information, I think they're more likely to be doing a shopping, can we work out what part of the shopping journey they're during, and bring up information more relevant to that. And so I think a lot of the more recent experiences in if you look at Google search, is there's so much diversity and experiences in trying to work out the intent of the user and what stage they are, and bringing up the most appropriate results based on that. And so more recently, I think that is where a lot of the advances are being made, not purely in the ranking experience, is trying to come up with better experiences and a better UX. And that's under unstructured data to help Google understand the different aspects of your products so that we can use that data to bring the products into those experiences, matters. And I think that's where things are heading more at the moment.

Shiva 29:43
Cool. Perfect. Alan, thank you so much for taking this extra time and clearing our answers, and answering all those question. Thank you so much.

Alan 29:51
Thank you for having me. It's been great.

Shiva 29:57
Thanks for tuning in the topics So we've covered in this episode are listed under the show links in the podcast description. And if you haven't subscribed already, just go ahead and hit that subscribe button so you keep getting notifications every Friday. And if you like this episode, just go ahead and share it with a friend or colleague of yours. might find this useful. That's it for this week. Catch you guys very soon in the next episode. Until then, take care.

powered by

More from this show

Episode 21