EP173: "The HARD Conversation, Stop Avoiding It"

[00:00:12] Welcome to AdExchanger Talks, the podcast devoted to examining the issues and trends

[00:00:17] in advertising and marketing technology that matter most to you.

[00:00:24] This episode is sponsored by Rain Barrel, a new audience graph that helps advertisers

[00:00:29] reach their best fit customers in today's privacy-conscious world.

[00:00:32] Rain Barrel builds both custom audiences and over 12,000 publicly available audiences

[00:00:38] for your next CTV, video, audio or display campaign.

[00:00:54] This is Allison Schiff, you're listening to AdExchanger Talks and this week I've got Omar Tawakol

[00:00:59] as my guest. Omar is an entrepreneur through and through. He co-founded BlueKy and sold it to Oracle in 2014

[00:01:06] and then sold another startup to Cisco a few years after that. Now he's onto his new venture,

[00:01:12] called Rembrandt. It's tech for a virtual product placement but it's also programmatic.

[00:01:18] I'll just let Omar explain it to you, but that'll be in the second half of this episode.

[00:01:23] In the first, Omar and I pour one out for Oracle Advertising, which went EOL,

[00:01:29] that's end of life on September 30th. A quick FYI before we begin, please save the date for next

[00:01:35] years CTV Connect, March 12th and 13th in New York City. AdExchanger is joining forces again

[00:01:42] with synopsis, ad monsters and chief marketer to host a can't miss summit on all the key issues

[00:01:47] and opportunities in connected TV. Learn more and register at cTVconnect.com.

[00:01:55] Hey Omar, welcome to the podcast. Thank you, glad to be here.

[00:01:59] So I'm going to ask you my favorite question that I ask of everybody. What is something about

[00:02:05] you that not a lot of other people already know, that they couldn't easily find out just by

[00:02:10] Googling you or going to your LinkedIn page? The question. First thing that comes to mind

[00:02:18] and some people do know is that I'm a father of four beautiful young ladies and the oldest one

[00:02:25] just got married and that's kind of a milestone for dad and it was awesome. It was just

[00:02:34] exhilarating wedding. So and in case you're wondering how I have a child who's already married,

[00:02:43] I got married when I was three so we're right. Well you look great and mothled of.

[00:02:54] Well I also have a fun fact about you that maybe some people know maybe they don't know but

[00:03:00] you have six patents right in your name. I saw that on your LinkedIn one from September 2021 for

[00:03:08] associating anonymous information with personally identifiable information in a non-identifiable

[00:03:13] manner which feels very edgy and then I scroll back to the beginning so the first one that I saw

[00:03:21] maybe there are others but this is just based on what but you have posted from April 2010,

[00:03:26] first filed in 2005 and taken us way back when you were CMO and GM of audience science and

[00:03:34] this is for accepting bids to advertiser's for users performing a specific activity. It just sounds

[00:03:43] like proto real-time bidding to me. Yeah I guess the file data is different when you know the

[00:03:52] when we started this stuff it was really envisioning a world where people would be bidding on the

[00:03:58] attributes of people and not kind of keywords and pages which was novel back then and now it's like everywhere.

[00:04:07] What do you do with a patent like that since it's still out there but the world has changed so

[00:04:12] greatly I guess other than listing them on your LinkedIn for people like me to find and ask you about

[00:04:17] you know those patents generally like all patents like that are done by individuals but owned

[00:04:23] by the corporations that get them and companies do patents first first and foremost largely defensively

[00:04:30] so that if people come after you you have you know weapons of self-defense so that's that's

[00:04:36] really one thing sometimes build people build them to actually go after other people and you

[00:04:43] protect their area that's the that's another goal. My personal tendency is I don't like the use of

[00:04:50] patents to go after people I think innovation should be more open I love open source and and I think

[00:04:58] that some of the times the people making judgments around patents aren't as much as in the

[00:05:03] know in the industry and so you know odd unintained behaviors happen so I'm I prefer patents

[00:05:09] as a defensive mechanism rather than as a way to go attack people because then you get this whole

[00:05:15] industry of like pure patent trolls and all these like unsavory things. Right when a patent trolling

[00:05:23] becomes a line item on your P&L you should probably look in a mirror. Exactly. So we're going to talk

[00:05:31] about Rembrandt more so in the second half it's your new or kind of newish at this point it's not

[00:05:36] brand new you've been doing it. AI powered product placement startup by the way are you doing any

[00:05:42] patents for Rembrandt? Of course yeah we've done we've done a few. Okay well we'll get into that later

[00:05:48] but I want to do a trip down memory lane for a little bit because just this week something

[00:05:54] big happened. Oracle advertising officially shut down all access to Oracle advertising services was

[00:06:03] on September 30th and this just to remind everybody the announcement happened right before

[00:06:09] can Oracle CEO Safra Katz basically just mentioned it in passing on an earnings call

[00:06:15] and a lot of people in Oracle's advertising division were really surprised by that news they apparently

[00:06:22] weren't given any warning really some people didn't have any warning. I mean you you spent two and a

[00:06:27] half years as SVP and GM of Oracle data cloud you joined back in the long ago year of 2014 after

[00:06:36] selling blue Kai to Oracle you founded blue Kai massive consumer intent a marketplace and

[00:06:43] DMP but did the shutdown surprise you or did it feel inevitable? I know you've been personally

[00:06:51] out of it for a while but I'm sure you still haven't interest. Well there was a there was a legal

[00:06:57] motivator that they were very silent about. There was something going on that was quite large

[00:07:02] for the industry in terms of you know what they settled on that they couldn't talk about.

[00:07:09] So I think that was part of the silence and I really didn't appreciate how impacted people but I'll

[00:07:18] tell you why I was silent about it. Number one they acquired my company and I benefited financially

[00:07:25] and when I was there I was the steward of trying to keep that that group safe but eventually I loved

[00:07:31] and so I can't turn back and point fingers at them after I benefited financially and after I

[00:07:39] selected to leave myself. That would seem disingenuous and kind of like not respectful to them

[00:07:45] to be like I took your money and you made it really nice for me to stay and I left and now that I'm

[00:07:49] gone I'm going to critique you. I just didn't feel that that was fair game even though I kind of wanted

[00:07:53] to say things that there were much more empathetic to the people who were impacted. I just thought

[00:07:59] it would have been hypocritical so I just kept my mouth shut and I have to say when I was there

[00:08:05] they both treated me well and did a lot of good things for that group and that group was brilliant.

[00:08:13] I mean there's so many great people there from not just blue skies so I'm not tuning my own horn

[00:08:17] but you know there are multiple acquired companies that were just like leaders in what they did

[00:08:21] you know, vote and data logics and there was just a whole bunch of companies acquired there

[00:08:27] with great people so I very fond memories and they made that possible and they treated us well

[00:08:32] while we were there and some things happened afterwards and I don't control and for for not to

[00:08:38] criticize even though it's kind of easy to criticize seeing how I went down. I just thought I'd keep

[00:08:43] my mouth shut. I do appreciate that but this is my podcast and I'm going to ask a few more

[00:08:51] go about it but I won't make you a criticize anybody. It's more like analysis.

[00:08:59] There has been reporting over the years that Oracle's been depryoratizing its ad tech assets

[00:09:06] in large part because of privacy related concern. Yes. Kind of felt like their heart just

[00:09:12] wasn't in it. I mean there was the shutdown of ad this and just general scrutiny certain lawsuits

[00:09:17] so is that primarily what you were talking about? Yeah well it was the lawsuits are and I'll tell

[00:09:23] you the biggest strategic mistake I did was that when we were acquired we required primarily

[00:09:30] for the DMP with some desire to take our data business and turn it into a bigger business

[00:09:37] and just before the day before the acquisition closed, I had a meeting with a bunch of the

[00:09:45] executive team there and I was going to go in and propose that they instead of becoming a data

[00:09:52] marketplace, that they shift to becoming the platform that enables data sharing between these entities

[00:09:59] but doesn't take a third party position. Doesn't become you know a data control or rather

[00:10:05] becomes a data processor because that would suit Oracle's long-term privacy stance more than

[00:10:11] becoming a third party data provider. That was my plan and I came in there and everybody took me

[00:10:17] aside and said don't do that, that's not what they want from you, they want you to go build the

[00:10:21] data business. I think that was my biggest strategic error. I should have fought that and said

[00:10:29] trust me, I know this industry the long-long run of this is going to it's going to test your

[00:10:36] metal on whether you really want to do this. If you just shift to becoming the technology

[00:10:41] for this rather than the third party owner, you may have a radically different stance in this

[00:10:46] industry and you'll stay in it longer. I didn't do that because they were adamant about hey

[00:10:51] we're really excited about this new data business go grow it but looking back, I wish I had done that.

[00:10:58] I mean it would have been a very different outcome when Facebook pulled the plug on

[00:11:02] it's third party data seller marketplace including a data logic's data segment so it would have been

[00:11:08] not really the blow that it was if you'd taken a different stance. Exactly exactly I think that's

[00:11:16] for sure and I because of this whole experience have developed even

[00:11:22] new thoughts about what the world would look like if we took some radically private sea

[00:11:28] saving technology approaches and maybe in a future podcast we can talk about stuff like that.

[00:11:36] Because I do think there are ways for the industry to evolve that just leapfrog where we

[00:11:41] are now we're constantly running from some new legislation and defensively reacting to it

[00:11:48] rather than using our considerable minds as an industry to figure out ways that consumers would

[00:11:54] actually be okay with. Right we're in this pose now where it's about defending the status quo

[00:12:00] or trying to twist ourselves into a pretzel to maintain what's already been built and to protect

[00:12:07] that when it could have been better for everybody really just in retrospect though it's

[00:12:13] so easy to say this now I guess to take a different approach and to think more about not what's

[00:12:20] cool to do or what is possible to do but what is good for the long term. Yes sustainable and that's

[00:12:28] you the theme you're going to hear when we talk about about Rembrandt it's just thinking through

[00:12:33] long term sustainability. Well we're going to talk about Rembrandt soon but I'm going to force you

[00:12:38] to talk about Oracle for the life of longer because also putting aside this you know

[00:12:46] can feature knowledge and if you had a time machine you would go back and maybe tap

[00:12:49] yourself on the shoulder and other people too and tell them what you knew and the industry would

[00:12:54] be different back in 2014 when you sold Blue Kai to Oracle an Oracle was in the process of spending

[00:13:01] billions of dollars on acquisitions to create what became Oracle advertising there were so many

[00:13:07] pieces I'm just going to rattle them off because to think of them into a telly is to really

[00:13:12] this was a vision didn't pan out but it was a vision there's L.O.C. for around 810 million in 2012

[00:13:20] responses for 1.5 billion the following year then Blue Kai in 2014 for I think it was reported somewhere

[00:13:28] between 350 and 400 million data logics the same year 1.2 billion maximizer in 2015 no deal price

[00:13:36] reported add this for 200 million crosswise for 50 million in 2016 mode in 2017 for 850 million

[00:13:45] and then grape shot same year for 400 million dollars and then by this year Oracle's advertising

[00:13:52] revenue fell to somewhere around 300 million dollars which is less than it paid for many of these

[00:13:57] acquisitions but when you were there between 2014 and 2016 like did it feel exciting, did it feel

[00:14:04] like something big was happening? Oh yeah I mean remember when they acquired us it was originally

[00:14:10] to plug the data management platform in to the Oracle marketing cloud so it was to bring data to

[00:14:17] the marketing stack it wasn't to build an antex stack that was the original part but in the process

[00:14:26] of doing that they said well looks like you have this data piece that also feeds advertising why don't

[00:14:31] separate that out we call the Oracle data cloud and the vision that we had back then was essentially

[00:14:37] to build the identity graph and the thinking was that the identity graph was going to be core

[00:14:43] to having a differentiated marketing cloud then the other marketing cloud players because if you could

[00:14:51] really understand consumers journeys across all these different touch points you could do far more

[00:14:57] effective marketing that was the glue of the original acquisitions while I was there and so if you

[00:15:04] you know if you think at data logics data logics provided you know more of the offline understanding

[00:15:10] consumers while blue kai was kind of online understanding of consumers add this brought an international

[00:15:18] component to the understanding of how people behave crosswise you know brought you know's

[00:15:25] understanding of identity in the mobile domain and how that might relate to other domains so

[00:15:34] those are all those early acquisitions were rounding out a very powerful identity graph and if you

[00:15:41] think about even even eloqua had a component of that because you had you know email as an input

[00:15:49] and so we were building at the time what was going to be the world's best identity graph

[00:15:55] and that was kind of the driving vision when I was there and then it altered of course

[00:16:00] what happens at a leader leaves different people have different visions and and it kind of took

[00:16:04] a different direction one thing that surprised me but maybe it wouldn't be surprising for someone

[00:16:11] who understands business better is that Oracle decided to just shut it all down rather than selling

[00:16:17] at least some of these assets because their assets with with they there's real name recognition

[00:16:22] there like blue kai for example but also grape shot and mo it's stuff that you would think other

[00:16:29] companies would be interested in oh for sure yeah I mean media mouth found a buyer

[00:16:35] but I I guess Oracle didn't want these pieces going to rivals perhaps or maybe they didn't

[00:16:40] like it was worth the effort I mean they do there but 465 billionish market cap company

[00:16:48] it may it may have more to do with the some of the aspects of the legal settlement

[00:16:53] I would point to that and not say much further because they better represent what they were thinking

[00:17:00] but I suspect had more to do without than anything for you personally though are you

[00:17:06] bummed that blue kai won't live on in some form I mean you you have actual children one who

[00:17:11] got married recently so congratulations again but a start-up which blue kai was you know is like

[00:17:18] a baby of course it became huge huge baby but you nurture it you know you're proud of it when

[00:17:24] it finds a home when it gets married got married to Oracle so I'm assuming there's you know

[00:17:30] it feels like there's a loss there at herds when it's no longer in the world yeah that's true

[00:17:35] I think the bigger thing that hurts is like the fact that there were people who are still there who

[00:17:40] this actually hurt the meaning you know meaning they had to look for a job and they're great

[00:17:45] people though many have found or well quickly find but that that was the one that more hurts

[00:17:51] because it's kind of like the practical thing the yes and of course it would be nicer it had gone on

[00:17:58] you kind of see this interesting legacy though which is that dmp's morphed into like cdp's

[00:18:04] so you know there were aspects of that that you know that Oracle still has which is you

[00:18:13] know technology that isn't a data marketplace that that took different forms so I'll say that but

[00:18:20] then the whole industry started growing for cdp's and now they've grown into you know even different

[00:18:29] gerions of a possible cdp's and that kind of thing that's an in clean rooms which is

[00:18:36] clean really just yeah it really just another angle and so in that sense you kind of feel part

[00:18:42] of how this this continued and you look at the people a bunch of these people started new companies

[00:18:50] and some of those new companies have already either succeeded or been acquired or grown and

[00:18:56] and you kind of feel part of all that and it's nice so yeah there was a downside here

[00:19:05] that's impacted the people and the loss of the brand name being out there but it's really on how

[00:19:12] the industry and the people evolve that that I tend to focus on that makes me happier

[00:19:17] and I know that double verify and I S have hired a bunch of people a former remote people

[00:19:23] in particular so that's been nice to hear on their earnings calls yeah that's right and you know

[00:19:30] over time a bunch of blue cires made it all sorts of places you know into a goal into double

[00:19:37] verify and to live ramp into you know T mobile advertising into all these different places

[00:19:45] so one last Oracle E thing and thank you for indulging me it's it's a doorky question

[00:19:52] guess the year of the first time that your name appeared in an ad exchange or article

[00:19:59] oh god I guess it would be in the very beginning because we at blue kai started

[00:20:09] talking to an exchange or really early on because you guys were so

[00:20:13] in so in the inside of our industry really understood everything it just stood out to us from day one

[00:20:20] well I'll give you the answer if you have to go guess no no tell me

[00:20:25] it was September 13 2008 16 almost exactly 16 years ago and the headline was blue kai

[00:20:33] enters with new advertising exchange and it's basically a summary of a scene at story about the launch

[00:20:40] and it mentions you and reading that story it feels like another lifetime this industry was just

[00:20:48] so different then that's right yep I was yeah 2008 is it absolutely that was fun time

[00:20:56] could you have envisioned though so September 13 2008 roughly six years later that you'd sell

[00:21:02] to Oracle like was that something that was that felt like a possibility definitely not Oracle

[00:21:07] yeah because Oracle was not really part of the ecosystem in the sense and it was surprising actually

[00:21:11] and even like the first time they approached us to discuss an acquisition I was surprising I just

[00:21:19] didn't have Oracle in mind as someone until they started articulating their needs for the marketing

[00:21:24] stack and for some weird reason I never could explain it we had just hit profitability

[00:21:31] and within two weeks three public companies came after us till kawaras it's just like a

[00:21:40] thing of beauty that was unplanned because obviously if you go through an acquisition you want

[00:21:45] multiple multiple people interested so yeah that was a surprising outcome

[00:21:51] well now we're going to fast forward to October 2023 which is the public launch of Rembrandt

[00:21:59] but first we're going to take a quick break so stick with us.

[00:22:15] Hello I'm Sarah Slues in the second of editor of ad exchange her and I have with me today

[00:22:20] Travis Reedelhooper executive vice president of Remboro welcome Travis.

[00:22:25] Thank you for having me Sarah. So Travis what differentiates Remboro from other audience providers?

[00:22:34] We founded Remboro on three principles transparency privacy and reach

[00:22:39] allow me to go through them quickly first transparency we found transparency to be rare when looking at

[00:22:46] other third party audience providers you can see the CPM and the size but that was about all

[00:22:53] digital advertising shouldn't be a black box we built an app with the directory of all of our

[00:22:59] audiences so users can see other important metrics like how the audience size has changed over time

[00:23:05] and how fresh the data is anyone can send up for our app for free and start learning more about

[00:23:11] the audiences they're using. Second privacy we built our audience graph using 100% consent to data

[00:23:18] we also have a robust opt-out mechanism to update our audiences as users with drop consent

[00:23:24] from an app or specific identifier we believe this attention to privacy and consent puts us in a

[00:23:30] great position for longevity in an ever evolving regulatory landscape. Lastly reach we know how important

[00:23:39] audience scale is for advertisers this is why we have adopted alternative identifiers like

[00:23:46] UID 2.0 and ramp ID in order for our audiences to reach as many people as possible on as many

[00:23:53] different platforms as possible. So much attention has been given to first party data are third

[00:24:01] audiences still important. We agree for as party data is a vital part of any advertising strategy

[00:24:07] however we know that it isn't as helpful when it comes to generating awareness and consideration

[00:24:13] amongst new potential customers. Third party data in our audiences at rainbow are a great way

[00:24:19] to reach qualified new customers one of our most common use cases are for conquressing campaigns

[00:24:25] that is where a brand is able to reach consumers that shop at a competitor's store.

[00:24:33] And now that Chrome isn't planning to ditch third party cookies does the fact that rainbow

[00:24:37] barrels audiences are cookie less still batter? From day one our audiences have never been built using cookies.

[00:24:45] The biggest reason for that is accuracy. We have put an unbelievable amount of time and effort

[00:24:51] developing our methodologies in order to provide our clients with accurate audiences.

[00:24:56] In the third party audience marketplace there is a lot of incentive for providers to loosen

[00:25:01] their methodology in order to achieve more scale and look better than their competitors.

[00:25:07] Rainbow Rale is different. Rainbow Rale is dedicated to putting accuracy first in order to ensure

[00:25:12] our clients can maximize performance. Well thank you Travis and thank you to Rainbow Rale for

[00:25:19] supporting our podcast. Okay we're back and yeah I mean congratulations by the way because just last

[00:25:43] week Rembrandt, one best in show at the Attics Change Er awards and you also to come at trophy for

[00:25:51] best AI innovator. I think that's a new category for us. I think that's the first time we had it

[00:25:55] and these are some big hefty gold trophies. They are inspired up purposely by the look and feel of

[00:26:02] an Academy Award and I do hope you'll display your hardware somewhere other than the bathroom

[00:26:09] because I say that because there's this I Graham Norton clip from a few years ago where Kate

[00:26:13] winslet says she keeps her Oscar in the bathroom because she knows people like to hold it in front

[00:26:19] of the mirror and pretend they're giving a thank you speech but they're like embarrassed so they need

[00:26:23] to close the door. But yeah so congratulations. Thank you by the way that was really quite

[00:26:31] exciting and I don't know which one we were more excited about because the AI category is exactly

[00:26:37] what we want. We want that this category to grow and be recognized and understood and so we're honored

[00:26:41] about that and then and then best in show was was like awesome so yes we will display them probably.

[00:26:48] I wish that you'd been able to say your speech in front of the crowd but well pretend like this was

[00:26:54] that and yes what is Rembrandt? I'm a programmatic product placement platform. I really like that

[00:27:00] but I want to go a layered deeper. Like what is it exactly for our listeners that might not be familiar

[00:27:07] with it yet? At the end of the day we're trying to build sustainable advertising because every

[00:27:12] ad exec I talk to when they go home they spend money on things like Amazon Prime and Netflix and

[00:27:19] when they get a chance to skip ads they skip home they basically avoid ads they spend their

[00:27:25] allocating funds to drive inter up to vads in other people's experiences but when they can they

[00:27:31] do everything they can to avoid seeing the ads and they're not alone consumers are doing that too.

[00:27:36] So if we want a long-term sustainable model the current model is like let's add more ads into the

[00:27:43] ad load so that if you're forced to not have an experience with ads in it you're suffering through

[00:27:50] several minutes more than you want to and as audience you know diminishes from linear TV there's

[00:27:58] more pressure for you to add more ads and it's just a not sustainable way to build this and so our

[00:28:06] whole model here is just to say hey look we're consumers too and we run away from ads why don't we

[00:28:12] build a future that is more sustainable where instead of interrupting the experience people really love

[00:28:18] what you should do is put the brand into the experience they they like non-interruptively

[00:28:25] and that is much more sustainable and by the way it gets noticed by the consumers it's

[00:28:29] appreciated by the consumers get attention for the brand it gets money for the content producer

[00:28:33] it's just not been technically doable until now so we think this is going to be a much more

[00:28:41] sustainable long-term future for advertising. And it is the intersection of a few different trends

[00:28:47] that you touched on. I'm people getting fed up with interruptive advertising which has been

[00:28:52] true for a long time but the rise of creator content especially on YouTube and AI so it feels like

[00:28:59] the right moment I guess. Absolutely and this is the first moment we could do it because

[00:29:05] if you look at product placement how is it evolved in the past if you're doing a movie two years

[00:29:09] before the script before the movie goes live you negotiate to get your product in the script

[00:29:14] or if you're doing a creator video this like this four to six weeks of back and fourth product is

[00:29:19] shipped you negotiate the script you do the shoot you might have to do a re-shoot you have to sign

[00:29:25] contracts all this can be done digitally without shipping product it could be done post production

[00:29:32] it could be done very quickly that's what changed you took a process that was you still

[00:29:40] big industry multi-billions of dollars but incredibly slow and only applied to a certain

[00:29:45] percentage of videos and by bringing AI to the table you can add a virtual product that looks

[00:29:50] like it was filmed in the original you can do it at the speed of compute you can give optionality

[00:29:54] to the person doing it you can eventually target it so the same you know item in a catalog and

[00:30:00] a video catalog could be you know showing you know a Pepsi in New York and a Perry A in Paris

[00:30:06] like like the flexibility here is amazing and AI has made that possible well make it a little

[00:30:13] more real for me say I'm on the mid-sized creator and I have an established audience maybe not

[00:30:19] massive but it is the kind of audience a brand would want to tap into and there's a shampoo brand

[00:30:25] or a soda brand that wants to work with me like what happens next where does Rembrandt come in

[00:30:31] that scenario yeah I mean so for example we we worked with this particular creator called

[00:30:37] Alexander but as she's like a chest stream or well known you know wonderful person and she

[00:30:43] has a following and we actually had a shampoo brand that that came in and basically looked at

[00:30:52] you know the thousands of creators in our network we recommended a subset that matched the audience

[00:30:57] requirements of you know they were looking at you know females are looking to certain age target

[00:31:03] and geographic target and Alexander boat has matched that and they you know they basically accepted

[00:31:10] that as one of the target creators within this much bigger list and so Alexander comes into our platform

[00:31:18] sees eligibility and basically says great let's put in the the virtual shampoo in my background

[00:31:24] and has done that many times across many products after they you know in the middle of their

[00:31:31] editing or near the very end they can log into our platform AI studio and they literally take

[00:31:36] the bottle of shampoo off the virtual shelf and they find a place in the video they I tell some

[00:31:42] eligible parts of the video and they actually place it themselves and then they publish it to

[00:31:47] YouTube and we collect the number of impressions that were shown for that video and that's how they

[00:31:52] see the emphasis on how do you make sure let's talk a little bit about the technology

[00:31:59] how do you make sure the thing so the shampoo bottle or the soda can or the bag of chips or

[00:32:04] whatever the item is that it looks good in the video and not weirdly tacked on like it should

[00:32:10] actually be there that makes sense that it's there in the video because you have to account for

[00:32:15] angle and light and movement and the reflectivity of surface it is very complicated.

[00:32:21] It's very complicated so let's talk about like that this is such the heart of everything

[00:32:26] let's spend a little time here. So first of all if you just understand how the I see

[00:32:32] it's it's it's very important because that's what you're going to mimic.

[00:32:36] What happens is two two dimensional images on printed in the back of your retina

[00:32:41] so your eyes are actually seeing 2D but it's your brain that sees three dimensions and it's

[00:32:50] dimensional images and look into them and understand you know an ocean of depth what's the

[00:32:57] scene look like you know where are the planes what's the lighting of the scene.

[00:33:03] The motion in the scene the the reflectivity in the scene the brain is able to capture and

[00:33:09] understand all of these things and synthesize an image it's even interesting because this happens

[00:33:15] like discreetly people guess somewhere between 25 and 50 frames per second it's not continuous

[00:33:23] so your brain gives you the illusion of continuity in terms of the way it process this amazing

[00:33:29] with the brain does. So what we have to do technically when you see a video which is essentially

[00:33:35] a two-dimensional asset is essentially try to reconstruct the three-dimensional reality represented

[00:33:41] in that two dimensional image. This is not like generative that that creates videos at the pixel up

[00:33:47] and comes up with an image and then you know string together some images this is

[00:33:52] understanding three-dimensional reality in such a way that if you have to alter it

[00:33:56] you have a map of the lighting at every pixel point in that image so what's the lighting look like

[00:34:03] because that's going to tell you how much lighting is going to apply to this new object you're

[00:34:07] you have to understand segmentation so you can handle occlusion so there's someone moves or hands it

[00:34:11] doesn't go straight through the shampoo bottle. You have to understand the material reflectivity of

[00:34:17] the surfaces so you can put a shadow that that reflects carefully and that's easy if the camera

[00:34:27] is fixed if the camera is moving like you get in professional content it's way harder

[00:34:32] and so this is the kind of stuff that you do with the AI to make it look like the bottle was

[00:34:38] there and you can't actually tell that it was virtual and what we find is that we animate our

[00:34:45] products on purpose because we want to get more attention and that's when you see the comments

[00:34:52] because people were like whoa I didn't realize that was virtual and so that's kind of how we guide

[00:34:58] how we know of doing a good job or a bad job to judge you know as the animation sensible how

[00:35:04] are they reacting to it what are they saying and the consumer comments are like they're really cool

[00:35:10] like I can share some of you who want but I think that's an important guide that we have

[00:35:16] from from all this because if you have consumer comments you have this asset you don't usually

[00:35:23] have on ads so for example one of the comments says I think it's noticeable but subtle enough not

[00:35:29] to be annoying I like this type of ad as it is less disruptive to content I thought it was very

[00:35:34] clever product placement it's very unique I like how the brand was seen and not heard but great

[00:35:39] brands don't have to say their name it was a new and way cool of implementing the ad I think

[00:35:45] this is smart way for all removed the name of the brand to say them to get the word out about

[00:35:49] products you know it was on and on or you know I'm really craving you know and if it was a drink

[00:35:55] the name of the drink and so on and so forth those are kind of comments consumer say that guide

[00:36:01] us to did we do to a good job is it right you know there's one case where I think we overanimated

[00:36:08] the product where it was a big product I didn't need to be moving as much as it did and that's what

[00:36:13] you expect when you get these new mediums you have to learn to get it right well that's I was actually

[00:36:19] gonna ask about that because you can make objects fly in and out of frame maybe fly in from

[00:36:25] the side and then settle on a table or something which you're purposely doing to try and

[00:36:28] call attention to it but like do people want bags of flame and hot cheetos or whatever flying

[00:36:34] into frame right again I think my opinion on this really doesn't matter what matters is looking

[00:36:42] at the data from the consumers and so what we did is we actually put in our platform below

[00:36:48] linked for every advertiser all the data I'll what customers are saying

[00:36:53] because we think that if you have that kind of very clear transparency they're going to basically say oh

[00:37:00] so in general what will do is we'll do multiple creative units that go in we'll do one that's

[00:37:04] unanimated we'll do the animation we'll do you know poster with a 3D pop out we'll do a 3D object

[00:37:11] and as the campaign goes if we see comments that tell us you know one unit is fair and the other

[00:37:17] will stop doing that unit so really date a guide this my opinion doesn't matter now I'll tell you my

[00:37:23] opinion is that we're gonna make these creatives and some of them are going to be

[00:37:29] to over the top and we will learn very quickly where is the line that gets attention but isn't

[00:37:35] annoying and that's exactly what we're doing with the data from consumers so luckily you know

[00:37:41] me being the CEO of a opinion it really doesn't matter its customers opinion the matters

[00:37:45] fair enough uh to the creators that you've been working with so far today stop using pre roll and

[00:37:52] mid roll all together I mean those are you know far more interactive um I'd rather see flying

[00:37:59] cheetos I think than then pre roll but is that something that you're seeing your creators do or

[00:38:06] did they do both still because you monetize where you can yeah so what we see with this is that largely

[00:38:13] our creators have not yet start turning it off but we're starting to be a monetization

[00:38:19] engine that matters so for instance one of our you know mid-sized creators made $200,000 off

[00:38:25] of another one um posted a single video made $49,000 off the single video just starting to matter

[00:38:30] them so they could get to the zone where they would do less um of other forms so that's a good

[00:38:37] now here's the surprising thing that we discovered we discovered that you know most ads on on YouTube

[00:38:44] let's say 30 second ad on average after 10 seconds 50% population is not there and so um we found

[00:38:51] this really curious result is that if you show an interactive ad to an audience member who's seen

[00:38:57] our you know virtual product placement sit in that video for several minutes when they see the ad

[00:39:05] after that they skip it less they skip it about 50% less meaning they're they're mentally

[00:39:12] primed to be willing to listen about that product because they saw that product in an environment

[00:39:17] from a trusted creator so it's just a it's just a very surprising result and I think that makes

[00:39:24] us more friendly to the platforms because we make their ads perform better that's interesting I hadn't

[00:39:31] thought of these things working in tandem it does make sense um i'm going to stick with my flying

[00:39:36] dos imagery because I think it's funny how do you measure the impact of something like this how do

[00:39:41] you measure the value of flying cheetos in the background of a video podcast for four ways um and

[00:39:51] we're all different the first one is look at the comments below it how are people reacting to it

[00:39:55] and how many people are liking those reactions so it's just a very rich form of feedback that's

[00:40:02] pretty instant meaning it's in the comments below the videos and the videos get most of the

[00:40:06] view in the first few days so so that's number one number two is a lot of our brands who can't

[00:40:10] throw studies and what we've seen with the can't throw studies is a very large lift and unated brand

[00:40:15] awareness and lift in other areas like you know favorability and persistent and so on so can't

[00:40:21] throw studies are like I'd say the most common third party a form of data um the data below

[00:40:27] they add from the consumers is available every video can't throw studies you do at the campaign level

[00:40:32] and then the third thing is actually measuring the skipperates of ads from the advertiser

[00:40:37] to the audience that saw that's another kind of signal that we found is useful

[00:40:42] and um the fourth one which is much more rare and something I'd like to see a lot more of

[00:40:47] is the ability to do sales lift studies to see okay like are we actually moving widgets in supermarkets

[00:40:53] or in other places? Do you partner with DSPs? How do you make it as accessible as possible

[00:40:59] to do a virtual product placement by which is still new to people I think? We have not yet

[00:41:06] partnered with DSPs that isn't what we want to do for next year because what we want to

[00:41:10] move into next year is something that's really, really cool which is um CTV with programmatic

[00:41:16] targetability so the the the the biggest innovation that was missing that would make CTV possible

[00:41:23] is the ability to do this photo realistically without VFX in a moving camera situation and that's

[00:41:29] what we built. We now are able to allow AI to place that can and as the cameras like jiggling and

[00:41:36] shifting the can is still sticking to the table and not sinking into it or floating above it

[00:41:43] and now we can do that so now that you can handle professional content without having all

[00:41:48] cost to go to VFX who takes too long to actually do this stuff we could do it in the speedy

[00:41:52] compute. Now that we have that one of the beauties of CTV that you don't have with YouTube and

[00:41:59] and TikTok is that you can use server side add insertion um that's technology that exists today

[00:42:05] in other words when you decide which 30 second slot to input inside of a piece of content it's

[00:42:11] the server side add insertion that's making that possible that same plumbing makes it possible for

[00:42:16] us to not just wait for the interruption to decide on what peach content show but before the

[00:42:21] interruption in the content itself you can show a piece of content with a Pepsi and it or a piece

[00:42:26] of content in in New York where another piece of content is viewed in Paris and you show the

[00:42:30] periay so um that is the realm of where programmatic and DSP's come in and and that's where we're

[00:42:38] putting some effort now and it's all really cool stuff but I want to zoom out a bit like as we near

[00:42:47] the end of our episode here and talk about what you might think of as the other side of AI because

[00:42:52] Rembrandt is helping creators monetize their videos and like that's that's awesome and then on

[00:42:57] the flip side this is also the kind of technology that could be used to mess with people's

[00:43:03] perception of reality in a subtle way like there's always a concern about flip sides and

[00:43:08] particularly during a presidential election year so it's like pretty easy to generate fake content

[00:43:15] it doesn't even have to be over the top to be convincing something virtually placed in the background

[00:43:20] could signal an affiliation for example a support for a cause or something very negative so

[00:43:26] how are you thinking about that or are you worried about that and how do you protect against it

[00:43:31] just the misuse of your own technology? This is huge what you're talking about is huge and by the way

[00:43:37] the next couple months it's gonna get terrible because if you think about the last watching

[00:43:41] what we did is we had the ability to take a piece of content that was huge and generated that was fake

[00:43:47] and use targeting on the social platforms to get it in front of people and the social platforms

[00:43:53] just you know love outrage so they would take this fake thing that generated outlayage and got

[00:43:58] a lot of attention to us I was last said election cycle oh what's much worse in this election cycle

[00:44:03] is now we can programmatically design a different piece of outrage to tailor the the the

[00:44:11] the sensibilities of each subgroup and and we can scalable create that content and we could make it

[00:44:20] in the form of audio and and video so whatever problem we had four years ago it's 10x now

[00:44:26] in terms of its danger so this is near and dear and my heart and hilarious piece of the history

[00:44:32] of how Rembrandt got created you know laugh at this I originally started building tech that was

[00:44:39] it gonna enable us to know the difference between something real and something fake my whole thesis

[00:44:45] was that from now on you were gonna you know no stream was going to be a pure stream it was always

[00:44:51] going to be augmented and altered by a high so let's figure out technology that let's say there's

[00:44:56] you know benign alterations like for example you know I do a podcast in English and I want to use

[00:45:02] a i to translate that and all the other languages in my voice and i want that as a content creator

[00:45:09] or you know the case of what Rembrandt does today is you know I want to put a product in

[00:45:14] in there and I'm allowing it and I'm actually the one actually picking up the the bottle off the shelf

[00:45:19] and putting it in myself and so these are like you know benvelin or benign uses of AI but

[00:45:24] they're um how do you know the real ones from the fake ones so i started you know looking at building

[00:45:30] the technology and actually did build um technology that would allow you to do this and then

[00:45:36] it turns out you know there's another organization called the C2PA which is the coalition for content

[00:45:42] providence and authenticity they're really dive into kind of the kind of standards that should

[00:45:48] be open source and adopted by by leaders and it is by the way it's adopted by a lot of folks

[00:45:54] then it's going to start to let us tame this world and be able to have something equivalent to a

[00:46:00] watermark or a nutrition label where you can look at any video and understand the alterations

[00:46:06] and what was the chain of alteration and was it done by the original content owner not so um that's

[00:46:13] originally the technology started looking out but I came to the conclusion I was building to

[00:46:16] all gee that this was better open source and not for for profit and it was better part of a coalition

[00:46:21] and I started looking for okay let's look at the the benign uses of augmentation of videos

[00:46:27] that's really going to be a big industry and um really settled on you know virtual product

[00:46:33] placement so it was part of my journey yeah it's an interesting origin story and it ties right

[00:46:39] into your feelings about open source and patents yeah it does and how do we make this whole

[00:46:45] thing sustainable from from the beginning so I'm going to bring us home with an anecdote and a

[00:46:53] question it's kind of a long anecdote but uh we we talked a lot about blue kai during the first half

[00:46:59] we didn't talk about voiciel um that's pronouncing it correctly right voiciel yeah okay

[00:47:05] wasn't sure because I've actually never heard it set out loud but that's another company that you

[00:47:09] founded later in 2017 sold it to Cisco in 2019 an enterprise voice assistant so you are a serial entrepreneur

[00:47:17] pretty good track record for selling companies late last year it was November I was with Ryan Joe

[00:47:24] this is back when he was at business insider and now he's editor and chief of ad leak

[00:47:29] I hope he remembers the little people and he and I were at a meta drinks get together thing

[00:47:34] I think it was a holiday party I actually don't remember and I also don't remember how Rembrandt came

[00:47:39] up but it did for some reason and Ryan said something like I wonder how long it'll take Omar to sell

[00:47:46] and you and I had only met once before I didn't really know you very well but I did have your email

[00:47:51] address from that meeting so Ryan and I decided to write you an email with the subject line our

[00:47:57] prediction and this is what we wrote like hi Omar this is Allison and Ryan from ad exchange

[00:48:03] your and business insider today we made a prediction that Rembrandt will be acquired by either Amazon

[00:48:08] or TikTok within two years we're sending this email to plant the flag in case it actually happens

[00:48:14] we're not drunk just in a goofy mood speaking of drinks though if a deal does transpire as

[00:48:20] predicted we'd like to take you out for a drink but you're paying all the best and then you go back

[00:48:27] you're like hi Allison and Ryan love this email made me laugh I wasn't sure how to respond

[00:48:32] I'm having too much fun to think about exits yet I'm glad to know that others also see the value

[00:48:38] so it's been many months since that exchange yeah I mean does does our prediction have a chance

[00:48:46] of coming true we have a better year you know what I'll tell you the interesting about these

[00:48:51] predictions is they're usually under estimate the long-term impact of it but they usually get the

[00:49:04] so what I mean to say is that I don't know if two years is the right prediction

[00:49:12] I'm certainly not building it to flip it because I'm just having fun and sort of the people in the

[00:49:16] company but I do think that this will be strategic to a lot of different players as as time rolls

[00:49:23] forward and we could really prove that there's a significant volume that we'd be done on this the

[00:49:28] other thing that I would say is that we're trying to append the way the ecosystem works so like if

[00:49:33] you look at the ecosystem works today as you create content a separate platform is where you

[00:49:37] go distribute it and then these this other stack called the add tech stack is where you actually

[00:49:42] monetize it what we're doing is we're shifting the power into the content creation stage

[00:49:47] so while you're literally making your video and editing in Adobe Premiere and FileCupro or

[00:49:52] all these different editorial tools you literally can you know create this virtual object put it in

[00:49:57] and now this one you've locked in monetization before you ever hit distribution and so that shifts

[00:50:02] power in a 250-billion dollar industry that's got to be interesting for a lot of players and so

[00:50:10] so in short I don't know if two years right I don't know whether we're a whole IPO or get bought

[00:50:14] we're just having fun but I do know this will be relevant to a lot of people

[00:50:19] more importantly when whatever happens are you gonna buy us a drink?

[00:50:24] No yeah for sure I know yep yep yep yep it's not on the records for so absolutely and I

[00:50:31] didn't enjoy that. This episode was sponsored by Rainbarrel a new audience graph that helps

[00:50:43] advertisers reach their best fit customers in today's privacy conscious world learn more about

[00:50:49] Rainbarrel or sign up for the app at rbarrel.com that's rbarrl.com you can also take their

[00:50:57] answers for a test drive on the trade desk dv360 stack it up Samsung ads and beyond