ABOUT: LinkedIn



ABOUT ROCKETSHIP HQ: Website | LinkedIn  | Twitter | YouTube

KEY HIGHLIGHTS

🤔 Why it’s important to understand what parameters a DSP takes into consideration in formulating their bids.

😲 Why a bidder might correctly identify a bid request that might drive an install or purchase – and yet lose an auction.

🤷 Why a bidder might win a bid request and might not end up showing an impression.

⏳ What the lifecycle of an ad is like – from when it’s loaded in the system to when it’s eventually shown.

🔅 How bidding on lowest priced inventory can be suboptimal in driving performance.

🗄️ What the best way is to think about privacy or confidentiality concerns around sharing data with DSPs.

💡 Why DSPs sometimes may not have access to high performing inventory from publishers – and how they should address this situation.

💯 What session depth means – and why it’s important for a marketer to be mindful of this.

KEY QUOTES

How to troubleshoot programmatic

Each DSP will operate with an exchange through the open RTB spec. And this specification will provide hundreds of various data points to the DSP about the device, the app, the user, and if the marketer can have these really interesting in depth conversations with their DSP about what parameters the DSP is taking into consideration with their exchange integrations and what data points they’re not taking into consideration and why aren’t they taking those data points into consideration,  and how do they weight those data points.

There’s a health check process

So coming from the exchange world, we would go through regular health checks with our DSP integrations to make sure that there was low latency, for example, because there could be instances when a DSP is bidding on an impression and that impression would drive an install for that marketer, but they ultimately lose the auction because they didn’t bid within the allotted timeframe that they needed to.

The event flow of an ad

There could be a handful of events that the ad transacts across in this life cycle that requires it to successfully end up showing to the user. So it’ll involve – when the ad is loaded in the system, the marketer uploads the campaign and the creative, then when the DSP bids with that creative into the programmatic auction, when they win the auction, then the ad is loaded potentially in the background of the app, then when the user has time to see it, then the ad is shown. And so through that whole lifecycle, there could be many reasons why it doesn’t make it to show to the user.

Placements impact performance

Typically publishers will be able to siphon off inventory based on high performing ad units or lower session within the user depth – and they’ll usually siphon the highest value inventory to direct sold ads, direct sold advertiser clients. 

Then a DSP will never have access to that high performing inventory because they’ve already sold it. And sometimes DSPs typically will rest in the middle or longer tail subset of the publishers inventory and so being a savvy marketer to ask your DSP, do you know your placement within this high performing publishers’ waterfall – this is absolutely going to drive higher performance.

Understanding session depth

So session depth will essentially indicate where the user is within one session of the app. So, for example, if I open up a music streaming app, and within the first 10 seconds of searching for music, I’m prompted with an ad, I will absolutely be more engaged with that ad than after I’ve been listening to music for an hour. And so it’s really important to be able to identify where in the session the user sees ads, so that you can make sure that you’re connecting with them throughout the right point in their session and attention span essentially.

FULL TRANSCRIPT BELOW:

Shamanth: I’m very excited to welcome Beth Gilmore, founder of the New Department, to the Mobile User Acquisition Show. Beth, welcome to the show.

Beth: Thank you. I’m excited to be here.

Shamanth: Yeah, I’m excited to have you because we’ve worked together – and I’ve certainly seen the sheer depth of insight you have around all things programmatic. You really are able to get into the weeds and really understand the nuts and bolts of how every bit of how the programmatic ecosystem works – and I’m excited to dive into a lot of that with you today.

Beth: Yeah, I think marketers really only see to a certain level what happens with programmatic and usually what a DSP shares with them. So it’s always good to lift a veil for marketers to see what’s really happening on the back end and exchange level.

Shamanth: Yeah. A lot of people say, oh, programmatic is a black box. And seen in a certain light, that’s true. But as my conversations with you have shown, that’s not entirely true at all.

Beth: No, I mean, I’ve always been a huge fan of programmatic because the framework is all about transacting data. And there’s so much data to drive smart advertising and, and if that data is exposed and transparent, then it’s not a black box. But it’s really up to the players in the industry who share that data who’re helping create that black box, like a mantra actually.

Shamanth: Indeed. So let’s assume a marketer’s working with a DSP and let’s assume this marketer is savvy enough to really take the veil off the black box. And they’re like, look, let’s work closely with the DSP to actually impact the performance. And if the marketer notices right our performance isn’t where it needs to be. So, what are some of the areas that you would recommend this marketer investigate, so as to troubleshoot programmatic performance?

Beth: Yeah, to give a little bit of context,

Each DSP will operate with an exchange through the open RTB spec. And this specification will provide hundreds of various data points to the DSP about the device, the app, the user, and if the marketer can have these really interesting in depth conversations with their DSP about what parameters the DSP is taking into consideration with their exchange integrations and what data points they’re not taking into consideration and why aren’t they taking those data points into consideration,  and how do they weight those data points.

So I think just really getting into the nitty gritty of what data signals are driving the bidder’s algorithms and how that potentially can be optimised. There’s a DSP’s overall bidder health.

So coming from the exchange world, we would go through regular health checks with our DSP integrations to make sure that there was low latency, for example, because there could be instances when a DSP is bidding on an impression and that impression would drive an install for that marketer, but they ultimately lose the auction because they didn’t bid within the allotted timeframe that they needed to.

There are times when the bidder will send malformed bid responses and not send data back to the exchange in the protocol that they’re supposed to. So they could lose out on bids from that. There are also various creative components that might not be known to the marketers. So, usually with the video and interstitial and more rich formats, if a bidder is bidding on inventory that has maybe low network connectivity or is in a geo that’s not ideal to support the weight of that ad load, here’s a lot of optimization that can even be done at the creative level on the back end that marketers might not know is happening in real time. 

I would actually suggest to savvy performance marketers to just go to the open RTB spec and learn what data points are in the spec and get a sense of what the integration with these exchanges looks like, so that they can better have these informed conversations with their DSPs on how to optimise.

Shamanth: Yeah, that is very interesting, because there clearly are a lot of these data points, which may have nothing to do with a marketer’s marketing strategy. 

Beth: Totally

Shamanth: Right. You know if the bidder isn’t bidding fast enough, that’s an engineering problem and it will absolutely impact marketing performance, even though it has nothing to do with the performance marketer’s strategy. That is so fascinating. You did speak about some of the creative components. I’m curious if you could elaborate on those because, again, I think when I speak to a lot of marketers who advertise on programmatic, one of the aspects that does come up is that creatives really don’t have that much of an impact. So can you speak to how some of the creative components influence bidder performance or programmatic performance?

Beth: Yeah, and we’ve discussed this before that an ad will have a life cycle between the DSP and when they bid and the inventory.

There could be a handful of events that the ad transacts across in this life cycle that requires it to successfully end up showing to the user. So it’ll involve – when the ad is loaded in the system, the marketer uploads the campaign and the creative, then when the DSP bids with that creative into the programmatic auction, when they win the auction, then the ad is loaded potentially in the background of the app, then when the user has time to see it, then the ad is shown. And so through that whole lifecycle, there could be many reasons why it doesn’t make it to show to the user.

Shamanth: Right.

Beth: And, I think what typically will happen again, with these interstitial and video and more rich media formats is that publishers can cache those ads in the background of their app for periods of time, so when the user is ready to see the ad, it’ll be a better experience for them. And a lot of ad servers will release the cache after a certain period of time but sometimes publishers overestimate how long the user is going to be in the app. And so there are many times that a DSP will see a very high value impression, bid on it, win it but then the ad never clears and is never shown to the user. 

Shamanth: Right.

Beth: So in those instances, it’s really important to see where there’s high win rates for certain subsets of inventory. And then if there are extremely high win rates with low clear rates, there could be an issue with the creative lifecycle or the SDK integration with the publisher – a whole slew of things that you can then try to troubleshoot.

Shamanth: Right. So even if you’re winning bids, you may not be actually getting the impression – and that could just be something to do with the publisher itself. Yeah, and you know, speaking of the publisher, or what can be broadly called the supply side of things, are there things a marketer can potentially do to make sure they are on the right kind of supply and they are winning bids on the right kinds of supply? How should a marketer think about these?

Beth: I think this is really where having historical performance data is key. And every marketer knowing that a certain app or a certain ad unit in a certain geo is driving higher performance for them, that means that they essentially should be spending a higher rate for that inventory. And I think what happens a lot of the time, DSP bidders will over index on spending lower CPMs on high volumes of inventory, and miss out on the higher performing creatives, for example interstitials or videos in North American locations on new devices, because they’re going to go to premium and it’s really up to the marketer to push on the DSPs to price accordingly and to not always go back to the lower CPMs that have a lower rate of conversion.

Shamanth: Right. I imagine that’s one way where if you leave everything to the algorithm, you’re just going to get the lowest CPM. And you want to be as a marketer influencing the algorithm by essentially biasing it in favor of perhaps the supply sources that have worked well in the past, or the users that have converted well in the past.

Beth: Exactly and having a test program where you really are able to see performance at an exchange level, at a publisher level, app level, at a creative level, ad unit level – I mean, there are apps that have a whole variety of different ad units within the content. You know, like in a game, you might have a full screen, rewarded video placement and then later on in the session, you might be given a banner ad, so like, knowing that rewarded video placement is going to drive higher performance and then maybe black-listing that banner ad unit so that you don’t have to waste a bid on it knowing that it’s not going to convert. So having a strong test and getting the right data from your DSP to put that test program into place.

Shamanth: That’s interesting. So, when you say test program, this is essentially to identify supply sources, app IDs that perform well and almost get a seed list. Is that a fair understanding of what you are proposing & recommending?

Beth: Absolutely. 

Shamanth: Got it. And some marketers might have questions like: “Oh, if I share my supply sources, particularly, if I share device IDs, I will have privacy concerns or confidentiality concerns.” How do you recommend thinking about this?

Beth: I think as a marketer, you will only increase your performance dramatically if you share that data with your DSP. You’d like to think that every DSPs algorithms are extremely intelligent and can find those high converting users. But it’s not always the case. And there’s so much inventory out there that it’s almost getting harder for the bidders to stay smart in the clutter and commoditized exchange world. And so, if a marketer has first party data and known users that convert, it’s a really good idea to share that data with your DSP. And then of course, you can work it out with them, whether they’re allowed to use that data to better their platform or not. But yeah, you almost have to share that data to drive real performance.

Shamanth: Yeah. Certainly.

Beth: And then I think, just sorry, highlighting the user privacy policy perspective, and that is really having a conversation with your DSP to make sure that they have the proper support across their own system and all the various integrations that they have, and just asking them the hard questions to make sure that they’re not going to get you in trouble, basically.

Shamanth: And shifting gears a little bit, when a marketer is buying on a DSP, does it matter as to where in the ad monetization waterfall of a publisher there ads are shown? And how might a marketer think about this and how it’s going to impact the performance? 

Beth: Yeah, it absolutely will impact a campaigns performance based on what inventory set they’re competing against within the publishers waterfall stack, I think it’s

Typically publishers will be able to siphon off inventory based on high performing ad units or lower session within the user depth – and they’ll usually siphon the highest value inventory to direct sold ads, direct sold advertiser clients. 

Then a DSP will never have access to that high performing inventory because they’ve already sold it. And sometimes DSPs typically will rest in the middle or longer tail subset of the publishers inventory and so being a savvy marketer to ask your DSP, do you know your placement within this high performing publishers’ waterfall – this is absolutely going to drive higher performance.

And it kind of just leads into the conversation of private marketplaces, where PMPs are essentially ways for marketers to work directly with publishers to pay a premium or just pay a static price for inventory where they won’t have to compete with the entire market. Or maybe they’ll compete with a smaller subset of advertisers or DSPs for that inventory, but that’s a great way for marketers to basically access inventory in the waterfall that they might not have been able to access.

Shamanth: Right. So what you’re recommending is really taking a two pronged approach, which is – see if you can get first dibs on premium inventory by doing private marketplace deals. And for the rest of the inventory, get as much visibility as possible as to where in the waterfall you are.

Beth: Absolutely and there is this trend now in mobile where exchanges are supporting in-app bidding, advanced bidding, which is essentially trying to create one unified auction for all demand to compete at the same priority within the waterfall. But there are just so many publishers that have large teams dedicated to yield management and sales and they’ll always continue to have a more advanced, I guess you could say, waterfall where a lot of demand won’t be able to compete for certain aspects of their inventory.

Shamanth: Right. So you’re saying that’s never going to be a completely open marketplace. There’s always going to be side deals like this.

Beth: The marketplace deals for inventory, you know, performs while you also compete in the open market to get a better sense for new inventory – or maybe pockets of high performing inventory that then you can set up private marketplace deals for.

Shamanth: Right. And when you were describing that, you spoke of the high performing inventory, and correct me if I got this wrong, but you spoke about users that were lower in the session. What does that mean? Can you speak to that?

Beth: Yeah.

So session depth will essentially indicate where the user is within one session of the app. So, for example, if I open up a music streaming app, and within the first 10 seconds of searching for music, I’m prompted with an ad, I will absolutely be more engaged with that ad than after I’ve been listening to music for an hour. And so it’s really important to be able to identify where in the session the user sees ads, so that you can make sure that you’re connecting with them throughout the right point in their session and attention span essentially.

Shamanth: Got it. That makes sense. Again for a marketer who’s working with a DSP, let’s assume the DSP has past IDFAs and Android IDs of past campaigns, perhaps in that genre. I know you talked about a marketer offering their own device IDs. But what about historical device IDs that the DSP might have? Is that typically something that can be beneficial without running privacy risks for either the DSP or the marketer?

Beth: Absolutely, I mean, being able to combine historical with current known first party behavioural data will not only increase the performance – but this can be done in a way where you’re complying with all the various policies in place. I mean, according to GDPR, if anyone hasn’t given consent, then you cannot store and log their first party data for advertising purposes. And if that user has, then that historical data is fair game, and I think that’s a very smart thing to do.

Shamanth: Indeed, indeed, I think GDPR is a completely different beast. And not to get into the weeds of that, but I totally get what you’re saying. Yeah. And again for a marketer who’s working with a DSP, are there forecasts that might be available for them to help plan better, either creative formats, or source apps or even exchanges that they could use to make more informed decisions?

Beth: Yeah, I’ll be honest that I haven’t really seen that many forecasting tools out there for programmatic. And really, I think it’s just hard for publishers themselves to forecast their own inventory. And there’s so many different variables at play, like external factors out of their control – so I think forecasting has been hard for DSPs and exchanges to do. 

But coming from the exchange world, there are very robust real time analytics tools that show what’s happening in the marketplace now. And DSPs should absolutely share exchange level data, app level, publisher, creative, geo, you know, all the basics with their advertisers. And likewise, I know that if an app, if a marketer works with an exchange directly, there are a lot of times that exchange will be able to give the exchange level data to them as well. So there’s definitely a lot of real time data available to make those informed decisions for the future.

Shamanth: Indeed, indeed. And as you said that while that data may not always be perfect, it’s certainly better than flying blind. 

Beth: Yeah

Shamanth: And certainly from everything you’ve been telling us that working with programmatic and with exchanges and DSPs is far, far from being a black box. And I think, for the smart marketers, I think it behooves them to be better informed so they can make educated decisions.

Beth: Yeah, definitely.

Shamanth: Beth, that’s perhaps a good place for us to wrap up this conversation. Much like every time I speak to you, this has been incredibly instructive. So thank you so much for being on the Mobile User Acquisition Show.

Beth: Thank you for having me.

A REQUEST BEFORE YOU GO

I have a very important favor to ask, which as those of you who know me know I don’t do often. If you get any pleasure or inspiration from this episode, could you PLEASE leave a review on your favorite podcasting platform – be it iTunes, Overcast, Spotify or wherever you get your podcast fix. This podcast is very much a labor of love – and each episode takes many many hours to put together. When you write a review, it will not only be a great deal of encouragement to us, but it will also support getting the word out about the Mobile User Acquisition Show.

Constructive criticism and suggestions for improvement are welcome, whether on podcasting platforms – or by email to shamanth at rocketshiphq.com. We read all reviews & I want to make this podcast better.

Thank you – and I look forward to seeing you with the next episode!