MediaMath Blog - Intelligence

main

Data Intelligence Media

Three Major Omnichannel Challenges Today

August 4, 2017 — by MediaMath

This article originally appears in DMN News.

Omnichannel has been top-of-mind for marketers since the advent of digital media, and it’s hard to argue with the progress businesses have made in omnichannel marketing over the last decade or so.

Indeed, the industry has come a long way from extolling the benefits of omnichannel to today’s world, where businesses not only understand the benefits of omnichannel marketing, but are increasingly facing pressure from customers and partners to be omnichannel as a standard.

“Today, omnichannel marketing across all addressable channels and inventory, coupled with identity resolution and machine-powered optimization are table stakes for all media buyers and as a result, enriches the consumer experience,” says Dan Rosenberg, chief strategy officer at MediaMath. Not only does omnichannel execution allow you to manage the frequency of ads… but by adopting a more audience-based approach, marketers will be able to consolidate as many addressable channels as possible to enable one-to-one storytelling and messaging, no matter where a customer is connected.”

There are a few key areas of contention that continue to challenge omnichannel marketing as a concept, and marketers will likely grapple with these for the next few years.

Managing the customer journey

The customer journey is extremely difficult to track these days. It’s harder than ever for marketers to distil the customer journey down into the neat funnels that were once standard to the marketing process. Still, marketers are going to have to figure out how to engage customers across disparate channels as best they can.

“Managing consumer data across channels is a challenge with teams that are historically silo’ed and not incentivized to share data. Marketers need to understand the 360-degree customer journey, so that a marketer can address a given consumer’s concern in the moment,” Rosenberg says.

Privacy

As is the case for practically all digital media, privacy and data ownership will continue to be big concerns for brands doing omnichannel marketing, particularly because of the multiple channels and touchpoints involved.

“As part of privacy, marketers should be good stewards of consumer data, and not advertising too aggressively or invasively with the use of frequency caps. Using frequency caps across channels curb the number of times a consumer sees advertisements from a given marketer on any device,” Rosenberg says.

Fraud

Similar to privacy, marketers doing omnichannel have a vested interest in the advertising industry’s battles with fraud.

“Fraud has been a longstanding issue within advertising where marketers are realizing that fraud is susceptible across all channels including fake bot data, fake social media profiles and not just an ad tech,” Rosenberg says.

There’s little in the way of best practice here, as these are issues that affect all of marketing, not just omnichannel, and the progression of technology advances and exacerbates problems like privacy and tracking the customer journey.

In the end though, omnichannel is well worth the effort.

Data Intelligence

The Power Of Agency-Tech Company Relationships: A MediaMath Case Study

July 11, 2017 — by MediaMath

This article originally appears on adotas.com.

Recently, MediaMath, partnered with digital agency PMG to help a global beauty brand address struggles they were experiencing when trying to connect with customers, in real-time.

The brand wanted to deliver its ads to high-affinity audiences in ways that were optimized and personalized. By working together, PMG was able to construct audience profiles in a way that would allow them to target new and dynamic users. The company collected past data into a centralized platform to create current audience profiles built on the history of the customer’s actions. The result was improved ROAS and proof that marketers can improve the user experience of online advertising for greater efficiency for brands.

This represents the potential future of relationships between agencies and tech providers. Lead by CEO Joe Zawadzki (pictured top left), MediaMath believes that by customizing approaches and developing a strategy, a partnership leads to better results for all involved. Agencies must embrace ad technology to stay ahead of their competitors, and by partnering with ad tech providers on a strategic level they can make sure to meet the goals of the brands they are working with. Download the case study, A Triumvirate Approach to Helping a Global Beauty Brand Target New and Dynamic Users in Real-Time.

Adotas explored to subject in a Q&A with Chris Keenan, Director Product Management at MediaMath (pictured left).

Q: This PMG case study, mentions that marketers struggle to connect with customers “due to audience pools that are incomplete, inaccurate, and siloed from activation.” Why does this happen and why is it so commonplace?

A: Incomplete and inaccurate audiences that are siloed from activation are the primary drivers for marketers moving to platforms that have tightly integrated demand side and data management platforms.

We have found that when pushing audience segments from standalone DMPs to DSPs for activation, there is a 10-20% loss in reach due to the cookie sync required. This is particularly troubling when marketers have a granular, niche segmentation strategy. Losing any portion of your low-recency, high-frequency, high-value users results in lower campaign performance.

In addition, cookie-less identity resolution has become a “must-have” as mobile is exploding and the cookie is going away. Three years ago, you could reach 75% of a user with desktop cookies alone. Today, that number is less than 40%. For mobile, you will miss 80% of available RTB impressions without a cookie-less identifier.

A common complaint I hear from friends when I tell them I work in the online advertising industry is “Why am I followed around the internet with ads for a product I purchased days ago?” Another complaint I hear when speaking with marketers is that frequency caps set at the audience level are not respected. This happens when DMPs only update their segments in batch, as opposed to real-time. DMPs that evaluate segment membership in real-time will immediately remove the user who just converted or the user who just received their 8th impression. Not only does real-time segment evaluation drive performance, it also leads to a more positive user experience.

Q: Can you talk more about the pixel resolution/limitations of this client and why they prevented audience targeting?

A: Like many online retailers entering the holiday season, the client had entered a ‘code freeze’ period where they were not able to make any modifications to their existing or place new tags onsite. Code freezes are enforced as a precaution against changes having unintended, negative consequences during historically busy periods that could result in hundreds of thousands of dollars in lost revenue in a very short amount of time.

Fortunately, the client already had global footer and conversion tags placed. From these tags, we were able to ingest standard retail variables, such as Product SKU, Product Brand, Product Price, Search Phrase, Cart Value, and Order Value. These variables, in conjunction with Page URL, were used to develop a granular audience profiling strategy.

Q: How are sophisticated agencies and brands utilizing data to drive campaign performance?

A: Everyone has heard the phrase ‘the right message, to the right person, at the right time’ before, but how do marketers position themselves to deliver on that principle? I have been fortunate to speak with our clients around the globe and here are two common themes amongst the most sophisticated marketers:

• Eliminate latency
DMPs need to have an incredibly tight feedback loop with their DSP counterparts to remove latency between the two platforms. The propensity for users to convert is highest within a few hours, but if your DMP is only syncing your audiences to your DSP once every 8 hours, or worse, once a day, you are losing the opportunity to reach your audience during the most optimal time. Then, once those users have converted, you will still be wasting your media budget on those users until the next time your segments are re-synced.

By eliminating this latency, tightly integrated DMPs can reach the user with a message for complementary product offers instead of wasting the impression with a creative for the product they just purchased. It also allows for capabilities like accurate global frequency capping and sequential messaging.

• Act quickly and decisively
Today, many marketers are waiting for campaigns to run for at least a few days before looking at any analytics regarding the audiences they are targeting. By storing the raw event, impression and click level data, modern DMPs will be able to tell you how your newly created audiences would have performed against prior campaigns, without spending any media dollars against those audiences. If there is not enough 1st party data available, smart marketers generate similar performance reports to determine which 3rd party audiences will help them accomplish their reach goals in a performant manner. These tactics allows marketers to test their hypotheses without burning through their media budgets; further improving their ROAS.

Another benefit of working with a DMP that stores data at the raw event level allows marketers to define an audience segment that is fully populated and ready for activation immediately. The days of creating a segment, waiting for the audience to populate before activating it in a campaign are over. This is particularly valuable unplanned budget is made available and needs to be spent against while still meeting high performance goals. It also means that you can redefine your segment definition on-the-fly while your campaign is still in-flight with your latest learnings/needs (e.g. drive additional reach vs. improve performance) without having to start from scratch.

Q: Beyond Adaptive Segments, what other tech innovations would you recommend marketers and publishers use to accomplish their goals?

A: One innovation that has had implications for both marketers and publishers, and become a hot topic within the last year, is header bidding. For those not familiar, header bidding allows publishers to run a unified auction across their demand partners (e.g. direct sold vs. RTB) for each impression and the highest bid wins. Header bidding provides a level playing field opposed to the legacy, waterfall method where demand sources would be ranked in priority order and only get the opportunity to win the impression if the partners above them passed on that impression first. This means that the publisher’s ad server would serve the $5 CPM direct sold campaign even when there was an RTB partner willing to pay $15 CPM for that same impression.

Q: What does this mean for publishers and advertisers?

A: For publishers, the primary benefit is an increase in revenue due to the increased the “true value” of each impression being recognized. Publishers commonly report a 30%+ increase in revenue after implementing header bidding. This increase in publisher revenue translates into higher media costs for advertisers. However, advertisers now have access to premium, first look inventory allowing them to scale their campaigns in ways that would not be possible in the waterfall world. While media cost is certainly a metric that marketers should pay attention to, they should ultimately be focused on outcome oriented metrics such as CPA and ROAS. Header bidding has the potential to create a win-win for publishers and marketers alike and is an innovation that I’ll be closely.

Intelligence Media

Why Multi-Touch Attribution Adoption Will Change More Than You May Think

June 29, 2017 — by MediaMath

Most people know that last-touch attribution is flawed. Yet it remains the leading gauge for channel performance. This ignores the fact that giving the last ad viewed credit for a sale or desired action is like giving the MVP award in a basketball game to the person who scores the final basket.

But last-touch has remained viable because it’s better than nothing. The alternative — multi-touch attribution — historically has been hard to implement.

It’s still hard, but organizations are making great strides toward making MTA a standard. In 2017, most top-tier marketers are working on fully implementing MTA. When they do have this completely up and running (and here I mean as more than just as a reporting mechanism), this will usher in profound changes for both digital marketers and for the organizations they service in terms.

Spending will change

The first change we’ll see is increased growth in programmatic spending. Consumers are spending more and more time online researching and purchasing. As a result, digital marketing spending, specifically programmatic, will continue to rise — eMarketer expects programmatic display spending to jump about 28% this year. MTA will help marketers better understand which forms of digital media are contributing to brand health, to an incremental sale or other desired customer behaviors.

Equally importantly, MTA will illuminate what the value is of each of those marketing channel contributions. Programmatic as both a cost-efficient and customer-focused method of marketing will continue to grow as incremental performance and percent of contribution become more refined.

MTA will also prompt less spending for channels which overly benefit from last-click attribution models, specifically social media and search. Though both are very important in the customer journey, they have overly benefited from last click measurement, for many reasons.

One reason is “Tab Distraction.” For example, let’s say you’re in Chrome and you have several tabs open. (Quick: How many do you have open right now? A lot, right?) So you’re at J. Crew’s website and about to buy something, but then you decide to visit Facebook before completing the purchase (because you get distracted by another tab). When you open the page, you are very likely to see a J. Crew retargeting ad. You then go back to continue completing your purchase on J. Crew. Even though it didn’t help make the sale in this case, Facebook will get credit for this sale on last touch models. It will also get future marketing spend as a result of this (distorting both that customer’s true behavior and media relationship).

Because most marketers who use last touch models are also still using clicks as the indicator of engagement for all digital channels, those channels where customers do actually click as the biggest sign of engagement are also the beneficiaries of last click models, meaning more budgets are dedicated to them than should be.

Not surprisingly, the channels where clicks are a strong indicator of engagement are search and social. Once marketers grok the fact that customers engage with them differently on different digital channels and understand which of those engagements correlate to improved brand health and return on ad spend, then marketing budgets will right-size. Of course, search and social are important channels in the customer conversation, and we are still in the period where customers are increasing their use of both. But once fully implemented, I believe MTA will show that many clients are overspending on channels where clicks are the predominant indicator of customer engagement.

Organizations will change

Additionally, MTA will herald significant changes to the planning and activation processes that generate marketing plans and performance measurement frameworks. Most organizations have separate siloed marketing channel teams i.e. email, social, search, display, etc. That means each of those teams is goaled based on their particular channel’s reported performance.

Without MTA, each team claims credit for as many of the sales as their channel contributed to, but they claim credit for 100% of that sale, so total sales credit claimed adds up to much more than 100% when each team’s “performance” is added together. Once implemented, MTA will continue to give credit to each channel which deserves credit, but will assign partial, not full credit. This is not a minor change in terms of reported sales by team.

In the case of retail, the industry average is seven marketing touchpoints before a conversion, so MTA will apportion each of those seven marketing touchpoints with (for instance) 1/7th of the credit that they were claiming prior to MTA. This means that marketing, finance, HR, merchants and all related teams will need to work together to goal marketing against new benchmarks. That will require new processes for media planning, channel performance measurement frameworks, and new frameworks for defining and rating employee reviews and goals.

There are many reasons that MTA is difficult for organizations to implement — technology constraints, data connections, internal skill sets, required process changes, costs. Even though LTA is flawed, it is The Way It’s Always Been Done and that’s hard to fight. But strategic organizations have realized that the rewards of MTA are big enough to overcome both this inertia and the other significant obstacles. Right now the main barrier to full implementation is still siloed data, so smart marketers are spending 2017 getting their data house in order to make way for full MTA implementation. And when they do, we are going to see that increased understanding of complete customer behaviors will lead to correction in the marketing spend for click-based channels and significant organizational changes.

Intelligence Media

How to Add Ads and Then Add Some More

June 28, 2017 — by MediaMath

You’ve built a new social network called Minivan, for soccer moms and dads. You’ve signed all the soccer moms on the eastern seaboard and they’re sharing content. Your user numbers are soaring, but you’re not making money. Some of your soccer moms are leveraging the platform for their businesses and want your help to reach new markets for their locally made sports gear. Unfortunately, making a full-featured advertising platform isn’t something your small engineering team can do overnight, especially not while you’re still trying to get funding yourself.

It’s not hard to create an individual advertising campaign nowadays. There is a large set of platforms and services where you can get started with your first ad. Facebook and Google AdWords are fantastic at getting first-time users to create their first ad on their platforms. But, as your business scales up and you want to run thousands of campaigns targeting segments of your customers, those platform tools slow you down, forcing you to make campaigns one by one.

Let’s go up a level. You’re an agency — you manage a variety of brands. Keeping all the campaigns in check, performing well, and keeping clients happy is a feat. You’ll need custom software to manage everything. Enter the Demand-Side Platform. DSP is a service that manages advertising campaigns. That’s what our programmatic advertising platform does. You probably knew that, arriving on our blog.

But let’s go up another level. MediaMath is designed to be API First. What does that mean? Everything in the platform is built on top of our API services. If you’re in our T1 platform and creating a campaign, that’s a series of API calls your browser makes to our services. The T1 platform is a shiny facade to the gritty advertising workshop inside.

Let me break that down with an analogy:

If you want to buy a chair, you go to a chair store, and pick one up. If you have custom needs, you can reach out to a chair factory who will custom build one for you. As of now, we’ve had lots of companies willing to customize the campaign to your needs. But what if you need more? What if you want the factory itself?

This is the idea behind MediaMath’s developer platform. The engineering team at Minivan, for instance, can build MediaMath’s DSP directly into Minivan’s system letting your users create their own campaigns to target their ideal audience. Now, instead of seeing car ads on your service, your customers are marketing to each other. Such ads can be distributed on the wider web as well.

APIs don’t just allow individual actions — they give users an ability to franchise their services, allowing deep integrations into new markets. This, at least, is one road that Minivan can take.

To learn more about MediaMath APIs, click here.

Intelligence Media

The Role of Viewability

June 23, 2017 — by MediaMath

Over the past few years there has been a lot of discussion about viewabilty. Does viewability really matter though? The answer is a frustrating “yes, but…”

Viewability tends to matter less for direct-response advertisers. Because such marketers have no trouble measuring the outcome they’re trying to achieve, they can afford to treat viewability as a means to an end rather than an end unto itself. DR advertisers care whether their ads are visible, but the visibility (or lack thereof) will manifest in incremental sales, which is the bottom line.

But brand marketers are in a different position. They often don’t have the luxury of an easy-to-measure success metric because the bulk of their sales take place in an environment that’s hard to attribute to digital marketing (such as a brick-and-mortar supermarket) or relate to a product with a very long consideration cycle (such as a car). If it’s hard to directly attribute media spend to the outcome you care about, measuring the quality of that media is the next best option. So for brand marketers, the primary goal is usually just to ensure that they convey an engaging, impactful message to the right audience. That makes viewability a natural fit.

Is Maximizing Viewability a Good Idea?

It might be tempting to aim for 100% viewability, but this probably isn’t achievable for most digital campaigns. That’s because you can’t know in advance whether most impressions will be viewed before they’re served. Most pages tend to load all of their content and advertisements at once, which means you have to bid on the impression before knowing whether the user will scroll to the middle of the page where it renders. As a result, advertisers usually shoot for somewhere in the 60-70% range.

Even if you could achieve 100% viewability, should that be your goal? If the campaign only reached a handful of users or bought inventory at an exorbitant cost, it would be tough to consider it a success. That’s the challenge with using viewability as the sole measure of a campaign’s performance. All things being equal, it’s better to reach 1,000 people in a campaign with 70% viewability than 500 people with 100% viewability. Cost per viewable impression (aka viewable CPM) might be a better metric than viewability in many cases.

Optimizing Viewability

With MediaMath’s viewability solution, you can select a third-party partner to verify whether impressions were viewable and use their measurement as the success metric for algorithmic optimization. Harnessing machine learning is critical—studies have shown that many variables influence the likelihood that an ad will be viewed, and the impact of these variables varies by vertical. We update our models every 24 hours to ensure we’re capturing up-to-date changes in inventory and performance.

And we offer these tools because our customers ask for them. But assigning the importance of viewability is every marketer’s individual judgment call. In other words, how much viewability matters varies depending on who you are and what goals you’re trying to achieve.

Intelligence Media People

Why MediaMath, Why Now

June 21, 2017 — by MediaMath

We’re living in challenging times. Publishers are fighting to win back media spend that has largely fled to a small handful of players. Buyers are fighting to make sure that they don’t run against unsafe — or untrue — content. And ad enablement platforms are fighting to justify the share of media dollars that they take out of the ecosystem. The result is an overcomplicated, often irrational framework that sits between marketers and consumers.

We would all agree that there’s got to be a better way.

In my view, buyers aren’t shunting the lion’s share of budgets to those few players because they necessarily want to king-make an oligopoly — rather, buying on those properties is easy to do, and carries an ostensible halo of brand safety. If you believe that premise to be true, then it naturally follows that if another entity were to offer equivalent levels of ease, brand safety, addressability and scale, it would create a compelling alternative for media buyers to consider.

Historically the task of curating inventory has fallen to individual exchanges and publishers. What that effort offers buyers is incomplete at best — it checks the safety box but doesn’t necessarily address ease and scale. Seeing this behavior across the industry, MediaMath is radically innovating around media curation and addressability.

In my conversations with members of the MediaMath leadership team, I was immediately struck by how much respect the organization has for the supply side. The company makes its money by selling to advertisers, of course, but is keenly aware of the critically important role that high-quality inventory plays in its ongoing success. Having run supply businesses at display, video and mobile platforms, I saw an acute connection between the passion I have for empowering publishers and making digital advertising simpler, and MediaMath’s core values.

Most would agree that programmatic in its current form needs improvement. MediaMath’s first major step was to introduce the Curated Market product — easily accessible premium, audience-informed, hand-picked inventory with a brand safety guarantee. Next comes addressing operational inefficiencies inherent to OpenRTB — bidding against yourself, dynamic floors, low win rates and unnecessarily high eCPMs. Taking these steps will permit the company to meet buyers’ needs while helping restore budgets to a diverse array of premium publishers.

We’re here to make programmatic a cleaner, safer, simpler, more cost-effective experience for buyers, and to help publishers take back control over their futures. I couldn’t be more honored to be a part of this team, and to work with so many of you as partners.

Intelligence Media

How Programmatic Will Change the Brand-Agency Relationship

May 15, 2017 — by MediaMath

This byline by Parker Noren, Senior Director, Programmatic Strategy & Optimization, originally appears in CommPRO.Biz.

Ten years after its birth, programmatic advertising is in its awkward teenage years. The existing technology has allowed those who have fully embraced its capabilities to build relationships with their consumers in ways previously inaccessible. But, we need to look forward to see when programmatic will fully deliver its value for all brands. Programmatic allows us to achieve greater brand growth as the result of algorithm-driven optimization and a better understanding of the consumer—all accomplished with less manpower due to automation.

Technology improvements will play a critical role in making this value more accessible. Algorithms will get smarter and more independent from human guidance. More touchpoints will become addressable as we’ve seen with channels like out-of-home. And, we will get to full integration of what today are considered separate but connected silos (e.g. DMP and DSP). Yet, what’s potentially more interesting is how the improvements in technology will influence the roles of brand and agency in advertising execution. Brand marketers will take direct control of the execution of their advertising dollars, while media agencies will need to find new avenues to prove value.

Some brands are already there, especially those with a heavy e-commerce presence where the benefits of direct control were most obvious from the start. But, for many brands, the marketer is still far removed from actual execution. The disconnect was necessary (and still is in linear channels) because of the heavy lifting and knowledge required to execute an ad buy. As we look forward, tactical skills specialization will no longer be needed to execute media to the degree necessary in the past or even today. Scalable technology will fulfill this role, and the job of a campaign planner/manager will shift towards campaign definition and strategy.

The move in this direction—enabled through technology—will reshape roles, activation strategy and success criteria:

Media agency as a strategy agency.

Agencies will largely pivot from value provided in execution manpower and knowledge to value as a strategic adviser. They’ll consult their clients on how to form the proper tech stack for their business needs and push them to update their approach to campaign definition so those capabilities can better be leveraged. This includes helping the brand reform channel- and product-based budgeting systems to focus purely on the relationship with the consumer.

Greater connective tissue between marketing strategy and advertising execution.

With marketers directly involved in execution, brands will gain ownership of a holistic view of their data and begin to develop a feedback loop between marketing strategy and advertising execution. This will produce more purposeful approaches to campaign definition, supply curation and consumer targeting better aligned with the marketer’s ingoing strategic intent.

Heightened pressure on marketers to demonstrate real performance.

Those easy, unsophisticated (“spend the budget”) dollars will all move towards accountability in true business outcomes. The move will be spurred by greater ease of accessibility to sophisticated measurement approaches, marketers having direct transparency and control of how budgets are spent, and the overall trend towards marketing as an accountable revenue center.

For more insights on programmatic’s past, present and future, check out our #10Then10Ahead Content Hub!

Intelligence Media

AdWeek: 3 Ways Programmatic Can Graduate From Adolescence

May 8, 2017 — by Michael Lamb

This article by MediaMath President, Mike lamb, originally appeared on AdWeek.

As programmatic advertising hits its 10th anniversary, it represents more than 50 percent of all non-search digital ad spending. What began as a quirky set of technologies and protocols for monetizing long-tail display inventory is poised to become the dominant framework for all of digital advertising.

The initial disruption is behind us, we’ve clearly reached the end of the beginning and the future is full of promise. But it’s equally clear that, in order to deliver on that promise, we need to make some fundamental changes, as the status quo is satisfying neither the marketer nor the consumer.

The rapid adoption of programmatic has demonstrated that marketers aren’t willing to settle for mere automation — and why should they? Consumers have been very clear that they expect control and relevance . I believe that, if accorded this respect, they will respect the business model that supports the internet.

This is the crucial year in our adolescence in which we begin taking responsibility for our actions. There are three urgent tasks the ad tech industry must address if we are to make it to full maturity without alienating our clients and consumers in 2017. Indeed, there are three promises unfulfilled that left ignored, will stunt our continued growth if we as an industry do not address them now.

Provide true addressability

Marketers were promised full addressability across all channels, and consumers were promised a personalized, customer-centric experience online. Marketers want to communicate with people, not “users” and “users” want to be treated like people. Strong, reliable identity solutions needs to go hand-in-hand with strong privacy, data governance and a consumer “bill of rights.”

Pull the curtain back

Marketers are demanding an experience that is reliably fraud-free and brand safe, with transparent and rational economics. Consumers deserve absolute protection from fraud and malware. Quantification will set us free here. There’s no retreat from granular, buy-side measurement of advertising effectiveness, and publishers deserve an equally rigorous toolset for quantifying the contribution of advertising to consumer experience.

Stop our own infighting

True interoperability and transparency will require a strong partnership between marketers and publisher with regard to business outcomes and consumer experience. Antagonistic, arms-length models won’t get us there, nor will closed, monolithic systems.

We’ve talked this talk for a long time, but if we continue to fail to deliver on this, then we are going to lose the faith of consumers and marketers, and we will have lost the opportunity to shape the internet. Luckily, I already see a burst of activity in meeting all of these challenges. Holding companies are holding Google accountable for lack of control and transparency over where their ads appear. DSPs are beginning to draw the line against fraud by refusing to pass the cost on to marketers.

And on Thursday, several of the largest independent companies that have traditionally competed with one another banded together to launch a “standard identity framework that enables buyers and sellers of programmatic digital advertising to create more relevant campaigns and improve consumer experience.”

It might take us another decade to get to the ultimate vision of digital — one where consumers are opting into an experience that empowers brands and publishers alike. It’s in our power to build the programmatic experience that marketers and consumers deserve, and it’s time to get started.

To read the full article via AdWeek, click here.

Intelligence Trends

MediaMath Weighs in on the Evolution of Machine Learning and Artificial Intelligence

April 27, 2017 — by MediaMath

Machine learning and artificial intelligence are being used across industries to help mine vast quantities of data. For a recent article on ComputerWeekly.com, Heather Blank, MediaMath SVP of Audiences, talks about how marketers are utilizing these technologies.

Marketers are also harnessing machine learning to better predict how certain customers react to various marketing efforts and how likely they are to make a purchase in what’s known as conversion.

This helps brands and agencies run more holistic marketing campaigns, targeting the right audience through the most optimal channels and at the best times, says Heather Blank, senior vice-president at MediaMath.

However, Blank says machine learning is not necessarily always predictive in nature and can be descriptive instead.

“Studying machine learning models can help explain which features or individual characteristics are important in predicting an event – usually a conversion – and which features may be meaningless or even predictive of the event not occurring at all,” she says.

“This can help marketers understand their consumer patterns more clearly, by filtering out the noise. It can also help challenge status quo notions of what is important to the purchase cycle or even what an ideal consumer looks like.”

Intelligence

How is AI Really Changing Advertising? We Asked Two Experts

April 24, 2017 — by MediaMath

AI. It’s one of the hottest topics in advertising these days. It seems like every ad tech firm is promoting their AI acumen.

As with “big data” a year or so ago though, it can be hard to separate the truth from bluster. Some argue that there isn’t enough data to make AI effective for advertising and AI is merely the industry’s latest buzzword.

To get a different perspective on this issue, we talked to two professors who specialize in AI, Richard Baraniuk of Rice University and Tim Kraska of Brown University. We interviewed both separately. Excerpts of those conversations appear below:

MM: What is the difference between deep learning and machine learning?

Richard Baraniuk: Deep learning is just a special case or an example of a certain kind of machine learning. Machine learning is a broad classification of a set of techniques for solving certain kinds of problems that involve using data to construct systems that can aid in solving problems like making decisions. Deep learning is just a particular example of one machine learning technique.

Tim Kraska: Deep learning is a sub-field of machine learning. It’s one very particular type. Machine learning is a bunch of techniques, including regression models. They are different types of algorithms you can use. The most traditional ones would be linear regressions, polynomial regressions. They are ways to find data to explain your model to predict future events while deep learning is a very different type of model building which inspired by what the brain does.

MM: How would deep learning come up with different solutions than machine learning?

RB: Let me think of a good analogy…Machine learning is like talking about automobiles and deep learning would be talking about Ford, which is a particular kind of automobile. So a particularly powerful set of algorithms in machine learning are these deep learning methods. It’s basically a rebranding of neural networks or artificial neural networks which folks have been working on for 60 years or more and are very loosely based on how the brain works or might work. The reason they’re called deep learning algorithms is you take some basic machine learning algorithm and you connect it to another one and connect it to another one and so forth so you make this cascade or chain of many algorithms in a row and that creates one of these so-called deep learning approaches. And they are probably the hype-iest of machine learning approaches today.

MM: Facebook’s News Feed is often cited as an example of machine learning. Is that correct?

RB: Exactly. Any mathematical algorithm that’s able to use data and learn from that data in order to do an even better job next time, that’s a machine learning technique.

TK: Machine learning nowadays is in almost every type of product you can think of. If you go to Netflix and get a video recommendation, that’s machine learning. If you go to a store on your phone and it makes a recommendation of what to buy next, that’s machine learning.

MM: Is there a threshold, like X-amount of terabytes in which it becomes more effective?

RB: Absolutely. There are thresholds, but there’s not a general rule of thumb you can apply. It really depends on the kind of data that you have access to and the kind of prediction that you want to make. Whether you’re dealing with text data or image data or medical records data or click data, those all contain very different amounts of information, and it becomes a difference between making some kind of judgment about whether a person has found some resource or ad useful or intriguing versus trying to find some fine-grained demographic data about a person like what age were they, what gender were they, etc. As you ask more and more deep questions, you need more data.

MM: Machine learning has been around for a while. Why is it getting so much hype now?

RB: The main reason is machine learning systems have gotten way better over the last few years. Going from yesterday’s really terrible telephone number recognition – speaking into a telephone and having a machine barely understand you – all the way to basically being able to drive a driverless car around. Really the reason these methods have gotten better is three things: Folks have come up with new kinds of algorithms, but, even more importantly, computers have become so much more powerful that we can crunch on the third element, which is that we have so much data available now. There’s so much data now and so much computing power that algorithms that would have been impossible to apply five years ago are now commonplace.

TK: On the one hand, the amount of data available is very different from even 10 or 15 years ago. The other thing is that processing power got much more advanced, so you can do more things were previously possible. Plus, there’s a huge advancement in the techniques used. But the biggest influence is the tools available – tools people can use in an open source way. The whole tool ecosystem exploded. More people are using them.

MM: Do you think at some point AI will be able to make creative decisions in advertising?

TK: We are still far from that. Maybe eventually it might happen, but it’s hard to tell.

RB: Well, machine learning is already being applied to advertising, big time. Any of the ads that are placed in a Google search page, any of that is done via machine learning. But [regarding creativity], this is where we start to move from machine learning to artificial intelligence, which is a label that’s used for any problem that’s too hard to solve right now. I would say, currently, machine learning systems are extraordinarily effective in taking some training data – some data that they used to apply rules from – and then applying that data to new data. But really, they’re just learning some rules and applying them to a piece of data. Figuring out the creative process and how to mimic that in a machine is still an unsolved, very difficult problem.

Newer Posts Older Posts