<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Patricio Dávila on Medium]]></title>
        <description><![CDATA[Stories by Patricio Dávila on Medium]]></description>
        <link>https://medium.com/@patricio_86906?source=rss-2c9863248be2------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*Av8korkReXzUTMmH.jpg</url>
            <title>Stories by Patricio Dávila on Medium</title>
            <link>https://medium.com/@patricio_86906?source=rss-2c9863248be2------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 13:07:17 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@patricio_86906/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Programmatic advertising: Getting the most out of your digital programmatic inventory and…]]></title>
            <link>https://medium.com/hexagondata/programmatic-advertising-getting-the-most-out-of-your-digital-programmatic-inventory-and-ec17dca237d2?source=rss-2c9863248be2------2</link>
            <guid isPermaLink="false">https://medium.com/p/ec17dca237d2</guid>
            <category><![CDATA[data-management]]></category>
            <category><![CDATA[supply-side-platform]]></category>
            <category><![CDATA[demand-side-platform]]></category>
            <category><![CDATA[programmatic]]></category>
            <category><![CDATA[hexagon-data]]></category>
            <dc:creator><![CDATA[Patricio Dávila]]></dc:creator>
            <pubDate>Mon, 04 Mar 2019 22:07:11 GMT</pubDate>
            <atom:updated>2019-03-04T22:07:11.938Z</atom:updated>
            <content:encoded><![CDATA[<h3>Programmatic advertising: Getting the most out of your digital programmatic inventory and campaigns.</h3><p><strong>What is programmatic advertising?</strong></p><p>Programmatic advertising is a catch all term that involves utilizing big data and technologies likes ad networks, Supply side platforms (SSPs), demand side platforms (DSPs), real time bidding (RTB) and Data Management Platform (DMPs) in order to serve personalized targeted ads that provide relevant experiences to consumers across different channels.</p><p><strong>So what exactly what are SSP and DSPs?</strong></p><p>SSPs are a transactional technology platform which caters to publishers and allows them to set up their digital inventory in real time marketplaces.</p><p>DSPs are the other side of the coin, they allow for media buyers to access digital inventory that is available in real time marketplaces. They allow for personalized targeted advertising and for the creation of marketing strategies in line with their goals.</p><p><strong>Back to programmatic!</strong></p><p>The combinations of these technologies allows for algorithms to interact with the buying and selling of digital spaces and creatives. Depending on which side of buying and selling you are these algorithms leverage consumer behavioral data or bidding data in order to create budgets and goals, or price floors and ad placement optimization.</p><p>These methods have only just now now become a part of the lumascape as increases in computing power, cheap data storage, automation, improvements in advertising infrastructure across the board and increased data on consumer’s preferences have allowed for dramatic changes to occur. Programmatic advertising is on the frontier of the digital marketing market and has created an ecosystem that rewards players who make data driven decisions and who are capable of using all of the tools available in the digital world intelligently in order to drive their most relevant performance indicators to success.</p><p><strong>Why is it important?</strong></p><p>Not only has the ecosystem changed to the point where programmatic advertising is not only capable of being implemented but rather programmatic buying is now going to be worth $14.88 billion of the approximately $58.6 billion digital advertising spending last year, according to <a href="https://adage.com/directory/emarketer/2910">eMarketer</a>. That’s a $5 billion leap from 2014, when it accounted for $9.9 billion. That’s a 40% year after year growth!</p><p><strong>How does it happen?</strong></p><p>Programmatic happens across a digital exchanges where automated selling of digital space across different channels likes desktop, mobile, search and video is supplied to advertisers. Essentially publishers (on the supply side) use SSPs to auction their digital space to DSPs that represent advertisers in real time. The whole bidding process happens within milliseconds of entering a digital property, when a Unique User Id enters a digital space the website corroborates the existence of consumer data on that Id on the DMP, then the SSP sets up the auction on the ad exchange (or multiple ad exchanges which is called Header Bidding), at which point the DSPs bid for that impression. Finally, the highest bidder gets the impression and the creative is generated at the digital property. The rules of the auction depend on the exchange where they happen, some auctions are 2nd price auctions and they come with different fixed rates for the transaction to happen.</p><p>At a high level programmatic advertising is much like buying and selling stocks on the stock market. Like our modern financial systems, the programmatic advertising environment rewards players that are capable of creating automated systems that are capable of fulfilling millions of orders that must happen within milliseconds.</p><p><strong>Using a DMP</strong></p><p>DMPs are platforms used by owners of digital properties with advertising space that they use to identify and centralize all the information that is gathered on their audience (This is first party data). They use these platforms to sync with their SSPs so that the attributes which make their audience valuable is represented on the marketplace. On the other hand, advertisers and brands use DMPs to model their ideal customers.</p><p>The tracking that DMPs provide to buyers and seller is what drives an audiences’ value. This personalization of creatives to consumers is already evolving to include both online and offline data which provides a real insight into who the “real” user is. Leveraging this data allows for the success of marketing initiatives and to make a media seller’s inventory acquire comparative value.</p><p><strong>Hexagon Data’s intelligent clients</strong></p><p>Here at Hexagon Data we are committed to have our publisher clients maximize their audiences’ value. This is why we have created personalized taxonomies for our clients that satisfy the needs of the global marketplace, while simultaneously reordering and re-categorizing their online consumer behaviors into their newly created taxonomies this allows them for the creation of audiences that are both large and interested in specific content.</p><p>Our services allow our client’s inventory to differentiate itself from the abundance of poorly categorized websites. This is what drives the price of impressions on their websites, since having intelligent design and creative infrastructure set up into their audience management makes users more likely to click on advertising content as well as creating behavioral changes in them. Our catered taxonomies also create the circumstances for the users of their websites to have more pleasant advertising experience as well as having more relevant content shown to them.</p><p><strong>Programmatic @</strong><a href="http://www.hexagondata.com/"><strong>Hexagon Data</strong></a></p><p><a href="http://www.hexagondata.com/">Hexagon Data</a> has a compromise to understanding who is that is buying our clients’ inventory and how much does each buyer value their inventory. Our programmatic team is constantly asking itself how is it that we must update business rules, price floors, and what inventory is being sold and to who in order to maximize our client’s revenue while simultaneously creating a relevant ad experience for their users.</p><p>This is why our programmatic team looks through our clients’ Ad Exchange data in order to decipher what are the price ranges on inventory we offer. What is the spread between winning bids and closing bids by our most important buyers? What is the demand that our inventory has and how does it change over time? Which are our client’s most important buyers and how can we attract to our site?</p><p><strong>Price floors and business rules</strong></p><p>By using historical data from our clients’ data Exchange we have unique insights that we apply to our pricing rules. For example, we know that advertisers tend to buy ads higher by the end of month and on holidays. Also, on Friday and weekends the eCPM prices are slightly higher (about 5–7% higher in ads eCPM).</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ec17dca237d2" width="1" height="1" alt=""><hr><p><a href="https://medium.com/hexagondata/programmatic-advertising-getting-the-most-out-of-your-digital-programmatic-inventory-and-ec17dca237d2">Programmatic advertising: Getting the most out of your digital programmatic inventory and…</a> was originally published in <a href="https://medium.com/hexagondata">HexagonData</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Collaborative Filtering: How predictive modelling tells us what we want.]]></title>
            <link>https://medium.com/hexagondata/collaborative-filtering-how-predictive-modelling-tells-us-what-we-want-37848533ece7?source=rss-2c9863248be2------2</link>
            <guid isPermaLink="false">https://medium.com/p/37848533ece7</guid>
            <category><![CDATA[collaborative-filtering]]></category>
            <category><![CDATA[marketing-technology]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[technology]]></category>
            <dc:creator><![CDATA[Patricio Dávila]]></dc:creator>
            <pubDate>Mon, 28 Jan 2019 23:02:43 GMT</pubDate>
            <atom:updated>2019-01-28T23:11:23.790Z</atom:updated>
            <content:encoded><![CDATA[<h3>Collaborative Filtering: How predictive modeling tells us what we want.</h3><p>Recommender systems are everywhere in our modern internet environment. They have become integrated into services that some of us use everyday like Amazon, Netflix, Spotify and many more. Recommender systems have been put into place to help us navigate the myriad of options and choices which our modern economy has given consumers. However, the implementation of these systems come with many known challenges and developers must strife to make them the most robust possible so as to avoid exploits like positive and negative feedback loops that would favor one item over another, this is usually referred to as overfitting and is one of the main drawbacks of collaborative filtering.</p><figure><img alt="Resultado de imagen para collaborative filtering" src="https://cdn-images-1.medium.com/proxy/0*8NDUGfEMdlPzU0RI.png" /></figure><p>So while a recommender system refers to a system that is capable of predicting the future preferences that a user has over a set of items we are going to focus on collaborative filtering algorithms in this entry. These systems come in two flavors: (i) User based systems which measure the similarity between target users and other users. (ii) Item based systems which measure the similarity between the items that target users interact with and other possible items. In layman terms, item collaborative filtering answers the question “Users who liked this item also liked…” while user filtering answers “users who are similar to you also liked…”.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/638/0*Htj6aTGRBSKgegaQ" /><figcaption>Difficulties in using these matrices</figcaption></figure><p>The theoretical framework of this algorithm relies on inputting a large matrix where its rows represent the entry for a user and the columns represent the entry for a particular item. These distance matrices can be calculated with a variety of different measures that vary from Jaccard, gower, cosine based similarity or correlation based similarity. However, due to the abundance of content (items) these matrices may be misleading as these calculations of distance/similarity are done with an abundance of missing values. This is particularly bad when using item based collaborative filtering since there are more missing values when trying to compare an item to all others rather than comparing users that are already likely to be consuming the same items. That’s why user based collaborative filtering is usually performed in order to actually improve service and customer satisfaction.</p><figure><img alt="Resultado de imagen para collaborative filtering" src="https://cdn-images-1.medium.com/proxy/1*9_a6dvYrVevBiX2FQG9-JQ.png" /><figcaption>One of the possible measures for matrix calculation</figcaption></figure><p>The value added of collaborative filtering for businesses comes from the fact that it helps alleviate consumers of their choice atrophy, caused from the amount of possibilities that a service offers them. Collaborative filtering helps limit how much we need to process at any given time and helps reduce the complexity of what we are perceiving. This in turn increases the interactions that customers have with products and contributes to having more successful transactions — Or in other words, more transactions.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/485/0*1-tCMpu4h7EwkKph.jpg" /><figcaption>Why user based CF is best</figcaption></figure><p>Here at <a href="https://www.hexagondata.com/">Hexagon Data</a>, we have explored the possibility of using a <a href="https://www.lotame.com/lotame-data-stream/"><em>DMP Data Data Stream</em></a><em> </em>in order to create exciting opportunities for our clients. Firehoses like the ones we are thinking of using are essentially a massive database which contain all the records of cookies, interactions with websites that are tagged with a DMP. These records contain lists of the content that cookies visited, their timestamp signaling when did it, what country they did it from and the website where they did it. We think that this product has tremendous potential to be modeled as a dynamic social network and that it can be decomposed as a bipartite graph with users on one side and content on the other, or even users on side and websites on the other. Due to the structure and form of the data we think that it would be possible to measure the distances between users, items and websites with a large robust data set.</p><p>One possibility for this product is precisely the creation of a service with a back end in collaborative filtering that would be capable of creating new and innovative opportunities that would be capable of improving our clients’ ability to create relevant content for their users and automatize audience detection and targeting. Essentially by measuring the similarity of the content which users visit we wish to be capable of not just identifying communities of like minded people but also further understand the attributes which make them most similar but also the content which members of a given community may be interested in but are not are currently not interacting with. Collaborative filtering would allow us to skip over traditional marketing segments like age, gender, household income and find naturally occurring communities of users which are interested in a sub section of content that might not be intuitive to think of as belonging together in a more traditional framework of thought.</p><p>In <a href="https://www.hexagondata.com/">Hexagon Data</a> we think that this technology would not only allow us to better understand the content which similar users are actually consuming but also allow us to target audiences which would be much more affine to the ads that we are showing them therefore increasing the click through rates of our campaigns. Furthermore, we also believe that there is potential for editorial consulting as well by being able to identify what the content that users that visit a publishers site are also viewing on other sites as well as identifying which websites have the most similar content featured on their site. We believe that these systems would be capable of improving campaign performance, help our clients better understand their audience and identify their main competitors.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/400/0*yBBll_kzq1l2GWcz.jpg" /><figcaption>Hexagon Data</figcaption></figure><p>By: Patricio Davila<em>— Data analyst</em></p><p>Citations and acknowledgements:</p><p><a href="https://d4datascience.wordpress.com/2016/07/22/recommender-systems-101/">https://d4datascience.wordpress.com/2016/07/22/recommender-systems-101/</a></p><p><a href="https://towardsdatascience.com/collaborative-filtering-based-recommendation-systems-exemplified-ecbffe1c20b1">https://towardsdatascience.com/collaborative-filtering-based-recommendation-systems-exemplified-ecbffe1c20b1</a></p><p><a href="https://medium.com/@cfpinela/recommender-systems-user-based-and-item-based-collaborative-filtering-5d5f375a127f">https://medium.com/@cfpinela/recommender-systems-user-based-and-item-based-collaborative-filtering-5d5f375a127f</a></p><ul><li><a href="https://towardsdatascience.com/various-implementations-of-collaborative-filtering-100385c6dfe0">Various Implementations of Collaborative Filtering</a></li><li><a href="https://planetachatbot.com/tagged/collaborative-filtering">Collaborative Filtering - Planeta Chatbot : todo sobre los Chatbots y la Inteligencia Artificial</a></li></ul><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.slideshare.net%2Fslideshow%2Fembed_code%2Fkey%2F77iTD977FDmyOC&amp;url=https%3A%2F%2Fwww.slideshare.net%2Ferikbern%2Fcollaborative-filtering-at-spotify-16182818%2F10-Supervised_collaborative_filtering_is_pretty&amp;image=https%3A%2F%2Fimage.slidesharecdn.com%2Frecommendationsatspotifyslideshareversion-130125172143-phpapp02%2F95%2Fslide-10-1024.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=slideshare" width="600" height="500" frameborder="0" scrolling="no"><a href="https://medium.com/media/8af811e81f43cd00b5f2e7be9a2d0b8c/href">https://medium.com/media/8af811e81f43cd00b5f2e7be9a2d0b8c/href</a></iframe><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=37848533ece7" width="1" height="1" alt=""><hr><p><a href="https://medium.com/hexagondata/collaborative-filtering-how-predictive-modelling-tells-us-what-we-want-37848533ece7">Collaborative Filtering: How predictive modelling tells us what we want.</a> was originally published in <a href="https://medium.com/hexagondata">HexagonData</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>