<?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 Tyler Rich on Medium]]></title>
        <description><![CDATA[Stories by Tyler Rich on Medium]]></description>
        <link>https://medium.com/@tylerprich?source=rss-3c8d5828c075------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*I1QT_Nfvk4lUlQkFdhmhFA.jpeg</url>
            <title>Stories by Tyler Rich on Medium</title>
            <link>https://medium.com/@tylerprich?source=rss-3c8d5828c075------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 18:39:49 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@tylerprich/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[How to Measure Advertising Effectiveness with Brand Lift Studies]]></title>
            <link>https://medium.com/@tylerprich/how-to-measure-advertising-effectiveness-with-brand-lift-studies-8eae4b2bdd04?source=rss-3c8d5828c075------2</link>
            <guid isPermaLink="false">https://medium.com/p/8eae4b2bdd04</guid>
            <category><![CDATA[python]]></category>
            <category><![CDATA[ad-measurement]]></category>
            <category><![CDATA[brand-lift-study]]></category>
            <category><![CDATA[ffmpeg]]></category>
            <category><![CDATA[z-test]]></category>
            <dc:creator><![CDATA[Tyler Rich]]></dc:creator>
            <pubDate>Mon, 04 Mar 2024 04:20:51 GMT</pubDate>
            <atom:updated>2024-03-07T17:57:34.964Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*tE-BUxeb8UAXUPHqJGI1JA.png" /></figure><p>In an era where consumer privacy is becoming an increasingly important priority, <strong><em>brand lift studies</em></strong> remain a crucial tool for advertisers to gauge the effectiveness of their ad campaigns. This article delves into the intricacies of brand lift studies, focusing on the application of these survey products in video streaming platforms. I’ve also included a short <a href="https://www.youtube.com/watch?v=RU4ih8IwRyw">video demonstration</a> along with a Python notebook example that is accessible on <a href="https://colab.research.google.com/drive/1FqDg8aLYgAqkEMIiG3tcBkolvCvKMEwV#scrollTo=TEs91LTJyCau">Colab</a> and <a href="https://gist.github.com/t-m0neybagz/0c2a88181ddbd9bb7971cca279aa36c0">GitHub</a>.</p><h3>How Does a Brand Lift Study Work?</h3><p>Brand lift studies measure consumer engagement with a brand. At the core of these studies lies the <strong><em>marketing funnel</em></strong>, a concept describing the journey of consumer interaction with a brand. This journey encompasses several stages including Recall, Awareness, Favorability, Consideration, and Purchase Intent. Each stage is measured with the use of specific survey questions, mapping out a comprehensive funnel of consumer behavior:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*wGPEjXGETCY2tW2D0zP4YQ.png" /><figcaption>Survey questions correspond to specific marketing metrics</figcaption></figure><p>Understanding the methodology behind brand lift studies is key. These studies are built around a hypothesis — that ad campaigns positively influence consumer actions towards a brand. <strong>This hypothesis is tested by comparing survey responses from two distinct groups:</strong></p><ol><li><strong><em>Exposure Group</em></strong><em>:</em> Individuals who have been exposed to the brand’s advertisement. Also known as experiment group.</li><li><strong><em>Control Group</em></strong>: Those who have not seen the brand’s advertisement. Also known as the baseline.</li></ol><p><strong>To quantify the ‘lift’, or the impact of the ad, statisticians often employ the Z-Test.</strong> This test helps calculate the Incremental Lift Percentage and assess the Significance (P-Value) of the results, ensuring the findings are not just by chance:</p><ol><li><strong><em>Incremental Lift Percentage</em></strong>: The proportion of the exposure group that responds positively to a brand in relation to the control group. Calculation: (Exposure Group Proportion - Control Group Proportion)</li><li><strong><em>Significance (P-Value)</em>: </strong>The likelihood that the incremental lift from the sample group can be broadly applied to the population. A lower p-value is preferred because it indicates a lower probability that the result is due to chance (i.e. the result has greater significance).</li></ol><h3>Creating a Brand Lift Study on YouTube</h3><p>Imagine we’re running a campaign for a popular streetwear brand, T-Moneybagz. To evaluate our campaign’s impact, <strong>we set up a brand lift study on YouTube, focusing on metrics like ad recall and purchase consideration</strong>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*V8UaMVSvOt8kwV38dqPjwQ.png" /><figcaption>Brand lift study settings on YouTube</figcaption></figure><p>Once the brand lift study is published, it undergoes a period of data collection while the campaign is delivered and survey responses are gathered. Results are typically made available for analysis after a period of several days.</p><h3>Measuring Brand Lift Study Results</h3><p>The outcomes of brand lift studies hinge on two factors: positive incremental lift and statistical significance (p &lt; 0.05). For T-Moneybagz, we observed a 10% incremental lift in Ad Recall but no significant lift in Consideration. This disparity is not uncommon in brand lift studies, where upper-funnel metrics like ad recall often show more significant results than lower-funnel metrics like consideration.</p><p>While most advertisers are only presented with a high-level summary of results, behind the scenes, we can review the full results of our hypothetical brand lift study study. We notice that there appears to be 2% incremental lift in consideration (7% — 5%), but the results are not significant according to the Z-Test and therefore will not be shared with the advertiser, T-Moneybagz.</p><h4><strong><em>Ad Recall — 10</em>% incremental lift </strong>😊 :</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xCj7oB_DWRus8WHQkobW1Q.png" /></figure><h4><strong><em>Purchase Consideration — </em>No visible results</strong><em> </em>😢 :</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*gQa7fboJk9Yw5UnGHmv4-g.png" /></figure><h3><strong>How Best to Achieve Significant Results</strong></h3><p>Achieving significant results in brand lift studies involves two key strategies that are showcased in the video below:</p><ol><li><strong><em>For Advertisers</em></strong>: Craft ads with a clear, resonant message targeted at the right audience. This ensures higher brand recognition and preference post-ad exposure.</li><li><strong><em>For Ad Platforms</em>: </strong>Increase the survey sample size. A larger dataset can often lead to more statistically significant results.</li></ol><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FRU4ih8IwRyw%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DRU4ih8IwRyw&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FRU4ih8IwRyw%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/aef147aea6a67d613e250f0e450f7a24/href">https://medium.com/media/aef147aea6a67d613e250f0e450f7a24/href</a></iframe><h3>Conclusion</h3><p>Brand lift studies are an invaluable tool for measuring ad effectiveness. By understanding their methodology, analyzing results correctly, and implementing strategic actions based on these insights, advertisers and ad platforms can significantly enhance the measurable impact of campaigns.</p><h3>Appendix:</h3><p>Complete Python Notebook:</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/4188e73e6e8c997cb56fde9615ff3810/href">https://medium.com/media/4188e73e6e8c997cb56fde9615ff3810/href</a></iframe><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=8eae4b2bdd04" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Creating a Lookalike Audience from a Source Audience — A Python Example]]></title>
            <link>https://medium.com/@tylerprich/creating-a-lookalike-audience-from-a-source-audience-a-python-example-670f3684c4fb?source=rss-3c8d5828c075------2</link>
            <guid isPermaLink="false">https://medium.com/p/670f3684c4fb</guid>
            <category><![CDATA[sklearn]]></category>
            <category><![CDATA[python]]></category>
            <category><![CDATA[ad-targeting]]></category>
            <category><![CDATA[lookalike]]></category>
            <category><![CDATA[matplotlib]]></category>
            <dc:creator><![CDATA[Tyler Rich]]></dc:creator>
            <pubDate>Tue, 20 Feb 2024 08:20:10 GMT</pubDate>
            <atom:updated>2024-02-27T02:23:10.279Z</atom:updated>
            <content:encoded><![CDATA[<h3>Creating Lookalike Audiences from a Source Audience — A Python Example</h3><p>In the realm of online advertising, one of the most efficient ad targeting features is the creation of <strong><em>lookalike audiences</em></strong>. These are groups whose characteristics resemble those of a <strong><em>source audience</em> </strong>provided by an advertiser. Businesses can leverage their customer data (emails, phone numbers, etc.) to form lookalike audiences on various ad platforms, enhancing their ability to attract new customers while increasing cost efficiencies of their advertising efforts. This article demonstrates how lookalike audiences are practically created and assessed before launching campaigns, and it is accompanied by a Python notebook example available in <a href="https://colab.research.google.com/drive/1vdF0oxJCryYXzbnRrmIjspaw00FQpTIs">Colab</a> and <a href="https://gist.github.com/t-m0neybagz/19660392736de4ca3ebf6846ae4cb753">Github</a>.</p><h3>Netflix: A Case Study in Ad Targeting</h3><p>Netflix, a well-known streaming service, introduced an <a href="https://about.netflix.com/en/news/announcing-basic-with-ads-us">ad-supported subscription tier</a> in 2022. For simplicity, let’s consider a fictional scenario with just 1,000 users in this ad tier. We <strong>create a table with data on these ad tier Netflix users</strong>, including demographics (age, gender, income percentile) and content viewing preferences (favorite genre, presence of kids profiles).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/793/1*JYupJBlC0pAVeEv1qqEiGA.png" /><figcaption>Netflix Audience Insights</figcaption></figure><h3>Creating and Analyzing Our Source Audience</h3><p>Let’s assume we run a trendy streetwear brand, T-Moneybagz, popular with younger demographics.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/500/1*V6Da4IXaggRFPFZTUkh8PA.png" /></figure><p>With an eye on expanding our marketing efforts during the back-to-school season, we aim to reach new customers through online ad platforms, including Netflix. We <strong>create a source audience using the email addresses of our five highest-spending customers</strong> and share this with Netflix. We then compare our source audience with the Netflix user population.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/793/1*oyHkFKPZMsmFcOW4hjaMtg.png" /><figcaption>Insight Comparisons b/w Source Audience and Netflix User Population</figcaption></figure><p>Analyzing this source audience data alongside the broader Netflix user base, we discern key traits of our top customers: predominantly female, middle-aged, generally higher income, fans of drama and reality TV, and most have kids’ profiles on their Netflix accounts. This insight shapes our approach in targeting similar users on Netflix’s ad platform.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/500/1*NmujPuf5pbTGXzKHXB33Cw.png" /></figure><h3>The Mechanics of Lookalike Audiences</h3><p>Ad platforms leverage various characteristics to determine the commonalities between a source and a lookalike audience. For instance, if Netflix’s lookalike audiences focused solely on age and income, they would align closely with those specific attributes of the source audience.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/568/1*Fg8vi1ERIdN6XMuBzkEbcw.png" /><figcaption>Lookalike Audience Example based on Age and Income Only</figcaption></figure><p>However, ad platforms typically use a more nuanced approach that considers a broader range of characteristics, as we do in this example, <strong>using all five demographic and content viewing data points from our Netflix user table to inform the makeup of our lookalike audience.</strong></p><h3>Selecting the Right Lookalike Audience Type</h3><p>Ad platforms typically offer a choice in the size of a lookalike audience, which affects its similarity to the source audience. Generally, a larger lookalike audience means less similarity, and vice versa. This trade-off is crucial in deciding the optimal audience type.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/500/1*epueE2Fu8zwUO7PS31fzIg.png" /></figure><h3>Creating and Analyzing Our Lookalike Audience</h3><p>The process for creating a lookalike audience is straightforward on most ad platforms:</p><ol><li>Choose the source audience</li><li>Select the audience type (based on size and similarity)</li><li>Pick target geographic locations</li></ol><p>For T-Moneybagz, we use our top spender source audience, opt for a “balanced” size and similarity, and target users from the United States. Once formed, we can compare the lookalike audience’s traits with those of the source audience and the general Netflix user base.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/793/1*mQ9MoRu0QFYENLxGxeWtMg.png" /><figcaption>Insight Comparisons b/w Lookalike Audience, Source Audience, and Netflix User Population</figcaption></figure><p>These insights validate that the lookalike audience characteristics resemble our T-Moneybagz source audience, and this gives us confidence in our ability to effectively target Netflix users in our upcoming ad campaign!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/500/1*_Ek6RtU9NRII-Xr4yfY72A.jpeg" /></figure><h3>Conclusion</h3><p><strong><em>Understanding the Power of Customer Data</em>: </strong>Customer data allows businesses to target potential customers sharing similar traits with their existing customer base more accurately, and it’s important for advertisers to develop a strategy for acquiring and retaining customer data.</p><p><strong><em>Maximizing ROI by Creating Lookalike Audiences</em></strong>: By creating lookalike audiences on ad platforms, advertisers can optimize ad spend and improve their return on investment.</p><p><strong><em>Validate the Accuracy of Lookalike Audiences</em></strong>: Prior to launching campaigns, it’s important to ensure that a lookalike audience closely mirrors the source audience. Ad platforms typically offer insight features to facilitate this comparison.</p><h3>Appendix</h3><p>Complete Python Notebook:</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/f8025d22602daca423ddef522a4c29ba/href">https://medium.com/media/f8025d22602daca423ddef522a4c29ba/href</a></iframe><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=670f3684c4fb" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>