<?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 Khalil Kafrouni on Medium]]></title>
        <description><![CDATA[Stories by Khalil Kafrouni on Medium]]></description>
        <link>https://medium.com/@khalil.kafrouni?source=rss-25ccfb6fe4c0------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*sp2_5P6zyK6jRMvBdM8iZQ.png</url>
            <title>Stories by Khalil Kafrouni on Medium</title>
            <link>https://medium.com/@khalil.kafrouni?source=rss-25ccfb6fe4c0------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sun, 22 Nov 2026 11:43:22 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@khalil.kafrouni/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[Profitable Marketing: Less Ads, Not More]]></title>
            <link>https://medium.com/@khalil.kafrouni/profitable-marketing-less-ads-not-more-3726112405be?source=rss-25ccfb6fe4c0------2</link>
            <guid isPermaLink="false">https://medium.com/p/3726112405be</guid>
            <category><![CDATA[statistics]]></category>
            <category><![CDATA[marketing-strategies]]></category>
            <category><![CDATA[business]]></category>
            <category><![CDATA[optimization]]></category>
            <category><![CDATA[data-science]]></category>
            <dc:creator><![CDATA[Khalil Kafrouni]]></dc:creator>
            <pubDate>Sat, 25 Nov 2023 23:15:58 GMT</pubDate>
            <atom:updated>2023-11-25T23:35:20.223Z</atom:updated>
            <content:encoded><![CDATA[<p>With the holiday season right around the corner, B2C businesses are readying their marketing engines to be working at full capacity. Creative teams are working day and night to produce new advertising material in anticipation of the festive period. Some campaigns will perform well, some will perform poorly, but the hope is always to strike gold early on and find that formula that will generate a return worth many times the investment. But when do you stop creating and testing new campaigns and settle for the good ones that you have already found? In other words, when does the cost of testing a new ad become higher than the potential profit you would get from it if it performed very well? It’s a delicate dance between risk and reward, and the final decision usually boils down to experience and intuition, but data science can help answer this question.</p><blockquote>TL;DR: To maximize the ROI of a marketing campaign, a good rule of thumb is to spend only a fifth of the budget on testing new ads.</blockquote><p>Suppose that you have a certain budget for social media advertising. To optimize your spending, you would use part of your budget to come up with new ads and test their performance, pick the most successful ones, and then invest the remainder of your budget into them. To figure out the best time to stop testing new campaigns and go with the best one that has been found, a statistical simulation would be useful. For this, we model the ad performance using the Return on Ad Spend (RoAS) metric, which is the revenue earned for every dollar spent on an ad. This is what a typical RoAS distribution might look like:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Ecg3phx9apKT95juGD65Uw.png" /><figcaption>Fig. 1: RoAS Distribution</figcaption></figure><p>In this simulation:</p><ul><li>about 40% of the ads have a RoAS less than 0.5</li><li>20% of the ads would at least break even (RoAS ≥ 1)</li><li>about 2% would have great performance (RoAS ≥ 2)</li></ul><p>With the distribution above, we can simulate what the ROI of different budget allocation strategies would look like. Say we have a budget of $100k for advertising, and that to create and test a new ad it would cost $500, meaning that the total budget allows for the creation of 200 ads at most. Now, how many ads do we create before we stop, pick the best among them, and invest the rest of the budget in it? For this, we run a simulation multiple times where we test every possibility of this stopping point, and then take the average of all the results. Here is what the results of this simulation look like:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*KAEloMTrKB60MAUx0xCQeQ.png" /><figcaption>Fig. 2: Simulation Results</figcaption></figure><p>As seen in the chart above, the peak occurs at around 20% of the budget, much earlier than one would intuitively think. This goes to show the importance of the risk-reward balance, where soon after passing the 20% mark, the prospect of finding a stellar-performing ad is outweighed by two factors:</p><ol><li>The cost of creating and testing a new ad</li><li>The reduction in budget that would otherwise be allocated to a well-performing ad that had already been found</li></ol><h3>Business Impact</h3><blockquote>“All models are wrong, but some are useful.” — George Box</blockquote><p>Although this simulation serves as a good generic guideline, its main goal is highlighting the need to keep the risk-reward balance of all marketing efforts always in check. Generally, extensive testing soon reaches diminishing returns, and the better decision would then be to stick with a good formula that had already been established.</p><p>There are of course multiple factors that come into play; for example, ad performances typically also reach diminishing returns at very high spend levels and/or over extended periods of time. In addition, a good marketing team could potentially find high-performing ads more easily after having tested the waters. Also, there may be certain cases where testing different ads requires extensive resources, or instances where the budget is very limited, and both of these scenarios interfere with the ability to reach statistically significant results, therefore a different approach should be employed. Finally, it is possible that the RoAS for a certain product has an entirely different distribution than the generic one described in this simulation, requiring a specific simulation that accounts for that aspect.</p><p>All in all, as a rule of thumb, companies should proceed with caution when spending beyond a fifth of their ad budget on testing. At the same time, marketing teams should not be blindly incentivized to keep creating new material when that is not what is needed. Even though there might be greener grass on the other side, the cost of getting there might be too high.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3726112405be" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>