What led a customer to buy from you?
\nIt’s a simple question that becomes difficult to answer the more you start digging into it.
\nDid they see an Instagram post or TikTok video, search for your product on Google, or open a promotional email?
\nMarketing attribution only gets more complicated the bigger your business becomes, with a website, social media properties, influencer programs, offline interactions, and other touchpoints to consider. And as consumer behavior evolves with the times, it becomes even trickier.
\nWhile it’s not a perfect science, the more marketing attribution data you can see across the customer journey, the better decisions you can make for your business. Ahead, learn the basics of marketing attribution, the different attribution models, and how to track and measure your marketing channels effectively.
\nTable of Contents
\nMarketing attribution is the process of evaluating and tracking the performance of your marketing channels.
\nThe goal of marketing attribution, of course, is to gain a clearer understanding of all the different interactions and touchpoints that customers have with your brand on the path to conversion.
\nIt allows you to credit the channels and specific campaigns that contribute to a conversion, which, in turn, helps you understand how and where to invest your money and attention.
\n![\"A](\"https://cdn.shopify.com/s/files/1/0070/7032/files/attribution-2.jpg?v=1682975644\")
While the marketing attribution definition sounds straightforward in theory, tracking your marketing touchpoints effectively can be quite complicated in practice. As technology progresses and consumer trends change, so do the ways we track attribution. Consider the following factors:
![\"Chart](\"https://cdn.shopify.com/s/files/1/0070/7032/files/marketing-attribution-history_0120bea2-6bd7-4069-a5ac-0ff84dc0ef8b.jpg?v=1682975690\")
Before we talk about how attribution works or the different marketing attribution models, we need to make one thing clear: There’s no such thing as 100% “true” marketing attribution.
\nYou can never fully understand exactly how each marketing touchpoint individually affected each customer journey, even when using the best marketing attribution software. All marketing attribution models and tools are merely an approximation of the real world.
\nThe only accuracy you can strive for is in:
\nUTM parameters are a string of labels beginning with a “?” or “&” that you might find after a URL:
\nwww.yourstore.com?utm_source=facebook&utm_medium=cpc
\nWhile it may look and sound foreign, UTM is a standardized system of tagging in digital marketing. Tags are easy enough to create using Google’s own Campaign URL Builder or a Chrome extension like UTM.io. Google’s Campaign URL Builder will automatically encode special characters such as spaces or question marks that might otherwise break your URL.
\nThere are five standard types of UTM parameters that can be used to describe incoming traffic for analytics tools so they can be grouped, organized, and analyzed in buckets.
\nYou can choose when and how to use them, but be sure to be consistent in your UTM labeling and tracking:
\nYou can also create your own custom UTM parameters to get even more granular with how you bucket your traffic. You could use “utm_season=fall” to track a specific seasonal campaign.
\nAdditionally, you can also use any of the valuetrack parameters to dynamically tag different marketing campaign settings or user attributes. For example, &utm_device={device} would automatically change {device} to identify what browser a user is visiting your site with.
\nHere’s an example of what this looks like. If I wanted to track the traffic and sales from a Google Ads search campaign for winter jackets by targeting the non-branded keyword “winter jackets,” my URL with UTM tracking might look like this:
\nwww.mystore.com?utm_source=google&utm_medium=cpc&utm_campaign=nonbranded
\n%20search%20winter%20jackets&utm_term=winter%20jackets
Breaking that down, each parameter tells you something about the traffic:
\nUTMs help you track your traffic down to specific sources so you can analyze how it performs on a more granular level—but only when you keep the following in mind:
\nWhen properly implemented, you can group and analyze traffic from different sources in Google Analytics and other reporting tools.
\n![\"Screen](\"https://cdn.shopify.com/s/files/1/0070/7032/files/GAsource_medium_2e93fe5b-9e21-4663-a0c8-146054d03235.jpg?v=1682975756\")
Proper UTM tracking is a step in the right direction, but by default, if the same user visits your site on multiple devices, each “visit” will be attributed as a separate user and a separate “journey.”
\nFor example, if a user sees an Instagram Story about a product, they may view the product but not buy it right away. Instead, they may research the product on their phone on the way home, before finally searching for it again on their laptop before bed and converting on a Google Shopping ad.
\nTo get around this and group all behavior coming from the same user, you’ll need to enable User-IDs in Google Analytics and integrate your CRM.
\nUser-IDs in Google Analytics create unique non-PII (not personally identifiable) IDs for each user, which gets included wherever their data is sent from. You can then use the ID to unify interactions across devices, as well as online and offline touchpoints, for each customer.
\nThe ability to turn what may look like many independent user journeys across different devices into a series of interactions from one user with your brand is crucial for painting a clearer picture of how your customers engage with you across different devices and campaigns.
\nDrive your business forward with Shopify’s analytics
\nShopify’s user-friendly reports and analytics capabilities help you make better decisions, faster. Choose from pre-built dashboards and reports or build your own to spot trends, capitalize on opportunities, and supercharge your decision-making.
\nExplore Shopify’s analytics
\n![\"A](\"https://cdn.shopify.com/s/files/1/0070/7032/files/attribution-4.jpg?v=1682975795\")
There are generally considered to be seven different types of marketing attribution models you can choose from based on the goals of your business and where in the funnel you want to place the most value:
Each model is outlined below to help you understand each attribution model across your marketing touchpoints. Note how the exact same customer journey can be interpreted differently depending on the model used.
\n![\"Chart](\"https://cdn.shopify.com/s/files/1/0070/7032/files/last-click_9526187f-3211-409d-ab84-3970aac8831e.jpg?v=1682975844\")
Last-click attribution is the most commonly used model and is the default for most marketing platforms and marketing attribution tools. This single-touch model is useful when you are aggressively trying to convert traffic into customers.
It gives 100% of the conversion credit to the last-clicked ad and corresponding keyword. As such, lower-funnel campaigns, such as branded search or retargeting campaigns, will be given more value while brand awareness and upper-funnel campaigns may receive none.
\n![\"Chart](\"https://cdn.shopify.com/s/files/1/0070/7032/files/first-click.jpg?v=1682975868\")
This single-touch attribution model sees the first touchpoint as the most important, as it gets 100% credit for bringing customers into your funnel in the first place. This is useful when you’re prioritizing spending money on campaigns that build traffic and find new audiences.
It gives all conversion credit to the ad or corresponding keyword that gets the first click. As such, high-value bottom-of-the-funnel activities, like remarketing, are missed in this model, potentially leading to a reduced investment in these efforts that actually drops your overall conversions and top-line revenue.
\n![\"Chart](\"https://cdn.shopify.com/s/files/1/0070/7032/files/last_click_non-direct_d1c2c0be-023f-44a5-8160-8aae4acd027b.jpg?v=1682975889\")
The last non-direct click model gives credit to the last click event before a buyer comes directly to your store to make a purchase. It gives 100% of the credit to the last, non-direct touchpoint.
In the example above, that would be Email.
\nDirect traffic is often from customers who have already made the decision to purchase because of marketing on a different channel. Last non-direct click attribution helps you filter out direct traffic and focus on the last marketing activity before conversion.
\nUsing last non-direct click attribution, you learn that it was actually Email that drove the final conversion. Direct will be an attributed channel in this model when the customers only trackable event is coming directly to your store.
\n![\"Chart](\"https://cdn.shopify.com/s/files/1/0070/7032/files/linear-attribution_00ff69c1-55c3-463b-b921-7d4e8a606dbe.jpg?v=1682975917\")
A linear attribution model distributes conversion credit equally across all clicks on the customer’s path to purchase. This is the simplest form of multi-touch attribution. With this model, you don’t miss out on crediting any interactions. However, it doesn’t tell you exactly what marketing channel had the most impact.
![\"Chart](\"https://cdn.shopify.com/s/files/1/0070/7032/files/time-decay_8c647b8c-aad0-4464-8994-be4060c4e9e6.jpg?v=1682975943\")
The time decay attribution model is similar to last click. However, it also gives some credit to the interactions that led up to the conversion, with more weight given to clicks that happened closer in time to the conversion.
![\"Chart](\"https://cdn.shopify.com/s/files/1/0070/7032/files/position-based_a0b6531d-1acc-46de-b5dd-a1090e5e0eed.jpg?v=1682975977\")
A position-based (or U-shaped) attribution model gives equal weight to the first and last click—each of these interactions gets 40% of the credit. The remaining 20% is spread out across the other clicks in between.
The assumption here, however, is that the first and last click are the most valuable interactions, while there may be campaigns or touchpoints in the middle that also play a significant factor.
\n![\"Chart](\"https://cdn.shopify.com/s/files/1/0070/7032/files/algorithmic_ba9eb2f1-a25d-45fb-81f0-0e1908b8c7c3.jpg?v=1682976004\")
This model is often referred to as custom attribution. When you have enough data available, you can allow machine learning to dictate which touchpoints deserve the most amount of credit in a customer’s journey.
Theoretically, this is the best model, but it relies on having enough historical data for machine learning to distribute weight across different touchpoints.
\nWhen you examine marketing attribution in Google Ads, Facebook Ads, Google Analytics, or Shopify reports, you might notice discrepancies. Which is the source of truth? Technically, all are “correct”—they just view marketing differently. Here’s a primer on how each one works.
\n![\"Screen](\"https://cdn.shopify.com/s/files/1/0070/7032/files/Channel_Performance_3_f1d3375f-40c3-4689-9a50-b01a9bbee3f0.jpg?v=1682976050\")
In Shopify’s Channel Performance report, marketing attribution models provide a holistic understanding of your customers’ journey to conversion. Through the report, you can toggle to the attribution model you prefer:
Shopify’s Marketing Insights tools and reports help you better understand your business. See insights at a glance, analyze marketing channel performance, and choose the attribution model that’s best for your business.
\nGoogle Ads only tracks Google Ads traffic. It doesn’t deduplicate conversions from other advertising campaigns on different platforms because it doesn’t “see” those touchpoints. Instead, it will take credit for any user who touches a Google campaign at any point—even if they later touch Facebook/Instagram, email, or visit your website directly and convert.
\nBy default, the Google Ads attribution window setting shows actions taken within 30 days of clicking on your ads using last-click attribution.
\nThe Meta advertising platform tracks only Facebook ads traffic and interactions (which also includes Meta-owned properties like Instagram).
\nIt also does not deduplicate data from other advertising campaigns on different platforms and will take credit for any user who sees or clicks a Facebook ad within a certain time span, even if they later interact with a Google Ads campaign or email, or visit your website directly and convert.
\nFacebook defaults to last-click attribution with attribution windows of within 24 hours of viewing your ad and within 28 days of clicking your ad.
\nFacebook Ads is the only one of the more detailed advertising platforms that will take credit for users who potentially “see” an ad (even without clicking it) and convert in another way. It’s recommended that you change the settings to be click-based if you’re looking for a better comparison of your results across platforms.
\nGoogle Analytics and other analytics platforms will track clickable actions across different paid and unpaid channels. Generally, analytics platforms can be configured to connect external/offline data sources, User-ID, and/or other web properties that are not directly part of your online store.
\nGoogle Analytics offers a Data Import feature that lets you upload data from other sources so you can analyze it all in Google Analytics. Adding additional data sources and incorporating User-ID are the best ways to include the majority of your customer interactions across platforms in one place.
\nGoogle Analytics will also deduplicate conversions from all channels and will give credit to the last touchpoint in a conversion journey, unless it was a direct visit to your site. In that case, it will give credit to the last non-direct touchpoint.
\nWhile most attribution is click-based, impression-based attribution and reporting is also possible.
\nThink about your own experience. Do you click on every ad that captures your interest? Even when you don’t, those ads can still impact your decision to buy a product in the future.
\nAn ad server allows you to consolidate and deduplicate all your marketing data in one platform, while also giving you access to impressions-level data. This data allows you to more clearly see your customer’s path to purchase and which channels you should invest in.
\nFor example, you may see great performance from your search ads at a click level. However, when you look at impression-level data, you see that those that convert on search were actually exposed to a video ad on YouTube beforehand and later searched for your product on Google.
\nThe Google Marketing Platform is one example of this type of technology, where you can have access to impression-level data for channels such as search, video, display, Gmail-sponsored ads, and some social media platforms.
\n![\"Three](\"https://cdn.shopify.com/s/files/1/0070/7032/files/attribution-1.jpg?v=1682976089\")
Understanding the landscape of marketing attribution, the holes in it, and the different models you can apply are a good first step toward better tracking, cleaner customer databases, and smarter decision making. Despite marketing attribution becoming increasingly complex, there’s no denying the many benefits of marketing attribution.
Marketing attribution can offer invaluable insights about how and where customers are interacting with your brand on their path to making a purchase. This attribution data can help inform marketing campaigns and your overall marketing strategy to maximize ROI.
\nThe most popular attribution models in marketing include:
\nMarketing attribution is important because it allows marketers to measure the effectiveness of their campaigns and track the customer journey. Attribution helps marketers understand which marketing channels and activities are driving conversions and customer engagement, and which are not. This data can be used to make informed decisions about future marketing strategies, allocate budgets more effectively, and optimize campaigns for better performance across your sales cycle.
\narketing attribution is the process of assigning credit for sales or leads to the various touch points in the customer journey. It is the practice of analyzing the impact of each touch point or marketing activity on a customer’s decision to purchase or take action. Different attribution models exist to attribute credit to different parts of the customer journey, depending on your preferences.
\n\n
\n
\n
\n
A new AI service for “parsing” music tracks.
\nThe Mikrotakt neural engine can split any track into vocals, instrumentals, bass, drums, and other audio tracks.
\nIn addition, you can upscale vocal tracks to clean up the voice and adjust the pitch. This feature is ideal for musicians and content makers.
\nThe new Chinese model DeepSeek-V3 becomes a competitor to ChatGPT:
\n– registration requires only your email and password
\n– the neural is quite fast: answers are generated in seconds
\n– Thinking mode, where AI turns into an academic on any topic
\n– the model can read diagrams and generate any texts
\n– files of different sizes can be uploaded
\n– the neuron is good at working with code
\n– absolutely free
\nHailuo AI from China can clone any voice without censorship in a few seconds. Its library includes more than 300 voices in different languages and intonations. The model also allows for full customization: you can change the tone, speed, and timbre and add sound effects.
\nThe Chinese have released a worthy competitor to ChatGPT.
\nThe Qwen2.5-Plus chatbot provides accurate answers, can read documents and images, and is completely free of charge.
\nKrea allows rotating models in its “Photoshop”
\nKrea developers have come up with an incredible feature for their neuro-Photoshop: you can drop any object and edit its position in 3D space.
\nThe Upscayl neuron has received a 16x upscale, so now it can save even the most hopeless photos. The download is free.
\nThe Wegic neural network has become even cooler.
\nNow websites are generated in seconds, and all your edits can be made with simple queries.
\nYou just throw in a promo and you have a perfectly(?) designed website.
\nA working AI meme generator.
\nThe Brainrot neuron follows trends, selects relevant images, and produces memes that are even funny to read. And all this is free.
\nGoogle has released a top free guide to AI agents.
\nIt contains everything: basic knowledge, tools, architecture, training methods, and a step-by-step recipe for creating agents on LangChain and LangGraph.
\nIt is a cool service for learning languages with an AI interlocutor.
\nThe Lingocat neuron imitates the native speaker, gives feedback, and has tasks to improve skills. And all this is free of charge.
\nThe free neural from the Chinese outperforms(?) ChatGPT.
\nA new model called R1 has been released that outperforms the best version of ChatGPT. R1 is twice as powerful as the previous version and it has literally everything you need: programming, math, rewriting, and everything else. It’s worth noting that if you look at the graph of how it “outperforms”, it doesn’t encourage you to switch from OpenAI to the Chinese neural.
\nUniversal AI photo editor.
\nGenerating, replacing, or deleting details, styling, changing the background, and upscaling – and this is not the whole list of functions of the new and free Dzine neural.
\nPrompt generator for ChatGPT
\nPromptEngine works according to the official recommendations of OpenAI. The service also supports other language models: Claude, Gemini, LLama, and Mistral, which can be selected from a list.
\nOpenAI President showed a feature for the o1 model.
\nAccording to Greg Brockman, the o1 model works a bit differently than other types of AI, and to achieve high performance, you need to follow a certain plan in writing promos:
\nthe goal, response format, warning, and context.
\n(The Anatomy of an o1 Prompt attached an example image)
\nChatGPT can now write first.
\nThe neuron will remind you of important things like a real assistant. For example, in the morning, it can provide a ready-made excuse for being late for work, and in the evening, a convincing explanation for a deadline shift.
\nSo far, only 10 regular tasks are available, but by the end of the week, all paid users will receive the new feature.
\nOpenAI offers free tokens for ChatGPT.
\nHowever, there is a caveat: in exchange, they ask for permission to use your promos to train models.
\nThe offer lasts until the end of February. Pro users can get up to 1 million tokens per day, and mini-version users can get up to 10 million!
\nTrump has canceled Biden’s decree on control over AI development.
\nThe document, signed in 2023, required developers to share test results with the government to reduce risks. However, Republicans insisted that this restricted freedom of speech and hindered progress. Now, AI development is unrestricted again.
\nChina will organize a competition between robots and humans
\nIn April, China will host the world’s first half marathon where people will compete with robots. The race will involve 12,000 runners and dozens of androids. Technology versus biology is a real race of the era.
\nAI tokens soar after the Stargate project announcement.
\nOn January 23, Trump announced a $500 billion investment from the private sector in the development of AI infrastructure in the United States. As part of this project, OpenAI, SoftBank, and Oracle will create a joint venture called Stargate. Immediately after this announcement, the AI and big data token market showed significant growth:
\n– AI16Z and AI Rig Complex (ARC) grew by 30% per day
\n– GRIFFAIN and ZEREBRO tokens doubled in price
\n– MetamonkeyAi (MMAI) showed an increase of 152.25%
\n– FLock.io (FLOCK) – by 57%
\nThe smartest o3-mini model will be free!
\nSam Altman said that this neural will be able to solve complex problems in programming, math, and data analysis – all for free.
\nFeatures of o3-mini:
\n– advanced analytical abilities
\n– accelerated data processing speed
\n– ease of use for everyone
\nThe release is planned for the near future.
\nOracle founder supports AI totalitarianism
\nBillionaire Larry Ellison stated that the total surveillance system with artificial intelligence based on cameras, drones, and police recorders will force citizens to “behave in the best way possible”.
\nOracle has recently entered into a partnership with OpenAI and Softbank. Digital dystopia is becoming a reality.
\nThe first Chinese base for robot training was opened in Shanghai
\nThey will learn to help around the house, play football, and assemble cars.
\nThe data obtained in the process will be shared with various companies.
\nCurrently, the center has more than 100 androids and plans to expand to a thousand by 2027.
\nChatGPT can now create any kind of application – a new Canvas feature has been released that runs the code right in the browser.
\n– There are no limits to ideas – we invent and ask AI to draw up a detailed project description.
\n– Copy and paste the request into the chat with the neuron, but add the “/canvas” command at the end.
\n– AI will produce a fully working application in React and HTML in a second – you can test it right away!
\n– There are no limits or restrictions – you can even clone YouTube in your browser.
\nThe feature is free.
\nTop 5 promos for working with any amount of information – the neural will squeeze out all the “water” from the source and give you only concise, verified facts:
\n![\"\"](\"/static/f0c3865203f0eb84451f04f5379021f7/85b06/A-weekly-inspirational-newsletter-1080-x-1080-px-1920-x-800-px-3.png)
![\"\"](\"/static/11ec55989bfb8fbe249559f786b70c9c/85b06/AI-NEWS-2.png)