How can you attribute conversions to marketing channels?

The attribution model you choose is the formula to tell you which steps were most important in leading to the desired action. I'll explain: Before taking a desired action, your user took a few steps. Whether they heard of your company from a friend or found it through Google, that user took steps that led them to the desired action. Sometimes, the users take multiple steps, and the ability to retrace their steps is crucial in recreating that action for another user.

Attribution model is an algorithm to assign credit for a success event and this success event could be anything you or your organization want measure. The model piece of this algorithm speaks about how this credit should be assigned. A real world example of this would be, who gets the credit for GenAi? 1. Would it be Sam Altman who made ChatGPT possible and revolutionized this sector? 2. or would it be Charles Babbage who is credited for modern day computers? 3. or would it be people in the middle like Steve Jobs, Bill Gates, Mike Dell etc. who made computers a house hold thing? Depending upon who you want to give credit to, how long back you want to go in time and what is your metric, tells about your attribution model.

Big giant ad platforms have their own definition of attribution models that you can set but that will tell you different stories in their language. To make it simple as a center point measurement, the mobile industry is relying on the MMP data and popular model is the last click that giving all the credit to the last touchpoint, (Channels) after converted as sale. Eventually, it's all depend on your business's shape and size to define the relevant model.

Attribution models play a crucial role in marketing analytics by offering insights into the effectiveness of various marketing channels in driving conversions. By identifying which channels contribute most significantly to conversions, businesses can allocate resources more efficiently, focusing on high-performing campaigns and adjusting or improving underperforming ones. Furthermore, attribution models provide valuable customer behavior, preferences, and needs data, empowering marketers to craft more targeted and personalized marketing messages. This strategic use of attribution models enhances marketing ROI and ensures that marketing efforts are aligned with consumer behavior and business objectives.

Simply saying, attribution models help businesses to understand what works and what doesn't in their ads investment. It's not always FB or Google are the winners, it could be other channels, local ad network, direct deal, affiliate, progrmmatic or influencer, MMP play a key role to measure efficiency and effficetiness and helping marketers to make smarter decisions to grow their business.

Given the prominence of 'last-touch attribution', I often feel like the idea of 'attribution' can be linked to a sales mindset = who closed the deal? The most common attribution models - which should not even be called 'models' - are just basic rules. First touch, last touch, decay etc. They are easy to apply, scale and understand. They can be of value for small or very strong e-commerce businesses. But mostly they are misleading as they are rather arbitrary and can be gamed too easily. Even algorithmic attribution models have serious flaws, particularly when based on web cookies (a dying species). If one has access to ID data, it's much better to consider experiments to assess the effectiveness of digital ads.

Attribution models include First-Touch, Last-Touch, Linear, Time-Decay, and Position-Based. First-Touch attributes all credit to the initial channel that directed the customer to the site, which is ideal for assessing brand awareness. Last-Touch credits the final "touch" before conversion, which helps evaluate sales performance. Linear distributes equal credit across all channels, which is suitable for understanding overall engagement. Time-Decay allocates more credit to channels closer to conversion, emphasizing recency. Position-based gives more weight to the first and last channels, with less to intermediate ones, balancing the measurement of awareness and conversion effectiveness. Each model serves different goals and data insights.

It depend on many factors, media investment, platform's capability, messaging strategy, creative asset, product's ticket size etc.. and there is no right or wrong answer to decide Last Touch point is better than Linear for example.

If your goal is to optimize media investments and increase investment in the channels that historically have driven more conversions, remember that giving conversion credit to more than one channel will result in more credits than actual conversions. Use the attribution model to map typical customer journeys to uncover the channels with highest customer touches and keep that journey in mind when rebalancing your media mix.

It depend on many factors, media investment, platform's capability, messaging strategy, creative asset, product's ticket size etc.. and there is no right or wrong answer to decide Last Touch point is better than Linear for example.

Attribution isn't solely the responsibility of one team; it requires cross-functional support to optimize its effectiveness. By integrating additional data sources like BI or other analytical data, a more holistic understanding of the customer journey and conversion process to understand where the gap is coming from. Again, all tools have their own definition in measurement so it will not be the same all the times.

Accept it, it will never be perfect! That doesn't mean you don't have to optimize it. It is an ongoing task that will need adjustments and refinements with time. Where to start you may ask? 1. Start with assessing the data being captured: With my experience, event the top tier analytics tools are also not immune to regulations and privacy initiatives. Make sure the data you need for attribution is being captured. 2. Set a baseline attribution model that fits your business model. Example: For a content page the model might be to just drive visitors and engage them. 3. Keep experimenting with you business stakeholders and have an open mind as the requirements can change depending upon which channel stakeholder you are working with.

When looking to change attribution models when business needs change, pay attention if the new model highlights dramatically different winners than the old model. Be prepared to explain any differences to leadership so they have appropriate context.

From my experience the most important thing after setting up the model is to have an active monitoring setup so that you can be aware about: 1. Any changes in your internal processes/systems: For example, a new joiner takes over the project from someone and ignores the analytics aspect of the campaign and goes for a launch or your tracking tags fall off the site. No matter how sophisticated your attribution model is, it will fail because it will never get the required data for channel identification. 2. Privacy regulations: Marketing has become a highly regulated space where even the 1st party data collection is getting impacted. Try to evaluate if your model depends on a restricted data point and find alternative data points.