Marketing is a vast and often confusing field drowning in data. There seem to be a million different methods, tools, and strategies for connecting with customers. It is hard to know which combination of touch points to choose for optimization. Manual trial and error helps, but it would be humanly impossible to test every possible variation. This would require time you don’t have, waste resources you could use elsewhere, and erode your brand image. Marketing attribution models provide a way to wrangle the beast.
Marketing attribution is a way of quantifying customer interactions with your brand and developing reasoned conclusions to guide your efforts. You look at which of your marketing strategies are working and how your customers’ actions form a pattern. Once a pattern is recognized, you create a plan that leverages that information to convert more sales.
Having a multi-touch attribution strategy will help you identify which tactics bring you the most value. According to a recent CMO Survey, marketing analytics is one of the top five important knowledge assets. It is also one of the top places where businesses of all sizes are devoting their budgets. Unfortunately, despite this focus, only a small percentage of marketers measure performance quantitatively.
Attribution is a “must” because it allows you to properly give credit to ALL the marketing touchpoints that contribute to a customer journey. This carries with it certain benefits.
Multi-touch attribution does not have to be confined to the digital realm. It can include your offline marketing efforts, traditional advertising, as well as non-marketing related variables. For instance, you may want to study if the weather or a particular sales process has had an impact on the effectiveness of your marketing and customer acquisition.
Adaptive marketing is a combination of marketing attribution and mix modeling. With the right integration and tools, you could start optimizing and tailoring (adapting) your marketing approach in real-time based on recorded data.
Your attribution modeling will change the way you invest across your suite of marketing efforts. You can manually or in some cases automatically update your media buying so that your ads are as effective as possible.
Multi-touch attribution does not have to measure the full timeline of your customer buying journey. You may want to measure marketing effectiveness by attributing touch points specifically caused by a rebranding , new product launch, or a new marketing message.
While you may think that you need an extensive amount of data to get started with marketing attribution, you only need to define your touch points. You can still infer conclusions. However, more massive data sets are more appropriate for more significant decisions. It’s important to pick an attribution strategy that is appropriate for your current business maturity.
Marketing attribution does not have to look at a single data set. You can also use it to establish a baseline that predicts purchases based on different touchpoints, then measure the effect of changing a variable. Variables that result in positive change are adopted into the evolving model.
Now that you know what marketing attribution does, let’s talk about how attribution modeling works.
Overall, multi-touch attribution follows a set of defined rules to assign value to one or many customer interactions. You could obtain information by polling every customer, but that would be time-consuming, and only a fraction of your total customers will respond. Your accuracy would be impacted by how well your customers remember all their exact interactions that led to the purchase.
Attribution models streamline all of this. Ideally, you will apply several models before deciding where and how to invest your marketing dollars. The best model is the one that provides you the greatest return for your investment. In this section, we will go over several models and explain which touch points carry the most weight in those strategies.
As a note, attribution is only as effective as the data you use. Because of this, you need to audit attribution data for integrity and accuracy. In general, think of it as Garbage in, Garbage out or GIGO. If you put garbage data into your model, you will get garbage conclusions.
Garbage data is any information that is incorrect, missing, duplicated, or obsolete. It’s essential to find the root causes of bad data so that you can change processes to prevent it. If the historical data cannot be repaired, it is better to throw these figures out or delete them. Leaving this information in your dataset will skew results and lead to conclusions that you cannot trust.
The First Interaction attribution model puts all of the weight on the first touch-point your customer has with your brand, regardless of how long it has been since he or she first interacted.
Example: The prospect sees a TV advertisement and visits your website directly via a tracking URL. After visiting the web property they decide to leave because they are not ready to buy. Later, after significant research on your competitors, they visit your domain directly and become a customer. In this model 100% of the credit will be given to the TV advertisement that generated that first interaction.
The Last Ad Click attribution model puts all the weight on your customer’s last paid ad touch-point. It operates on the assumption that your most recent paid ad was your most productive in generating desired prospective customer activity.
Example: A prospect hears a radio advertisement and later that day visits your domain directly via a tracking URL. As a result, they get added to your Google re-marketing list and later get presented with a paid Google re-marketing Ad. They click that link and visit your property for the second time and become a customer. In this model 100% of the credit would be given to the second Google re-marketing ad click.
You could also choose a Last Interaction attribution model. Under this method, you would look at each customer’s last interaction before buying. It has a similar premise to the Last Ad Click marketing attribution model, but this model holds that your customer’s last interaction with your brand was compelling enough to lead to a conversion, so it doesn’t matter if it was an ad or not.
Example: A prospect does an organic search and visits a blog page on your site. They later hear a radio advertisement, which served as a reminder that they were still interested in your service. As a result, they visit a unique tracking URL called out in the radio ad and become a customer. In this model 100% of the credit is given to the radio advertisement.
For lead generation companies with long customer funnels like education or mortgage, a click/visit interaction is not as valuable as a lead submission. It is also common for these type of companies to receive several leads for the same prospect over time before converting them. In this model, full credit would be given to the interaction that generated the first lead submission prior to becoming a customer.
Example: A prospect clicks a Facebook ad related to your product and does some initial research. Later they receive a Google re-marketing ad via display network and they click and end up submitting a lead form. They speak with your sales rep but aren’t ready to purchase. Later they receive an email from a nurturing campaign which they click, submit a new lead, speak to sales, and become a customer. 100% credit would be given to the Google re-marketing display ad that generated the first lead.
This model is similar to first lead attribution and also applies to lead generation companies. However, in this model credit would be given to the interaction that generated the last lead submission prior to be becoming a customer.
Example: A prospect clicks a Google Display network banner ad and visits your landing page where they submit a lead. After submitting the lead the prospect realizes they are not quite ready to talk to a sales person so they don’t answer any email or phone calls. A few months later the time is right and they perform a voice search “Hey Google! What is XYZ Company’s website”. This triggers a Google Brand Paid Search ad, which they click, submit a lead, and become a customer. 100% credit would be given to the Google Brand Paid Search ad that generated the last lead prior to them becoming a customer.
Some attribution models put all the emphasis on your customer’s Last Non-Direct Click instead. Basically, this weighting system ignores any direct interaction your customer has with your site. It assumes that his or her last non-direct click (the one that did not include typing your website URL) was the most convincing.
Example: A prospect interested in a remortgage starts by using Cortana Bing voice search on their laptop for “mortgage companies near me”. They click your organic search link. Later they talk to their friend about remortgaging their house and the friend recommends your company (a customer referral). The prospect visits the domain directly (by typing the domain in the browser) and starts the remortgaging process. 100% of the credit would be given to Bing Non-Brand Organic Search.
So far, each attribution model we covered involved assigning 100% of the credit to a single touch-point. There are other ways to evaluate your customer conversion using fractional weighting across multiple touchpoints. One option is to use a Linear attribution model. This system applies equal weighting to each interaction your customer takes before buying. If there are 5 interactions, the weighting on each one would be 20%. You can also apply this fractional model to only the interactions that generated leads.
Example: A prospect visits your competitor’s website and is interested in their services. They check out reviews for your competitor and subsequently find your Yelp Reviews page. They click a link on your Yelp reviews page and visit your site (1. Yelp Organic Reviews) . Later they continue their research and visit your page after typing your company name into Google (2. Google Organic Brand Search). Lastly, they see your YouTube advertisement (3. Paid YouTube Video) which re-sparks their interest and they click the ad, visit the discount page, and make a purchase. 33.3% credit for the customer acquisition would be given to each of the three interactions.
There is also a Position Based attribution model that is sometimes referred to as the u-shaped model. This one distributes the relative importance over each touch-point, but not evenly. The first interaction and the last interaction each receive 40% of the total weighting. You would distribute the remaining 20% across any other prospect interactions that led to a sale. You can also apply this model to only the interactions that generated leads versus all interactions.
Example: A prospect researching services performs a Google search, visits your page and reads about your services (1. Organic Google Non-Brand Search). A few weeks later they are on Facebook and click your re-marketing ad where they submit a lead, and do not convert (2. Paid Social Facebook Re-Marketing). A month later they feel like it may be time to buy so they search your brand name on Bing and click a paid brand ad where they research more about your services (3. Paid Search Bing Brand). Lastly, a week later they visit your domain directly, submit a lead, and become a customer (4. Organic Web Direct). In this model 40% credit would be given to the first Organic Google Non-Brand Search interaction and the last Organic Web Direct interaction. The two interactions between the first and last would receive 10% each (splitting the remaining 20%).
You may want to consider the Time Decay attribution model as well. This system applies unequal weighting to all customer touch points. The most recent interactions have the most weighting, while the oldest touch points have the least. Exactly how much credit you apply to each touchpoint will depend on your industry, your business, and what you already know impacts your business most. If you are unsure how to distribute your touchpoint weighting, start with a balance like 10:20:30:40 and tweak from there. You can also apply this model to only the interactions that resulted in lead submissions.
Example: A prospect is listening to Pandora while on a jog and hears your advertisement. They click and read about a specific service offering you have (1. Paid Pandora) . Later they see a Facebook re-marketing ad which entices them to click and continue more research (2. Paid Facebook Re-marketing). A week later they visit your Yelp Reviews page to see what others are saying about you. This results in a click through to your site for even more research (3. Organic Yelp Reviews). A week goes by and they visit your Organic Instagram page, click through, and purchase your services (4. Organic Instagram). Credit for the purchase would be distributed by giving 10% to Paid Pandora, 20% Paid Facebook Re-marketing, 30% Organic Yelp Reviews, and 40% Organic Instagram.
For many businesses it is best to create your own attribution model using a combination of these explained approaches. A custom model is usually best established after you have implemented other models and learned from them. This allows you to optimize into a custom model that provides the absolute best returns.
For a lead generation company that often experiences multiple leads prior to a purchase, they could create a custom model that gives equal (linear) weight to specific touchpoints they believe are most significant in each phase of the customer journey. A common example would be giving 25% credit to the First Click/Visit Interaction (Awareness Top Funnel), 25% to the First Lead Interaction (Developing Interest Upper Middle Funnel), 25% to the Last Lead Interaction (Serious Interest Lower Middle Funnel), and 25% to the last interaction just prior the actual purchasing decision (Closed/Won Bottom Funnel). Any touchpoints between these major funnel stages would not be given any credit in the model.
Example: A prospect watching daytime TV sees a commercial and then visits a custom tracking domain to research (1. Paid Television Commercial). After, they have two more search interactions (2. Paid Google Non-Brand, 3. Organic Bing Non-Brand and eventually click a search result on Google and submit a lead (4. Organic Google Non-Brand). A sales rep calls them, but they decide not to proceed for financial reasons. In the meantime, they check out your reviews and your website (5. Organic Yelp Reviews, 6. Organic Web Direct). Later their financial situation improves and they click a Google Display Re-marketing ad where they submit a new lead (7. Paid Google Display Re-marketing). They speak with sales and schedule an onsite appointment. During that waiting period they view a YouTube Re-marketing video that re-affirms their interest (8. Paid YouTube Re-marketing). They show up to the scheduled appointment and become a customer.
In this custom model credit for the customer would be 25% to first interaction Paid Television, 25% to first lead interaction Organic Google Non-Brand, 25% to last least interaction Paid Google Display Re-marketing, and 25% to last interaction Paid YouTube Re-marketing.
In today’s hyper-connected world, your customers have a million different ways to connect with your brand. They see your ads, they interact with your social media profiles, and they see posts from influencers (or happy customers). This goes on and on. Marketing attribution helps you see these efforts as part of a holistic approach you can execute on. The analysis assigns values to individual efforts that are not readily apparent or easily quantifiable with the human eye. It also puts emphasis on the order of those interactions as opposed to a simple correlation.
Marketing attribution might sound complicated – but don’t let that fool you into thinking that it is something that only big business can do. Having a marketing attribution strategy is essential for effective customer acquisition regardless of company size. Even small and midsize enterprises will see benefit in attribution modeling. They should choose an attribution path that is appropriate for their current level of marketing, data maturity, and budget.
If you need helping building a data foundation and choosing the best attribution model for your business, contact us today.