Marketing attribution – different models and how to use them
Marketing attribution seeks to give credit where credit’s due: assigning the value of a sale to the customer touchpoint responsible for it, whether that’s an ad, an email, a pageview or a clicked post. After all, even smaller brands make use of multiple sales channels – so to allocate resources efficiently between them, marketers must...
Marketing attribution seeks to give credit where credit’s due: assigning the value of a sale to the customer touchpoint responsible for it, whether that’s an ad, an email, a pageview or a clicked post. After all, even smaller brands make use of multiple sales channels – so to allocate resources efficiently between them, marketers must know which are really driving ROAS and ROI.
Table of Contents
- What is marketing attribution?
- Why do you need it?
- What makes it difficult?
- Types of models
- When & how to use these models
What is marketing attribution?
Marketing attribution shows marketers which marketing activities – the ads and promotions where prospects experience you – contributed most to a sale, so campaigns can be managed and budgets set in the most efficient way. Sometimes it’s a single event like someone clicking an ad. But more often it’s a combination of touchpoints – emails, ads, pageviews, search, and social along the customer journey. With each touch contributing to the outcome.
There are many models for marketing attribution. Some focus on the top of the funnel, where leads are gathered; some on the bottom, where the hottest prospects are ready to commit; and some are nuanced, seeking to understand how a series of actions delivered the result so marketers can tweak and tune their actions. All have potential – if understood in context.
Why do you need it?
Understanding what drives your sales is more complex than it used to be. Decades ago, it was easy to see what role the mailed-in coupon or the salesman’s doorknock played in getting to a Yes: to make more sales, you just ran more ad pages or employed more door-to-door people. But today’s customers experience your brand through many more channels … and have many more choices.
Without a sophisticated attribution model that shows you which marketing actions led to the sale, it’s hard to calculate the ROI of your marketing efforts and make decisions on those results. Perhaps nobody clicks on your Instagram ads… but what if those entertaining ads are keeping you top-of-mind, increasing clickthrough rates when those same people see your email campaign? Or you decide to cut a content strategy due to low views … and you’re wondering why overall conversion rates suddenly dropped. An attribution model, even in its simpler forms, can help. Below we’ll look at six.
What makes it difficult?
But first, take note of attribution’s complexities. Across channels, sites, devices, and campaigns, tracking the customer journey and assessing each contribution accurately is far from easy. Here’s a list of issues.
- Today’s customer journey is long and diverse. It’s vital to be able to track the full journey and assign value to each touchpoint to see full-funnel impact.
- The not-so-sad demise of third-party cookies makes a set of touches harder to stitch into a sequence. If the customer journey breaks into 2 pieces, your top-funnel campaigns show no results when they should. At the same time, your bottom-funnel campaigns will get all the results, when they shouldn’t.
- Bias in the attribution model also plays a part. Favoring last-touch may ignore important brand-building further up the funnel, or a focus on first-touch can play down the hard work of sales progression later.
- Tendencies towards double-reporting are common, too. If a buyer came into contact with your CRM outreach, web ad campaign, and content marketing at some point on their journey, you can bet three channel managers will all claim it all as their win, meaning sales numbers from the silos will look a lot more optimistic than they should.
- Comparing and acting on attribution data is difficult too. Every advertising channel uses its own attribution model – so if the same metrics are calculated in different ways, how can you compare this data?
- Above all, turning attribution data into insight can be tough. Without integrations that connect data from different sources and easy-to-understand dashboards that present large volumes of data graphically, key findings can be missed or misunderstood.
The good news: all these difficulties can be answered. Let’s see how those different models address them.
Types of models
Most marketing attribution models fall into one of two camps: traditional or modern. You can guess which are designed for simpler customer journeys and which for more complex ones.
Traditional models
First-touch attribution models
A first-touch attribution model (FTA) looks at the top of the sales funnel: the broad brim where large numbers of suspects and prospects are gathered, often by content marketing and brand advertising that introduces your products and services. Looking only at the initial interaction, first-touch models are simple and straightforward: all the sales value goes to the very first click in the customer journey.
Of course, by assigning credit for 100% of a sale to the very first touchpoint, this model misses a lot of detail. That’s why it’s mostly used for brand awareness metrics, when you’re trying to assess which channel attracts the most new customers. That’s not to say it’s useless – but it is limited in scope, and not helpful if you’re looking for insights across many channels and multi-touch customer journeys.
Last-touch attribution models
LTAs do the opposite: only the last customer touchpoint gets credited with winning the sale. It counts the link in your email that went straight to the purchase page, or the PPC offer with a discount code that was used in the next click. It’s a pure-play BOFU tactic, and gives a useful sense of how effective your closing is. 10,000 ad impressions had a clickthrough-to-sale of5%? Sounds great.
Of course, last-touch ignores all branding, presales, and market development efforts that prime the customer for that purchase decision – and without those, sales might nosedive. So while it’s useful for straightforward metrics, like FTA and LTA, should never be used as your only model.
Linear attribution models
Linear AMs are a halfway house between simple FTA / LTA and newer models. They look at the customer journey as a whole – all the touchpoints, every interaction – and assign credit evenly along the sequence, as if the sales process was as simple and regular as rungs on a ladder. If you use multi-channel marketing strategies, it gives each channel a fair chance to have its voice heard at budget time.
You can guess the problem, though: with a linear approach, all touchpoints are credited equally. And not all interactions have equal value. Is the download page for a whitepaper precisely as valuable as a customer review on social media? And what’s more, this method assumes you’ve mapped every point on the customer journey into your data model – and real journeys today aren’t that straightforward.
So where do modern models differ? Let’s switch to them.
Modern models
Multi-touch attribution models
Multi-touch models (MTAs) are a great step up in fixing the problems with linear AMs. By assigning different “weights” of credit to each interaction, marketers can customize the model in ways that bring it closer to reality. And when that weighting is driven by AI rather than manual labor, the advantages multiply.
For a given customer journey, an AI-assisted MTA model weights each interaction based on what actually happened – a longer dwell time on a web page, for example, or a full-page scroll instead of a half-page one (both suggesting the customer is more interested). This may mean the model gets weighted towards first and last touches (known as “position-based” or “U-shaped” attribution) or controls for “time decay”, where interactions closer to the sale are weighted heavier … whatever presents a more realistic picture of the customer’s behavior and intentions.
While flexible and customizable, multi-touch models still have issues. They’re still based on easily measurable interactions like a click; even with a complete view of all relevant data, the model may not reflect the totality of the customer journey. So while MTAs offer considerable optimization potential and flexibility, it’s still possible to go better.
Unified Marketing Measurement models
With UMM, marketers can combine the advantages of several models in one, choosing the most apt techniques for the task at hand. That could mean going MTA-style to work out the effectiveness of clicks, swinging in FTA and LTA methods for simpler campaign analysis, and using Marketing Mix Modeling (below) to make sure offline factors like the seasonal sales uptick in Q4 are given the right weight at the right time.
That makes UMM terrific if your marketing spans both offline and online. Of course, it’s more complex to manage, especially since data from offline sources can be harder to collect and integrate. And market dynamics can make your assumptions obsolete fast. So UMM, however, still has challenges – many of which can be overcome with a dose of AI.
Marketing Mix Modeling
MMM isn’t a set model as such, but a set of statistical modeling techniques that analyze data for larger trends and patterns that influence your sales. For instance, MMM could show if (or not!) a 3-month campaign of TV ads caused an increase in overall sales during that period. It’s all about the correlations: was there a relationship between Event A and Outcome B? If there was, MMM can track it down.
This makes MMM incredibly powerful. Because it’s looking for actual relationships between marketing activities and sales outcomes, there doesn’t need to be a straightforward connection between interaction and purchase; the model can infer that connection instead, such as when it sees a correlation between increased TV spend and higher in-store sales figures. And if the correlation is strong – perhaps between PPC adspend and same-week unit sales – you can use it to optimize your media spend, making sure you get maximum return on your marketing investments.
When & how to use these models
It’s time to summarize marketing attribution models. FTA and LTA shine for simpler customer journeys where sales are closely connected to top-of-funnel or bottom-of-funnel activities. Linear and multi-touch models are for longer and more complex journeys, where each touchpoint contributes to the whole. While UMM and MMM deal with large volumes of data from diverse sources and apply statistical analysis techniques to work out what led to what.
And their use cases? Obviously there are exceptions and special scenarios. But in general, limit first-touch models to brand advertising when you’re first reaching out to your market; adopt last-touch models when almost all of your marketing spend goes on direct response channels like PPC ads and email offers.
Linear attribution models work best when your marketing budget is evenly spread across several channels and you want to give each a fair chance in your results metrics. If your customer journey spans many channels but you’re certain of what the main touchpoints are and how each affects outcomes, multi-touch attribution will make your modeling more precise; it has uses if your sales cycles are long or your media mix is complex.
If you’re interested in seeing correlations and insights in your data extracted with statistical methods, it’s time for MMM: Marketing Mix Modeling. With the right integrations to bring together offline activity, online ads, web content, search keywords, social media posts, and more, it’s a model capable of doing your marketing attribution hands-free, drawing information directly from the datasets and presenting it to you in easy-to-understand charts and graphs, even optimizing your campaigns and budgets for you.
And if your marketing covers both tactical drive-to-sales and pure brand-building – combining (say) TV ads and web content views with PPC clickthroughs and email CTAs – move towards a Unified Marketing Measurement framework to see how the soft branding efforts affect the hard outcomes. UMM with AI assistance (as with Billy Grace!) can assign weightings to different parts of the customer journey to make sure every channel, campaign, or advertisement gets the appropriate credit.
That’s what we do here at Billy Grace: giving you a complete view of your market data, and offering insights and optimizations that make the most of your resources. Ready to talk? https://www.billygrace.com/book-a-demo/
