“Half the money I spend on advertising is wasted, the trouble is I don’t know which half.” This statement was made by John Wanamaker at the end of the 19th century, but its essence still lingers today. Marketers understand that all advertising has some impact on sales, but the nut to crack is how to value that impact. Attribution is the exercise that enables companies to calculate the value.
At its core, attribution is simple to explain. Instead of just valuing the last interaction that took place before a sale, attribution allows us to give more or less credit to all touchpoints in the customer journey to purchase.
However, attribution remains an unexplored territory for many companies. Why? There are several major factors that make advertisers procrastinate about moving on from the last click.
- There is limited consensus around attribution (except that last click is no longer a valid model), no best practice and few industry standards since this field is still in its infancy.
- There’s no such thing as a “perfect model,” which makes the perception of attribution daunting.
- The technology that comprises attribution has matured to the extent that it covers advertisers’ digital presence, but many advertisers have offline stores and invest a large portion of their marketing budget in traditional media. Although it’s possible to include these parameters into an attribution analysis, currently there is no self-service or off-the-shelf solution that covers all of them.
- Marketers underestimate the resources needed to succeed with attribution, ranging from single source attribution software to analysts interpreting the data. Also, it’s often the case that people fail to understand that attribution is an iterative process, not a one-time project producing truths that will last forever.
Understanding that attribution is an iterative process is absolutely fundamental, and that’s the reason companies need to get started today in spite of the limitations outlined above. Working with attribution analysis in an iterative manner will take you closer to the truth one step at a time. Here’s how to start.
Getting started with attribution analysis
Begin by gathering your marketing data in an attribution analysis software package such as Google Analytics and compiling conversions and investment levels in distinct line items (often called channel groupings). The more granular these can be, the better the analysis of your customers’ journey – but also the more complex and therefore harder to produce an overview.
A good trade-off between granularity and actionability is to create channel groupings that reflect your optimization scheme. If you invest in search, you’re probably present on different types of keywords, including your brand, generic keywords and special search products (like Google Shopping if you’re a retailer). These all play a different role in the customer journey; generic keywords correspond to the customer’s consideration phase, while Google Shopping and brand terms play an important part at the end of the path to purchase. Instead of looking at search as one channel, segmenting it will give you insights on how each grouping interacts with other channels and ultimately how much value each grouping contributes towards your business.
After splitting up your marketing efforts into separate line items, it’s now time to do that simple thing we mentioned at the beginning – attribution modelling. In other words, you now need to give credit to channels interacting with your customers along the way to making a purchase with you. You won’t find the perfect model for your business right away, but you will be less wrong. So where do you even begin picking a model? Your business goals are always a great starting point. This illustration provides insights into each of the standard models currently being discussed by the industry as ways to achieve business goals:
As you can see, it’s the usual suspects of rule-based models outlined here. You may have heard of data-driven attribution (DDA), which distributes credits based on algorithms taking statistics and probability theory into account to maximize sales. While DDA is a great approach in valuing marketing channels in the customer journey, it’s probably the next step of your efforts. With rule-based models, you have a clearer understanding of how the different channels are valued. Besides, DDA often comes with a cost.
When you’ve decided on a model (don’t worry, your choice is not finite!), your attribution software will enlighten you with recalculated sales levels for each of your marketing channels. Most of your favorite channels will inevitably look worse and the ones you never really thought paid off will suddenly look better. But if you do a quick calculation, the sum of all changes in sales will be zero. Attribution doesn’t create any sales; it’s simply looking at historic data from another perspective. So what’s the point with this exercise?
The point is this is your new way of evaluating marketing efforts. And this will enable you to optimize your presence in these channels based on their attributed performance. A channel with higher attributed sales will make you increase your bids and investment levels in that channel, and vice versa. If your attribution software comes with a bid optimization module for all your channels, you’re lucky – push the button and deploy your attribution model within your bidding strategy. (If you’re not quite so lucky, you’ll need to change your bids manually.)
Bear in mind that the difference between your new attribution model and your old one (yes, last click is also an attribution model) is calculated on the average of each line item. This will be sufficient as a first venture into the world of attribution. When you go deeper, you should invest in that optimization module that will not only make your life easier but will also allow you to be more granular in your attribution efforts. Invest in software that is able to drill down to the level of creative, keyword or the site your advertisement was present on so you can distribute credit accordingly.
How to utilise your data
Once you’ve gathered your data, distilled your marketing efforts into logical, actionable line items on which you are able to conduct your attribution analysis and started to act upon that analysis in your optimization work, then it’s the time to understand what effect this exercise has had on your business. Testing is essential in all marketing efforts in order to distinguish the half of your media budget that really has an impact on your business. Deploying a different attribution model should also be put to the test. However in contrast to a conventional marketing test, this will require changing several variables at the same time, which makes it harder to pinpoint the exact impact that each change has on your business.
One way of dealing with that is having the mindset that what you’re really testing is the whole approach of calculating value across channels. When your test results show you an incremental effect on your sales as an outcome of your attribution model, then keep that model, alter those parameters that seem to have had the most impact on the uplift and test it again. Results show a detrimental effect? Start over, maybe try a different model or significantly alter the one you chose in the beginning. Discouraged about starting with attribution because you’re worried the time and resources invested won’t pay off? Then answer this question, and be honest with yourself: when did you put your current last click attribution model to the test?
This iterative process will not only get you closer to the truth of what marketing channels aid your customers doing business with you, it will also ease the introduction of new dimensions to your attribution efforts. As cross-device reporting becomes standard for your attribution analysis software or when technology evolves to enable the measurement of offline media investments, you will be quick to include this in your next analysis iteration. These innovations are already starting to form, so you’ll be ahead of the competition if you enter this field as of today.
