Every Marketer knows that they need a good handle on marketing attribution, i.e., to be able to discern which efforts are causing which sales. It represents a fundamental concept. You need to understand what effect your efforts are having, to improve what will happen going forward. As with every widespread need, there are inevitably many widespread solutions, all claiming to be the best, the cheapest, and the fastest. Marketing attribution has become a ubiquitous term referring to a lot of things. This guide will walk you through everything you need to know to obtain an effective marketing attribution solution for your business.
On this page:
What is Marketing Attribution?
Marketing attribution is the parsing out of sales into the causes of those sales. The keyword in that definition is causes. The purpose of marketing is to induce more sales, so the cause of sales is what’s important to Marketers. Our very job as Marketers is to cause more sales.
Nevertheless, it’s common for marketing attribution vendors to get lost in the technical details and lose sight of the Marketer’s goal of understanding causes. I’ve seen marketing attribution vendors of various stripes mistakenly do this. More details below.
Benefits from Marketing Attribution
Marketing attribution at its core gives critical feedback to Marketers. By understanding what marketing initiatives are working better than others, Marketers obtain the necessary feedback to rapidly improve campaigns and earn more sales. Marketing attribution solutions alone can’t answer all the questions, but knowing the rank order effectiveness of different campaigns is tremendously helpful for stimulating the creative juices that so many Marketers have in abundance. There are often many good, but conflicting theories about why an ad should be effective. Therefore, knowing which advertisement produced more sales this time helps clarify the circumstances in which one line of reasoning is more dominant and increases productivity.
Increased ROAS
There are two primary ways you can use a good marketing attribution to increase your ROAS. The most obvious path to higher ROAS is to simply shift marketing dollars from medias attributed with fewer sales to the ones attributed with more sales.
The less obvious, but more important path to increasing ROAS, is to focus media attribution on a specific channel, to understand what drives sales from within that channel. In a media channel like TV, there are many decisions to be made regarding stations, dayparts, weekdays, creatives, frequency, etc. Other media channels have a similar array of choices of their own. In essence, there are many more managerial levers within a channel than between channels, and therefore opportunities to improve. More levers to change translates directly into more opportunities to improve your ROAS.
The less obvious, but more important path to increasing ROAS, is to focus media attribution on a specific channel, to understand what drives sales from within that channel. In a media channel like TV, there are many decisions to be made regarding stations, dayparts, weekdays, creatives, frequency, etc. Other media channels have a similar array of choices of their own. In essence, there are many more managerial levers within a channel than between channels, and therefore opportunities to improve. More levers to change translates directly into more opportunities to improve your ROAS.
ROAS Adjacent Marketing Goals
Return on advertising spend is not the only advertising goal, but most other advertising goals represent accomplishments that are meant to contribute to ROAS. For instance, a good creative should grab attention, be memorable, persuasive, and promote brand values, but all those good qualities are “good,” because they contribute to the advertising being more effective at selling more products per dollar spent. Consequently, an accurate marketing attribution can provide valuable feedback to many kindred marketing processes to enhance them. Here are a few related marketing goals that benefit from marketing attribution:
- Business growth from more effective internal investments
- Improved customer reach from using dollars more effectively
- Enhanced creative development due to better performance feedback
- Personal career development derived from a better understanding of what’s working
- Market segment identification deduced from understanding which creatives, geographies, and customer profiles are showing greater response
Marketing Attribution Software vs Marketing Attribution Services
One of the major choices in selecting a marketing attribution solution is that of software versus service. Every marketing attribution provider will use analytic software of some kind, and even the most automated solutions should still be able to resolve some of your custom problems. But there can be a wide spectrum between the least and the most customized solutions. The following represent some of the questions you’ll want to ask to determine what’s the best marketing attribution solution for you.
Pre-Defined Algorithms vs Custom Solutions
How will the marketing attribution solution deal with issues that are likely to vary between your business and others? Marketing attribution isn’t a one size fits all process. There are commonalities, of course, but some typical advertising issues that vary across businesses are: length of the sales cycle, product pricing, repeat business percentage, medias used, geographic distribution of any retail outlets, online vs offline sales, seasonality, and different sales channels like agents, direct, and retail. There are many more, but for illustration purposes, let’s take the first one listed, sales cycle, and think about how that would change the marketing attribution process.
One of the relationships that your marketing attribution will need to account for is the time between the advertisement and the sale. How many of the sales are dispersed over time and for how long? Does the sale happen on the same day as the advertising or weeks later? Part of this will depend upon the media channel and the phase of the sales cycle it catches your prospect at, but another factor is the product you’re selling and the length of its sales cycle. The longer it takes the prospect to make a buying decision, the greater the time delay will be between the advertising and the sale. This means that the marketing attribution models need to account for the time delay between the advertising and the sale in your business. A formula that works well for a quick impulse purchase won’t fit the longer sales cycles of durable products. That’s just one example, but it illustrates how a predefined very automated marketing attribution might not work out for you, if your business has distinctive characteristics.
How "bad data" is Used
There’s no business that doesn’t have some bad data. Your resources are limited, and you have to prioritize. It’s only natural that there will be occasions when important information doesn’t get recorded as thoroughly, or as correctly, as you wish it had. It’s not a question of whether some of your data is incorrect, it’s how the incorrect data will be detected, and consequently dealt with, to arrive at fair estimates.
This is one of the key distinctions between a primarily service driven or a software driven marketing attribution solution. Is anyone checking over the data and looking for signs indicating problems? The more complicated your business is and the more issues you suspect your data may have, the more important data and modeling services will be to your success.
Missing Information: Judgment vs Omission
Missing data can be a thorny problem, and unfortunately, the remedies can range from “bad,” all the way up to “okay”. There’s never a good time for important information to be missing, but nevertheless it happens. A software only marketing attribution will ignore the problem and make no adjustments.
A good analyst can mitigate the damage of missing data when making other estimates, by using projections and averages to fill in the blanks, and even provide estimates of the probable impact.
Price
Price is the primary dimension on which a software focused marketing attribution can come out on top. The expertise and time associated with a custom solution cost money and there’s no getting around that. The question then is whether the additional price of the services will be worth it to you or not. It’s hard to generalize value to all the different situations that the readers of this article could be in, but I can give you some benchmarks. A good marketing attribution done across medias, at the aggregate level, could be expected to improve your ROAS by 10%-15%. That’s a reasonable expectation from simply shifting the media budget across medias, quarters, etc. to the more profitable media investments. If 10% of your media budget is worth more than 5x what the services would cost, then that’d be a great return for you, and you would probably want the full-service solution. It wouldn’t make sense to risk a big media budget on a low-service marketing attribution solution that might give you the wrong answers. If 10% of your media budget is between 2x and 5x of the full-service price, then it comes down to what your other business opportunities are. If you’re not expected to get at least a 2x investment return, then I’d consider a low-cost solution, or pass on media attribution altogether, given that it will also take up your managerial attention.
How Does Marketing Attribution Work?
Let’s walk step by step through what it takes to assess marketing attribution performance.
Step “0”: Plan to do Marketing Attribution
They say that the best time to plant a tree was twenty years ago, but the second-best time is today. That’s the situation most companies find themselves in when it comes to marketing attribution, but like the tree situation you can only start from where you’re at.
Ideally, your marketing agency would spend your money to not only obtain the most sales, but also to maximize learning. Normally media budgets go up and down at the same time, which makes it harder to determine how many sales were caused by each media channel. There’s always some variation, and you can work with that, but if your marketing agency plans ahead, utilizes some experimental design principles, and alternates some of the spending, you can learn a lot more. This can go a long way to being more effective and growing faster.
Step 1: Gather Data
These days a lot of spend is concentrated into Google and Meta so some of your work will be done for you. That said, expect to spend some time collecting and assembling records of your media spend and your sales records into a database that can be analyzed.
Step 2: Correct for Missing and Suspicious Data
Inevitably some of the data will be missing or recorded incorrectly. No, you’re not alone in that. It’s that way to some extent at every company no matter the size or industry. Your marketing attribution vendor will need to spend some time with your team reviewing your data and verifying that everyone agrees that the data streams and the final database look correct.
Step 3: Estimate Attributions
This is the hard part, but not for you. This is the reason you’re paying your marketing attribution vendor.
Step 4: Understand Causes of ROAS Differences
An aggregate level marketing attribution primarily measures the returns on your media channels and gives less information about why the ROAS is what it is. Marketing attribution focused on individual media channels, such as paid search, break the sales down into more actionable media characteristics like campaigns, creatives, match types, bid price, brand name, audiences, and geographies. This is where significant competitive advantages can be found because it provides the understanding to improve marketing performance. Both cross channel and within channel marketing attributions are useful and profitable, but the within channel variety usually provides a deeper understanding, while the cross channel marketing attributions impact more total dollars and provides an overarching framework that even handedly compares channel productivities.
Step 5: Leverage ROAS Learnings
Knowing what to do is the first step, but taking action is what seals the deal. Many modern media buying processes are automated, which is efficient, but it also reduces your control. Nevertheless, even Google’s automated bidding can be guided into giving you an advantage over your competitors. You just have to nudge it in the right direction.
Step 6: Make Step “0” a Reality for Future Quarters
Congratulations, you’ve made it to step 6. Remember step “0” when we talked about how it’d be ideal if your spending had been done with an eye on learning and growing? If you’ve completed a marketing attribution exercise, it’s time to plan future spending to not only grow sales, but to give yourself the information to keep learning and growing. It doesn’t take much to add a little experimental design and variation into your spending patterns, and it can go a long way towards better measurement and understanding what’s causing your sales.
Marketing Attribution Methodologies
There are several different approaches that cast themselves as marketing attribution solutions. The following highlights the main differences between them.
Marketing Attribution vs Multi-Touch Attribution
Muti-touch attribution is often referred to, by its vendors, as marketing attribution. MTA attempts to track all the clicks customers make on digital medias, and sometimes the website, and determine how many online sales should be attributed to each click.
MTA, or what I prefer to call, digital tracking, can be very useful in determining which paid search keywords are getting more clicks per dollar spent, and how frequently those clicks lead to an immediate sale. It’s an important source of information, but it doesn’t give an unbiased answer to what caused sales. All sales that can be tracked to a digital click, are credited to a digital click, and no sale is credited to anything else.
Marketers know that many things influence sales, such as sale prices, product improvements, store openings, offline advertising, customer referrals, product selection, customer service, and product experiences, not to mention uncontrollable factors such as economic swings and pandemics. MTA ignores all that and attributes everything possible to digital media. It’s kind of like a proud father at the track meet whose child came in third. The story he tells back home makes his child out to be the grand champion. Third place is good, but should every online sale be attributed to online media? Of course not, it’s clearly heavily biased, but it’s also conveniently available via Google, Meta, and others, so it gets used. As they say “80% of life is just being there.”
Marketing Attribution vs Marketing Mix Models
Marketing Mix Models also offer an answer to the question of Marketing Attribution. MMM uses historical data to measure changes in advertising, and other factors, and compares that to changes in sales to assess how much each change impacted sales volume. Any change recorded over time can be assessed, so factors such as price changes or non-marketing factors such as the pandemic can also be accounted for. MMM is clearly on a better theoretical footing than MTA, and if done well, it can offer a more honest answer to what caused sales, i.e. marketing attribution.
The biggest limitation of marketing mix models comes down to granularity. Because they use total sales as the benchmark on which to look for impacts, only the large and medium influences can be easily measured. Small impacts tend to be masked by the shadow of the larger impacts, and thus hard to measure. That said, the big impacts are the important ones, and I always operate with at least some type of mix model technology in play, more extensive if you’re spending a lot, or more basic if you’re not.
Marketing Attribution vs A/B Testing
A/B tests are one-off tests that compare two conditions (A and B) under similar circumstances. An example could be increasing radio advertising in one city, but not in a similar city, to measure how sales change with and without the advertising. Presumably the sales in the city with the radio advertising will go up more, or go down less, than in the city without the advertising. If done well, these tests can be reliable measures of the effect of one change, in this example the radio advertising.
While not generally thought of as marketing attribution, they are measuring a single cause of sales and they can be fairly accurate, so you can go reasonably far by doing a lot of them. They offer two primary advantages: 1) limited budgets and 2) control over other factors.
The two major limitations of A/B tests are: 1) everything depends on the two populations being similar, and 2) it only gives you a one-off measurement, and because circumstances can vary, it can sometimes lead you into trouble when you try to generalize the findings to a new situation.
Marketing Attribution vs Artificial Intelligence (AI)
Artificial intelligence is the buzzword of 2023. Chat GPT both excited and alarmed many people when it produced fairly good answers to most questions in a short time, by interpreting information off the internet. This has led many companies to make claims about doing “’FILL IN THE BLANK’ with AI.” Marketing attribution has been no exception to this.
Artificial intelligence is a vague term that could mean a lot of things. One form of AI, is deep neural networks. They´re good at predictions if you have a very large dataset, but poor at predicting using smaller datasets. Another form of AI, are large language models (LLM). They are good at interpreting questions and writing sequences of phrases that could be likely answers. There is no current form of AI that is good at working with small datasets, with the sometimes problematic data streams, that characterize marketing attribution.
Therefore, when a marketing attribution vendor says they are “using AI” it could mean anything, or nothing, and if the explanation goes no deeper than “uses AI,” I can confidently predict it’s more likely to be window dressing than a solid basis for creating a superior marketing attribution.
When Does Marketing Attribution Payoff?
There are three basic requirements for marketing attribution to payoff.
- Quality Solution: The answer must improve your understanding of what’s driving sales.
- Organizational Change: Management must change how money is being spent, to take advantage of the better understanding of what’s driving sales.
- Financial Leverage: The media budget must be big enough so that the improvements in that media performance more than offset the cost of the marketing attribution.
Marketing Attribution: Financial Leverage
The budget requirement is the most straightforward. A good marketing attribution can shift media budgets to your best opportunities, and it will typically increase your overall ROAS by 10-15%. If that’s larger than the cost of doing it, then you’ve checked that box.
Marketing Attribution: Quality Solution
One of the harder problems to solve is identifying a good solution provider. My personal experience is that I began building mix models when I already had 15 years of experience in statistical modeling and about 10 statistics courses under my belt, mostly at the graduate level. Even so, I was lucky to get training in building mix models, some twelve years ago now. The most common statistical and machine learning models, that I’d spent a lot of time working on before that, had the goal of predicting something. The accuracy of those models can be gauged by how closely they predict the target variable using a holdout dataset whose data has been set aside from the modeling just for testing.
Mix models have the goal of comparing different media returns, to optimize across them. That’s a higher bar. If prediction is your only goal, you can eliminate any bad data because it doesn’t help you predict. If two data sources are very correlated so that their effects are hard to disentangle, you can simply use the one that predicts better and exclude the other one. None of that is true for mix models. All medias must be estimated, in order to optimize across them. The medias may have bad data, and they will be correlated. The analyst must use that data and adjust for any biases. Every coefficient matters, because they’ll be compared, rather than just the final prediction. It’s a niche specialty, within the niche of statistical modeling, and you’ll want to look for significant experience within that specific area.
Marketing Attribution: Organizational Change
It would be easy to say something trite here like “Findings are only useful when they’re used,” there, I said it! But the reality is that life is more complicated than that. Beyond the burden of good communication and stakeholder buy-in, there are real operational challenges that come from shifting significant media dollars from one media channel to another.
Let’s say that historically, non-branded paid search (NBPS) advertising has been punching above its weight with high ROI’s, and the business could improve profitability if they could shift money from other media into NBPS. Operationally, NBPS is actually a collection of campaigns, keywords, and audiences. If the lower performing parts of NBPS have the greatest room for budget expansion, an influx of funds could lower the overall performance more than expected. Alternatively, if all the money is put into the best performing keywords the business might bid up the price in the auctions, which could also limit performance.
In order to leverage marketing attribution, it needs to come along with a detailed plan of implementation, and better yet, an agency that will take responsibility for increasing sales from the highest-level budget plans to the lowest-level implementation details. Goal Aligned Media is one such agency. Let’s get started getting you more sales per dollar invested.
Ready to leverage Marketing Attribution?
© COPYRIGHT 2021. ALL RIGHTS RESERVED.