Choosing an 1 Attribution Solution a short guide Introduction Online marketing has grown at a fast clip over the years. Recently, however, measurability has led marketers to question the efficiency of digital marketing. Marketers reach target customers in a variety of ways across a variety of channels. Astute marketers have come to realize that simple attribution methodologies such as first click and last click attribution are flawed and result in misallocation of marketing budgets because they provide only partial insight into how customers interact with a brand before converting. Cross-attribution, that is, attribution across on- and offline as an additional tool for optimization, is also gaining in prominence with technological advances. Multi-attribution and cross-attribution, initially explored by only a few innovative companies, has now gained traction among seasoned analytics providers. There are now several new solutions competing with one another and with the existing multi-attribution solutions in offering their own flavor of multi-attribution. With so many options in the market, it is important for firms to understand the key differences between offerings and evaluate them in the context of their own requirements. This report gives you a concise yet
2 comprehensive set of criteria with which you can assess multi-attribution solutions in order to help you determine which solution best meets your needs. What is Multi-Attribution? Marketers use multiple types of media to reach their audience, but not all of them are always accounted for when measuring ROI. According to a Microsoft Atlas Institute study, 93 to 95% of all touch points are ignored when you attribute conversions to the last click. Consider the case of a homeowner trying to buy a piece of custom modern lighting. As a first step, he conducts a Google search for modern lighting and sees SEM PPC ads from several ecommerce companies, including YLighting. He then searches for a YLighting branded term, visits the YLighting website by clicking on an organic result, looks at a few products, and leaves before making a decision. YLighting retargets him a few days later through Display banners, and he comes back to the site, identifies a product to buy, leaves to search for YLighting coupons, comes back through an affiliate site, and eventually completes the transaction. This simple example is a classic case for multi-attribution and is typical behavior in online transactions. Yet most analytics packages use last click attribution by default. In this case, using last click would result the affiliate getting sole credit for the sale. All the assists in the form of organic search, SEO, and retargeting fail to get any credit. In contrast, multi-attribution understands and accounts for the fact that a classic marketing campaign touches potential customers in multiple ways. Multi-attribution tools track all marketing sources that the customer was exposed to before the purchase and apportion credit among all these sources accordingly.
Within multi-attribution there are a few different frameworks for allocating credit: 1. Even Credit simply split the revenue credit for the sale evenly among all known sources. 2. Funnel Position categorize sources as introducers (the first known source in the path to conversion), influencers (sources in the middle of the funnel) and closers (last known source before conversion) and apportion credit among them based on marketers preferences/beliefs on which is more valuable. 3. Algorithmic use sophisticated statistical and predictive models involving regression/hypothesis testing to arrive at a model taking time decay into account. Cross-Attribution Traditionally, marketers siloed their advertising efforts, ignoring any overlaps between their online and offline campaigns. But with ever-increasing interdependency (for example television ads that drive viewers to complete a sale online), it is clear that a holistic approach to attribution is required. With the sheer size of offline advertising budgets TV alone typically comprises 70% of the overall advertising budget for companies with a TV presence even small optimizations can have meaningful impact. Why does offline command the lion s share of budgets? According to Deloitte's fifth edition "State of the Media Democracy" survey, 71% of Americans rate watching TV on any device among their favorite media activities, and 86% stated that TV advertising has the most impact on their buying decisions. Of course, because it is delivered and consumed differently, the methods of tracking, attribution and optimization for offline are different from online. Measuring performance of offline and quantifying its impact have 3
4 been particularly daunting with the limited solutions available in the marketplace. However, Convertro now offers a robust solution that captures the impact of all offline media, including not only television, but also radio, direct mail, print and out-of-home channels, empowering marketers with the tools to make holistic, data-driven decisions. Multi-Attribution Key Selection Criteria As you make the transition to multi-attribution, be sure to ask tough questions in the following areas to ensure that the solution you pick is a good fit for your requirements. Below we ve provided a comprehensive set of questions in each important category that you can use to gauge the quality of the attribution solutions you re evaluating. A Guide to the Selection Process When selecting a multi-attribution analytics solution, we recommend using the below steps as a guide: 1. Identify the pain points and shortcomings of your current analytics implementation. Make a list of weaknesses of your current analytics/marketing measurement system and understand what s preventing you from getting an accurate picture of how your marketing campaigns are performing. Make a list of marketing insights you re trying to get and crosscheck these with your current capabilities. 2. Shortlist the minimum requirements for consideration. Go over your feature wish list and come up with a list of minimum requirements that would need to be met in order to make a switch from your current implementation worth it.
5 3. Research analytics providers and compare them using the framework provided below. Below we ve provided a comprehensive set of questions you should ask when choosing a multi-attribution solution. The analytics solution you choose need not satisfy every single item, but in order for any analytics solution to accommodate your growing needs over the years, it is essential that the solution can answer yes to most of them. 4. Avoid making these common mistakes during vendor selection: Following the lemmings: Don t pick a solution just because everyone else in the industry seems to be picking it. A solution has to fit your needs and provide solutions to your unique problems. The reason most firms go with the majority opinion is that it serves as an endorsement for their decision and helps deflect criticism in the future. By refusing to be a lemming you guarantee that you will make the right choice for your unique situation. Preferring form over substance: People often make decisions based on data presentation without paying close attention to the quality of the underlying data. Choosing form over substance is a definite path to failure. Data presentation is critical if you can t easily understand the data, how will you use it effectively? But poor data quality will render even the best presentation layer useless. Make sure that your analytics package is backed by granular and quality data. Investing in a product vs. a partner: Analytics packages can never be entirely automated. To arrive at actionable, powerful insights you need a partner to guide you through the analysis. Most analytics packages are nothing more than a data dump. To use this data effectively, you need not just another software vendor, but rather a partner who is truly committed to your success and helps you analyze and derive actionable insights from the data for optimization and strategy.
Selecting a product with teaser rates that lock you into higher rates in the future: Don t make the mistake of selecting a product that offers teaser pricing for the first few billing cycles but locks you into steep increases thereafter. Make sure that the pricing is simple and transparent. Buying a product that is too complex and not transparent: A product that is too complex to explain to senior management is more of a black box, and will always be questioned. When you do not understand how a specific recommendation is being made, then it is not easy to trust it to be reliable. The output should take complex problems and illustrate them in a simple manner for all levels of an organization to easily consume and make effective decisions regarding overall marketing mix, with predictive and forecasting capabilities. 5. Evaluate using demos and trial version An important step in the evaluation process is to schedule product demos. Make sure to ask questions and seek clarifications. If possible, use a trial version of the software to get familiar with the feature set, develop consistent KPI s with time-series to accurately compare offerings. Below we ve provided a comprehensive set of questions you should ask when choosing a multi-attribution solution. The analytics solution you choose need not satisfy every single item, but in order for any analytics solution to accommodate your growing needs over the years, it is essential that the solution be able to answer yes to most of them. 6
7 Data integrity: Data is king. We know from experience that data is more persistent than the applications built around it. The application is nothing more than a way to visualize data and although it can and does change over time, quality data is permanent and continues to provide insights far beyond its projected useful life. Does the product account for cookie deletion, use of multiple browsers or devices, and track users across multiple sessions? Can the product track users interactions with your marketing across both online and offline channels? Does the product store data at a granular level including click and order data, or does it store data in an aggregated form? Does the product retain all historical data, or does it truncate or age out data to save on costs? Data is only as good as its ability to identify users uniquely over long periods of time. Robust tracking is critical for getting comprehensive information about campaigns a user is exposed to. Without a robust tracking system that can maintain the unique identity of users over long periods of time, the benefits of multiattribution are severely diminished. Users are often driven to interact with your online marketing after being exposed to offline marketing such as television, radio, print or direct mailers. If your attribution solution can t track these data points onand offline, you ll be missing a big part of the marketing picture and the ability to drill-down into the data which drives your overall performance. Having the most granular level of data means you can always rerun analysis with new models or frameworks to gain additional insight as your business evolves. It s your data, and having all of it helps you run year-over-year and other important comparisons side-by-side. Solutions that truncate data over time hinder your ability to maintain and refer to historical data.
8 Data integrity: Data is king. We know from experience that data is more persistent than the applications built around it. The application is nothing more than a way to visualize data and although it can and does change over time, quality data is permanent and continues to provide insights far beyond its projected useful life. Does the product report on all known traffic sources that influenced conversions, or does it limit the number tracked to an arbitrary number or to a specific cookie duration? Limiting data analyzed to a set number of touch points or a specific cookie window or channel results in losing out information on valuable introducers and influencers in the traffic source funnel. Deep insights: As Avinash Kaushik puts it: a good analytics solution should lead to insights, not just a data dump of information. Does the product give actionable insights into how to allocate your marketing dollars, or is it more of a reporting tool? Does the product result in growth and ROI/ROAS improvement or is it just an added expense? Reporting is different from analytics. A pure data dump is not analytics. The product you choose should provide you with actionable insights and analysis well beyond simple data points. The product should provide quantifiable results and an easy way to verify actual vs. expected lift of recommended changes in a short period of time. Campaign/marketing management: Great insights don t translate into great results if they don t easily drive campaign management systems such as bid management, retargeting, creative optimization, etc. Does the product integrate well with campaign and bid management tools? The product s recommendations should automatically populate PPC bid management tools so you can optimize spend most efficiently.
9 Campaign/marketing management: Great insights don t translate into great results if they don t easily drive campaign management systems such as bid management, retargeting, creative optimization, etc. Does the product integrate with retargeting, creative and landing page optimization tools? Does the product allow you to measure campaign, channel, and vendor performance? Does the product give you insights into accretive vs. cannibalistic sources? Does the product give you insights into the lifetime value of channels based on their ability to drive repeat customers? The product should integrate the results of attribution with optimizing retargeting ads and landing pages. By measuring their relative contribution to your ROI and stacking vendors against one another, you can revisit vendor payment agreements to align with their true contribution. The tool should have built-in forensic analytics to determine which sources are accretive and which are cannibalistic, so you can increase spend on accretive traffic sources and decrease or eliminate cannibalistic sources. Some traffic sources are better at driving repeat customers than others, and the tool you choose should be able to calculate various channels true lifetime value based on such insights, including recommendations on budget allocations with significant time-series indicators. Transparent attribution models: Don t choose a black box model that you can t audit or validate. A good solution is easy to understand and explain. Are the attribution modeling and algorithms transparent and easily verifiable? Transparency and verifiability are keys to trust. When you intend to make business decisions based on attribution data, you need to be able to verify and re-verify it.
10 Transparent attribution models: Don t choose a black box model that you can t audit or validate. A good solution is easy to understand and explain. Is the product simple and easily understandable? All effective models are easy to understand and explain. This becomes especially critical when you have to explain increasing or decreasing marketing spend to executive management based on recommendations from the multi-attribution tool. Fraud detection/forensics: Online fraud is a reality and a good analytics solution should help enforce channel contract guidelines in order to reduce and/or avoid it. Does the product allow you to detect affiliate and/or vendor fraud? Does the product give you a clear indication of overlap among traffic sources driving conversions? The tool should report on fraud to help you make sure you re not overpaying for traffic sources that are stealing credit from others. On average, each user is exposed to 3.3 different sources of marketing before they convert, and having an insight into how these sources overlap can help you optimize spend. For example, a report that details how traffic sources work together in getting a visitor to convert and how often these sources overlap with others vs. operate independently would help derive this kind of insight.
11 Fraud detection/forensics: Online fraud is a reality and a good analytics solution should help enforce channel contract guidelines in order to reduce and/or avoid it. Can the product recognize sources that enhance brand value? Not all marketing has a direct and immediate effect on ROI. Some traffic sources may be brand enhancers that deliver financial gain in the longer run. A good analytics solution is able to detect and report on these sources so you can optimize your mix of brand-enhancing halo marketing to build brand equity over time as well as ROI-based marketing. Ease of deployment: IT resources at most companies are already tightly constrained and don t have the flexibility to allocate dedicated resources for managing web analytics solutions. As such, ease of deployment and maintenance are extremely critical for deriving full benefit from an analytics solution. Is the analytics solution easy to deploy? Is the analytics solution easy to maintain on an ongoing basis when new web pages are deployed and/or changes are made for conversion and event tracking? Does the solution play nicely with existing tag container solutions? Complexity results in deployment mistakes and can result in incorrect and/or missing data. Web pages, business rules and processes constantly change and maintenance of any tags and other mechanisms used for tracking should be simple enough that IT departments aren t overly burdened. Even if you haven t yet implemented a tag container solution, if you choose to use one down the line to make managing your marketing tags easier, you ll want to make sure that the attribution solution you choose can integrate with it using standard methods.
12 Conclusion There are several multi-attribution analytics solutions in the market right now, and each has its own strengths. When choosing among attribution solutions, while there are several selection criteria you should consider, keep in mind that data integrity in particular is key. In order to truly compare the performance of your marketing sources, your solution needs to have a robust tracking mechanism that can account for all customer touch points on- and offline, so that you get the whole picture of how your marketing is performing. As any former student of statistics remembers: given the choice between better data and a better algorithm, always choose better data. If your analysis is based on insufficient or flawed data, your conclusion will inevitably be wrong, regardless of the sophistication of your algorithms or the appeal of your presentation. And of course, the ability to take action on valuable insights about what s working and what isn t is the difference between just another analytics package and driving real incremental profit and revenue. Solutions that integrate into existing marketing management software, provide specific recommendations, and quantify the impact, will provide the highest return on investment.