Online behavioral data usually has quality issues
As online behavioral data rapidly becomes a remarkably more important source of customer insight, it is rarely treated as a real asset compared to many other data sources, except in digitally born companies. Due to parallel shifts in business priority and fundamental changes in online behavioral data collection, manipulation and usage, it is time to start treating the online touchpoints more seriously as a data source. Cookies becoming of less importance or even obsolete to some degree, growing privacy concerns and control mechanisms within browsers or applications, should be an alert sign for businesses wanting to compete also in a digitalizing business environment.
In this blogpost, you will familiarize yourself with:
- Reasons why online behavioral data faces quality issues.
- How online behavioral data collection is changing at the moment.
- Solutions and concrete actions on how to improve online behavior data quality and be equipped for the changing environment.
To name the biggest challenge for poor quality across online behavioral data sets, it is the lack of clear ownership in terms of treating online behavioral data as a proper data-asset. In addition to lack of ownership, online data collection has been strongly driven by other parties than the Data Controller or Marketer themselves. Instead of having full control, like the legal term Data Controller would indicate, different agencies with no real responsibility at stake have been in the driver’s seat. This usually leads to websites looking like the Wild West, where data is collected even in worst cases for other parties’ own purposes, without the Data Controller having any proper data collection inventory or necessarily noticing how their users’ data is misused.
One common pitfall is also focusing on quantity instead of quality. It is tempting not to define and prioritize the KPIs or clear action oriented use cases in advance, and instead try to collect all kinds of data that might be needed one day. The outcome is easily a set up that is both complex and contains many data quality issues, and thus it is not used to derive new insights. The data collection should be done in steps based on well defined KPIs, and after each step the quality should be validated based on actively using each data point.
To build an actionable marketing data-asset, online behavioral data is discussed in particular detail, since that is in our experience an area that will take time to be on a level as a data-asset comparable to customer data, transactional data or contact data, to mention a few common data sources at stake.
Digital analytics is not an exact science, instead it is a collection of definitions and conventions
Investments into digital advertising platforms have been growing rapidly during the past two decades, foremostly in performance driven channels. Growth has partly been fueled by the misconception or by an argument that everything is exactly measurable.
That argument might have some truth, but it is mostly misunderstood. Firstly the ability to measure is fading steadily, and secondly digital analytics has always been a form of art, rather than science. The exact measurability is at the end only a collection of rules, which affects everything in the range of basic KPIs and metrics, reports and action oriented use cases built on online behavioral data. But for digital analytics, the definitions have been to a large extent done by measurement tool providers, rather than the Data Controller themself.
Despite digital channels already for quite some time being the preferred channels for either customers or potential customers to interact with companies, digital analytics has not (except for digitally born companies) picked up in importance as a discipline in itself, or as part of traditional business intelligence. A very common way to place web or digital analysts, if a company has invested in hiring ones, is under marketing or communications departments, which often dictate their focus on either marketing or being statisticians for website visits. Rarely digital analytics is combined with identified customer data to produce real business insights, despite digital services being preferred channels by customers. Could you see a CFO or CXO placing the business intelligence function similarly?
Defining and agreeing on relevant KPIs, which drive your business, is not actually as easy as one might think with an overload of possible data points available. Digital analytics platforms have enabled it to get easily started with collecting and digesting data, but at the same time they might make you blind without paying proper attention. The devil is in the details, especially when it comes to metrics provided out of the box, such as sessions, unique visitor, time spent on site, traffic source, bounce rate, conversion rate and so on. Providing these easily readable metrics, there have already been many decisions made to calculate them.
Implementing a web-analytics platform with out-of-the-box settings has been made rather easy, and immediately metrics defined by tools, but not derived from business needs, are made available to an overwhelming extent. For businesses and organizations not born digital, this illusion of ease has serious implications. Vanity metrics, with little relevance for business impact, distract from thorough thinking and trying to identify what kind of behavior drives business outcomes, both in the very short and long term.
Organizations working with multiple brands, web domains or businesses spread across different geographical regions, face the most severe challenges. Since digital analytics rarely is organized in a co-ordinated and systematic manner, it also reflects data quality and compatibility across different business entities or an IT driven systems landscape. In the next chapter I have collected four key areas to focus on, increasing the value and quality of online behavioral data.
Key actions to improve digital analytics capabilities
As the need for understanding digital services usage grows at a fast pace, so should investments too into analytical capabilities and skills. To avoid many pitfalls described earlier and to gain new customer insight, digital services should not be looked at in isolation from other channels, data sources and analyst people skills across the organization.
Before jumping into details on how to improve online behavioral data quality and digital analytics capabilities, it is good to set the scene of management expectations for progress, either if you are building fully in-housed competencies or using consultants. Capturing of online behavioral data logic is moving steadily from services’ front end to back end applications and there these applications have also a larger responsibility to provide event flows as they occur, which is a totally different way of working as it used to be. Over time this for sure will change the way how online services are built, applying a data collection by design principle rather than following a waterfall model with data collection as a separate track usually closer to the launch of a service.
Since developer resources are mostly scarce and backlogs full, the ability to adapt digital services to serve analytics needs will be a dependency not to be overlooked. In other words, in order to succeed you need both front end and back end resources and capacity on backlogs, to make changes in any customer facing digital touchpoints.
Here are four key action items to consider, in bringing digital analytics to the next level:
- Customer perspective instead of browsers or devices:
Combining your online behavioral data with other customer data sources, enables to shift the perspective for analysis to an identified data subject or a group of them with similar characteristics of interest. (This topic has been discussed in more detail in an earlier blogpost: Combining behavioral data with customer data requires privacy compliance and technical enablers.) When merging different data sources together, it is important to think of common taxonomies across attributes, not only as a one-off exercise, but for the long term. To create a meaningful and consistent data set, products, event types, conversions, customer definitions and segments, to mention a few examples, should be harmonized being compatible to avoid misinterpretations or totally misleading conclusions. Looking at online behavioral data in isolation from other data sources, unfortunately brings little added value and makes your data less actionable.
- Identifying the value drivers affecting business outcomes:
The very first action is to take control over KPIs, metrics, definitions and structures among them, to reflect your business reality, not standardized interpretations of technology vendors. In practice that means identifying value drivers affecting real tangible business outcomes and understanding thoroughly how KPIs and metrics serving you are actually calculated. Unfortunately, this is rarely something that data analytics platforms can provide as part of their user interfaces, instead they should have a role in providing ad-hoc insights or statistics for non analyst or business users. Structured and for business purposes, a tailored dashboard should be built on raw data in separate dashboarding tools, such as Google Data Studio, Looker, Microsoft Power BI, Qlik Sense or Tableau. As an illustrative example, visits on websites and apps may correlate poorly with desired conversions, but certain behavioral patterns can be identified, leading with high probability during a certain time frame to a desired outcome.
- In-housing skills providing analytical insights on daily or regular basis:
To focus efforts on relevant areas in driving business results, core analytical skills are advisable to keep in-house, as close as possible with the persons having P/L responsibility over the business. Since insights and understanding deriving from online behavior is becoming an even more important source for competitive advantage, as more mature areas of customer insight or business intelligence, these analysts need to have a thorough business context expertise. Gaining that sensitivity for the actual business, besides digital analytics skills, is hard to outsource or to purchase as a service.
- Actionability of data to turn customer insight into concrete activities or actions:
To fully start leveraging value out of your analytics deriving from online behavior, the actionability needs to be kept in mind along with business value drivers. Actionability could be seen in three different dimensions to start with:
- Legal grounds and usage purpose integrated as native part of the data set. From a data utilization perspective, you naturally want to maximize the extent that data can be used for activation purposes, besides customer insight creation.
- As part of value driver identification to adapt early on a way of thinking how to foster and cultivate behavior, leading to expected business outcomes.
- Ensuring to have integrations to channels or touchpoints, where your customer or potential customer naturally selects to interact with your company.
Most important questions to ask your digital analytics ?
In this paragraph, I have translated my observations and key actions, to improve online behavioral data quality into actionable questions to ask stakeholders involved in this field. If you are able to respond to these five questions, you are on the right track to gain value and keep up with competition:
- Do you understand the correlation and causality of drivers for business outcomes?
Have you gained a good level of understanding of what kind of behavior is driving tangible business outcomes, such as sales, leads and so on, over time? If the answer is a clear yes, you are in the position to draw conclusions and initiate action. That means you can see what roles different touchpoints play in customer processes. Understanding business drivers enables you to take action, both on a strategic and tactical level. Strategic level means for example from a CXO perspective analyzing if you are either under- or over-investing in digital services and capabilities. From a tactical perspective you can initiate communication to drive wanted behavior, such as accelerating deal flow, identifying when existing clients are in buying mode.
- Are you able to understand user intent for a visit or multiple visits?
For customer insight creation, one of the most interesting analytical insights deriving from online behavior is the ability to explain the user’s intent for a visit or series of visits. To serve a customer or potential customer in a relevant and personalized way, the very first thing is to understand the intent of a visit. Let us say a user is searching for contact information, finds it directly coming via an organic search from Google and ends up looking at one view of the contact. Even though this is a short visit, the user intent is fulfilled. For another user, who is classified as high probability to convert for a sale within the next 14 days, driving the user towards the conversion and increasing the value of sales via personalization would be a desired outcome.
- Are you able to separate how behavior differs across different personas?
Looking at online behavior through different segments, such as existing customer, potential customer, other stakeholder, customer lifetime value and so on, helps to gain user intent knowledge and to serve these users with relevant content. Enabling separation across user personas requires combining data sets. (Which is discussed in more detail in the following blogpost: Combining behavioral data with customer data requires privacy compliance and technical enablers.)
- What actionable insights are you gaining from each visit and how to act upon them?
Since each visit mostly, except for accidental clicks, has an intent to make data actionable, understanding user intent should be turned immediately into a follow-up action. Doing nothing is an action too, especially not reacting to accidental clicks.
- What data is collected for what specific purpose?
If you cannot answer this question easily, most probably there is unnecessary noise created into data sets, due to collecting data with no clear purpose. Having a clear purpose for each data point has especially two upsides: firstly analyzing data is more straightforward, due to less noise around, and secondly from a compliance perspective the data minimization principle is followed to stand on a more safe side regarding compliance.
Concrete consequences of measurement getting harder
There is no doubt that in the post cookie era, data collection and analytics is becoming more complex all the time and requires an evolving set of competencies. We will explore these changes in more detail in a follow-up blogpost, focusing on the post cookie era. But to give you a concrete idea of the post cookie era consequences, we have identified at least the following five areas being affected:
Marketing effectiveness analysis is becoming less exact:
- Attribution analysis based on clicks, referral paths and visits as known in the past is coming to an end. Instead, methodologies using statistics and probabilities will be used more widely, since following individual users’ behavioral patterns across websites is restricted as a new default standard.
- As new methodologies and conventions are evolving, this could mean increased bias of conversions toward last click channels.
Traditional web-analytics is becoming more complex:
- In addition to marketing effectiveness analysis becoming less exact, the whole practise of web-analytics will be many levels more complex and requiring technical developer skills. Obviously, from an absolute amount of sessions and so on, less visits, events and conversions will actually be measured.
- To address online behavioral data collection and web-analytics becoming more complex, the only way is to explore new measurement methodologies. From a technical standpoint, these require remarkably different skills than before. Obvious skills are related to front end and back end development, raw data engineering and manipulation skills.
Online advertising targeting and retargeting is becoming less efficient:
- Targeting and retargeting capabilities are in general affected negatively. From a data subject or end user perspective, this could be an improvement towards a better experience using the Internet. Annoying retargeting banners may be less common and advertising could become better personalized, due to marketers being forced to rump-up their ability to utilize first-party data for targeting, across advertising and other channels.
- Historical timelines and comparisons make less sense, due to traditional metrics, such as view-through, click-through and so on, becoming obsolete.
Alternative technology development is speeding up:
- The industry will reinvent itself constantly, whereas server-side tracking is currently the most hot topic. For some years, the industry has been talking about so-called data cleaning rooms, which proved not to solve the problem at all, except for justifying to keep increasing advertising spending. Coming up with any valid predictions is guess work, besides ethics and good corporate citizenship being at stake more than ever.
- Besides alternative technological solution development, a closer collaboration between advertisers, technology vendors and publishers in order to exchange data collaboratively, will drive the digital advertising ecosystem in new forms in the coming years.
The illusion of more control for end users:
- The claim of users or data subjects gaining more control over what personal data they give away to whom is really contradictory. We witness the trend of choices given by browsers, platforms and websites, themselves driven by regulatory demands from either authorities or industry self-regulation. The contradiction comes from the fact that at the same time, we also witness a totally opposite trend of new technical advances or rather complex work to go around these privacy controls.
- The ability to control how data is used from a data subject perspective remains to be seen in the coming years. The only point to be taken for granted, is the fact that it will be constantly harder for an average citizen to understand where, by whom and how their personal data is used.
How to prepare for a world where measurement might be less exact
Surrounded by a lot of uncertainty, two aspects are certain for all businesses across any industry, small or large:
- The importance of customer insight and business intelligence derived from digital services usage is only growing, as long as there are processes that will shift to digital channels.
- The accuracy of measurement becomes less exact, and the effort of data collection for any purpose only gets more complex day-by-day.
Looking at the development described above, there are rather logical and straightforward actions to be taken, in order to either stay or become competitive also in the future, despite all the changes around us:
Build trust and direct customer relationships: Nothing else matters than earning and cultivating the trust of customers or users. Trust can be turned into wider permissions and even expectations to communicate directly with your customers and audiences. Treat data as a valuable asset and currency for exchange, which is way more easily said than done in an everyday business context. Contact information will be more valuable than ever for direct or indirect communication. Log-in services will have an advantage for data collection and usage, due to privacy reasons. Taking data-driven sales and marketing seriously, should include having a clear plan to get users using your services in logged-in mode, and maximizing the amount of identified online users.
Flexibility in data collection and processing logic: By adapting a technology agnostic approach for online behavioral data collection, tools or similar mechanisms collecting data can be changed with less effort on the fly. Changes for data collection logics are inevitable due to either legislative or technological drivers. Unfortunately these changes tend to be surprises for most organizations even though they usually are visible for specialists following the industry closely. Besides agility in data collection in itself, agility and speed to make adjustments regarding consent and privacy mechanisms are important too. As online behavioral data collection is becoming more challenging all the time, an ability to run statistical or probabilistic analysis analyzing marketing effectiveness on cloud environments should be seen as a way to fill potential gaps in data capture. Well thought purposeful structure and taxonomy in data models are a key requirement for agility and speed discussed here.
MarTech and analytics architecture design: Build with a solution approach, instead of treating each channel separately, to take advantage of the common IDs that are available within tools in platforms, such as Adobe Experience Cloud, Google Marketing Platform or Salesforce Marketing Cloud. Prepare to handle event level raw data on cloud platforms, rather than aggregated data via tool UIs. Pseudonymize customer and personal identifiers whenever possible, and pair them with online identifiers. Avoid too much vendor lock by identifying critical components to run your use cases delivering the most value.
Workarounds and hacks: Server-side tracking as one solution to be used, either to gain more control over data collection and distribution or as a hack to bypass ad blockers, ITP and so on. Workarounds can be used either responsibly or less responsibly. Relying on other identifiers than cookies or the like, for example fingerprinting, which legally might be questionable, but common practice with a long history.
Advisor and digital business evangelist