In the latest Bitesize webinar, Adverity’s Andy Fairclough and Joe Caston explain what data governance is and how best to implement it. Here we outline their 7 key building blocks for creating the right data structure for your business.
Simply put, data governance means having a solid data foundation by establishing guides and standards for managing and controlling your data. Sounds straightforward enough, but the truth is that many businesses are not doing enough to ensure good data governance principles, and this is costing them.
Studies from Experian have shown that some businesses are losing as much as 12% of their revenue each year from inaccurate data, while according to the Harvard Business Review, it costs 10 times as much to correctly complete a piece of work with bad data.
To find out how you can avoid this trap and build a robust data governance system, check out the latest Adverity Bitesize webinar with Andy Fairclough and Joe Caston, or read on for the 7 key building blocks for good data governance.
1. Data Access and Ownership
The building blocks of sound data governance start with setting the right access privileges. This means not only restricting who has access to certain types of data, but also what people can do with it. Handing over the ability to delete large historical datasets comes with a certain responsibility, so you need to make sure that only the right people have this level of control.
This is important, especially when it comes to Personally Identifiable Information (PII) data, which needs to be ring-fenced and accessed on a need-to-know basis. It’s also worth talking to your designated Information Security officer about how long you can keep hold of that PII data before purging it.
On the flip side, you also need to make sure the people who need access to the right data actually have it. The best way to break down all those doors and get ownership of all your data is to do it systematically, and patiently, with shared documentation that you update as you gain access to each data source, listing the account owner, the type of authorization required, and tracking whether the account owner has been sent a request.
That authorization data needs to be stored safely, if it's managed from a third party software, and/or is data that isn’t hosted on your organization’s owned properties, then you need to ensure it's encrypted and stored to ISO and/or SOC2 security standards, as there is a level of vulnerability there.
2. Unification and Consolidation
The most obvious way of improving the value of data is getting it all in one place. Once it’s all in one environment and flowing in regularly you can really start to step up your data analysis.
Your data will arrive in many different file types and formats, and so a key part of the data unification step is the process of consolidating data into a single format before storing it. Getting this right means that subsequent processes can be done in a more efficient way and at scale.
Once unified, your data needs to be properly classified. This means separating it into different data types - for example, quantitative data, qualitative data, dates, strings, whole numbers, decimals, currencies, percentages, floats, etc.
You can also never presume that all incoming data will be correctly classified. Some datasets will have completely lost all their data-types post-unification. Different organizations and software solutions will have their own ways of recognizing and classifying different data-types. For smaller organizations without the necessary software, this may mean hours spent in Excel manually ‘scrubbing’ reports to make sure the data-type is formatted correctly.
Enrichment is a fundamental aspect of data governance and consists of many different facets.
Similar to the step of unifying your data into a single environment, enrichment is another key process in squeezing every last drop of value from your data.
There are several ways in which your data can be enriched. Naming conventions, for example, involve creating a systemized approach to naming all campaigns, ad sets, and creative assets - a bit like how you would structure all your different files and folders to make it easier to sort and search for them.
Likewise, data can be joined together. For example, if you want to calculate your Return on Ad Spend (or ROAS), then you’ll nearly always need some form of ‘join’ between your Revenue data and your Ad Spend data. This is a really important step when it comes to getting the most value out of your data at a later stage and there is nothing more satisfying than breaking down data silos with a nice simple join transformation.
Next comes the all-important process of harmonization, and this is all about making sure that data from different sources (but essentially about the same thing) matchup for reporting. For instance, Facebook calls Media Spend ‘Spend’ while Google calls it ‘Cost’ and for Linkedin, it’s ‘costinlocalCurrency’.
Each platform has its own idiosyncratic way of naming things, and if your data is full of those variations, you need to harmonize them otherwise you won't be able to compare the same data types across platforms and marketing channels.
It’s critical to be able to trust your data. Any quality issues, incomplete data sets, and discrepancies between platforms all undermine that trust. Discrepancies between two or more platforms that essentially track the same user event (such as an impression click or conversion), can be particularly baffling. It's best practice to set up an automated way to monitor discrepancies between platforms that measure the same metric. If those differences get too large then some difficult questions need to be asked about who is counting that event correctly.
You should also run regular ‘data sanity’ checks. This is perhaps the most challenging part of data governance as there is no silver bullet for this one. Most problems around data sanity have to be addressed by changing the way organizations work in the marketing platforms themselves, perhaps even more importantly, looking at the accuracy of the data being generated by their owned properties such as their website, app, eCommerce site, etc.
Nonetheless, if you really want to take your data governance to the next level, you should try to tackle this head-on. Setting up and automating a weekly data sanity check will help you identify weaknesses in your data and get to the root cause faster.
7. Augmented Analytics
Finally, we come to the last building block - augmented analytics. It is a proactive approach to presenting critical marketing intelligence by applying AI to uncover deep, hidden insights and issues within multi-channel data.
- Immediate detection of risky outliers caused by unforeseen human errors;
- Identifying potentially costly problems as early as possible, before they become a major issue; and
- Uncovering key anomalies which would be impossible to spot by conventional analysis.
Revealing anomalies is perhaps one of the most useful functions of the augmented analytics component in your data set-up. For example, AI-powered systems can quickly spot problems such as a drop in conversion rate (indicating a webpage pixel malfunction), a dip in ROAS (showing ineffective campaign spend increase), a drop in engagement rate (indicating unsuccessful new creatives), or CPM decreasing to zero (indicating unseen ad account suspension).
Where are you on the data maturity curve?
Putting in place these building blocks will help establish a solid system for data governance within your organization. And, with each step, a business will become more mature, able to extract more and more knowledge and intelligence from their data. This is what is called the “data maturity curve” and most businesses who are serious about the role data plays are somewhere on this trajectory. As you think about your own data governance, think about where you currently are on this curve, and what steps you can take to reach higher up.