data

Grow roots for advanced analytics through better data quality

COVID-19 is speeding up digitalization and eCommerce, there’s no doubt about it. Because of the unexpected crisis, many organizations have been forced to stop and adjust their plans. The storm cloud might have a silver lining as this is a good time to analyze how you are acquiring new customers, which sales channels are effective and how do you deliver your products or services. Some companies are decreasing their advertising spend but at the same time marketing in digital channels becomes affordable even for the smaller businesses. For all, this creates an opportunity to evaluate the marketing channel strategy and improve marketing spend effectiveness.

In this article I explain how you can make sure that your data is valid and relevant for evaluating options and taking action. High quality data is also the foundation for more advanced analytics solutions.

Petri Mertanen

audit and validate data quality

Whatever you decide or plan to do with your marketing spend during this crisis, data driven culture is the way to go. And when it comes to data, the quality of data is very important when making decisions and building the ability to use it in marketing. High quality data is the foundation of your house of analytics. If you want your house to last and have multiple floors, you need to have a very solid foundation. If you don’t, your house will collapse sooner or later. Every now and then, you should make data audits or have a more continuous way to check the quality of your data. Although we would like data to be 100% correct, there’s no such a thing as perfect data, especially when it comes to digital channels. By doing an audit to your data, you know how good your data is.

When eCommerce is blooming, it’s obvious that your web/mobile analytics system is playing a key role. According to my previous research and current estimation, Google Analytics market share of (website) installations, is up to 80-90 % in Finland. I have been involved with hundreds of digital analytics audits and only few companies pass. Audits are must, but when you have collected data regarding your goals and possible online purchases, you should validate the data (against back-end) from time to time as well. And every time when you’re doing a test or launching a new campaign, you should debug and test the analytics setup and data collection. This way you’re able to maintain already achieved good data quality.

 

improve data relevancy

If you have done the default implementation, you’re not going to get real business value out of your analytics system. Luckily, this is quite easy to check: Do you measure business outputs as goals? Likewise, if you have zero events or zero custom dimensions in place, you’re probably not tracking your website or mobile app close enough. Advanced setup usually means that there is a data layer in place, and you are tracking your website or app in very detailed level (with events).

Google Analytics Events

Debugging and testing can be very time consuming to do.  Fortunately, there are systems to automate data quality checks: Observe Point has many products and one of the newest one, Datail, to mention few. Unfortunately, good data quality in one channel is not enough. In many organizations the data is spread in different kind of systems. Important data may be in CRM, advertising platforms, content management and web analytics systems.

 

define, test, integrate and visualize data

In order to get and report real Key Performance Indicators (KPIs), you need to centralize data from silos to somewhere. Especially advertising spend can be in various ad management systems. You need that data in order to count Cost per Acquisition or Cost per Action (CPA) or Return on ad Spend (ROAS). These are examples of important KPIs, as you need to know how much certain business output costs or how much revenue you are getting back from each channel, you’re advertising in. If you need to test whether your metric is a KPI, there’s a nice So what –test by Avinash Kaushik. With this test in mind, most of the people learn and ditch metrics like average time on site or average page views per session.

It is equally important that you don’t just report your KPIs – you need to set target levels for your goals. Your dashboard should show your key metrics in a timeline and compared to target level and previous year. You should see this data in a daily, weekly and monthly trend and year-to-date cumulative graph.

Petri Mertanen analysis of goals and actuals

In addition, you should have multiple data sources in your dashboard in order to succeed. If the need for integrating multiple data sources is permanent, there are many ways for building a Marketing Data Warehouse or a Data Lake solution. If you have a centralized Data Warehouse, then you don't need to do the integration work in the dashboard tool. Either way, there are also tools to help the integration work and keeping the metric metadata updated, such as Supermetrics

 
analyze data to take action

Sharing data in your organization with a dashboard is not enough. Someone or something has to analyze the data - otherwise the numbers are useless. The most basic approach is heuristic analysis which is based on testing and observing website usability and comparing it to other similar sites or previous user experiences. This is not a very scientific as the output of analysis depends much on the person. Anyway, it’s better to do heuristic analysis than analysis at all. In this case we are usually describing what has happened and then why it happened? Good analysis includes clear recommendations or testing hypothesis to act on. When you are doing AB-testing and conversion rate optimization by testing, you have already taken a step or two forward in your data driven path.  The key is to let the testing results guide you also in those situations when they are not the same as your intuition would say.

 

grow roots for more advanced analytics

When the organization’s maturity evolves, you are probably using algorithms and Machine Learning (ML) models to predict certain behavior. There are so many areas in marketing where advanced modelling can be used. Advertising systems have used algorithms for years for targeting, profiling and inventory optimization. There is also predictive analysis and anomaly detection in Google Analytics. Marketing Mix Modelling (MMM) is definitely one of the most well-known advanced analytics solutions. With MMM you should get the big picture of your marketing actions, optimize the spend across marketing channels and analyze the effect of all the marketing Ps in use.

During the last couple of years, many start-ups have risen in the field of Machine Learning. These software vendors usually connect to several data sources like Google Ads, Analytics and Facebook. You can check for example Diib and Uplifter in this scene. With these kind of vendors, and Google Analytics Intelligence, you have to know that you’re dealing with black-box solutions that have not been trained to fit to your business context. If you are statistician, mathematical person, programmer or Data Scientist, you may want to see and know what is under the hood. In this case you may want to test ML as a service type of platforms, like BigML.com. And yes, some of you just want to use R or Python and that’s ok too. No matter if you’re struggling with the first solid implementation or fighting to find explaining algorithm, it’s important to know where you stand and what to develop next in order to get competitive edge from analytics. You can test your Measurement Maturity in here.

 

contact us for help

If you are interested to improving the quality of your data, optimizing your marketing spend with data and growing your roots for more advanced analytics solutions, contact us at AI Roots and our network of data professionals for help. AI Roots is also starting collaboration with MarkkinointiKollektiivi, a network of marketing professionals in Finland, to help marketing teams to utilize data and advanced analytics in their work.