The long term trend in communication is that advertising is becoming digital, digital is becoming addressable and addressable is being bought programmatically. All this is being driven by data.
However, data knowledge and expertise are becoming increasingly specialised and siloed. As a result, there are gaps in people’s and organisations knowledge and understanding of data. Not only are there gaps between different specialists but there are also gaps between practitioners and leaders. This makes it hard to have a broad conversation about the different elements of a data strategy and what a data development roadmap might look like.
There are also practitioners who profit from muddying the waters and adding complexity and confusion. Typically they will use a Lumascape chart to further bamboozle adding granularity and detail ("look this is really complicated") just when the opposite is needed.
The roadmap outlined below seeks to identify the key stages you should be thinking about in defining a data roadmap and also highlights the key questions and discussions you might have at each stage. I am not a data expert but I regularly work with people who are. This roadmap is far from perfect but is my attempt at providing a simple guide for non-experts to able them to understand (and better question and interrogate) the work of people who are.
You can’t do anything until you have identified what data you’ve got and started collecting the data you already have access to. This stage typically involves a data audit and ensuring you are collecting and tagging the data from your key owned media (your website, CRM programme and social channels).
Typical discussions would be about first party data (data you own) such as cookie data, social, email and web site data. You would also discuss how you are tagging and collecting data from your website and social channels.
Once you have identified and started to collect your own data you will need to bring it all together in one place. This stage typically involves connecting up data sources, consolidating the platforms you are using and deploying new data management platforms.
Typical discussions would be about moving to a single ad server and building a DMP (data management platform) to house your data. You might debate whether to manage your data yourself or to outsource it and whether to use one data providers ‘stack’ or use different providers for different elements.
The oft used quote ‘shit in, shit out’ is never truer than using data. Once you have gathered your data together it is important to remove rogue fields and rogue entries to ensure the quality of your base data set. Sometimes a simple ranking exercise can be illuminating – for example should Nigeria’s ad spend be higher than the USA or has someone entered that number in local currency?
Not only does your data need to be clean but you also need to store and process it in line with the latest regulations. This has become more of a business imperative in the run up to the implementation of General Data Protection Regulation (GDPR) on 25th May 2018. The reason this is getting so much attention is that firms can be fined up to 4% of their turnover if they are not compliant with the regulations.
Typical discussions would be around how data has been collected and inputted, how old your data is and has is expired. You might debate what protocols and checks you have in place to scrutinise data and how these sources have been audited (or not!).
You might recall NASA losing a $125 Mars orbiter because one engineering team were using the metric system and the other imperial. There was nothing ‘wrong’ with the team’s data but how they had organised and labeled it has disastrous consequences. Establishing a clear and consistent taxonomy to classify your data might seem like a trivial step but it is fundamental to the success of you long term data strategy.
Typical discussions would be around how you tag and classify different data sets. You might debate how you group and cluster different elements of your communications. And you might discuss when you need to reclassify and reorganise your data as this is a step that will continually evolve as your data and data needs expand.
There is a lot of foundational work that needs to be done before you can start to analyse and draw conclusions from your data. But these are essential to ensure you can draw valid conclusions from the data you are analysing. Rather than jump into specific analysis and granular investigations it is worth starting by looking at some broad questions to get a ‘feel’ for the data set you are looking at. Data visualisation can help at this stage as it allows you to start to look at and investigate your data.
Typical discussions would be around dashboards, Reporting and KPIs. For your initial analysis you might look at seasonality, the size, and distribution of customers and prospects and the broad channels that are performing well. As a result of this analysis you might debate the validity of information and insights produced and if further investigations are needed
Once you have an understanding of what is happening and what is working (or not) you can start to tweak and adapt your plans. The insights you have identified in the previous step can be used to optimise your plan.
Typical discussions would be around measurement and evaluation using various modelling techniques from simple regression analysis to attribution modelling and more complex methodologies. You might debate whether you have uncovered a cause or a correlation between data points and if you are attributing success to the right factor or just focusing on the elements that are just the easiest to measure.
As your data knowledge and understanding develops you will start to be able to identify holes in your knowledge – I think Donald Rumsfeld would call those ‘known unknowns’. There is a limit to what you can infer from a data set and this is where you might consider adding data to improve your knowledge and understanding as well as your ability to execute.
Typical discussions would be about second party data (data you borrow from partners) and third party data (data you buy from other providers) You might debate which data providers you use – such as Axciom, Oracle and Blue Kai – the cost of that data and what fields you want – location, age, interests, purchase history etc. If you are combining different data sets you might also discuss data legislation and whether you need to anonymise your data and host it in safe harbour.
Now your data strategy should be purring and you can start to look at how you can speed up the process and to automate the repetitive and manually intensive stages. These are things that computers excel at.
Typical discussions would be about machine learning and automation. You might debate what can and can’t be automated and how you can check that you are optimising towards the right KPI.
I hope that has provided you with a simple framework and a set of questions to guide your own thinking about data. I'd be interested in hearing your thoughts. Do you agree with those questions? Are they in the right order and what is missing?
Paul is Global Head of Strategy at Vizeum. He is a Global Strategist with experience that spans a variety of sectors (CPG, Tech, Pharma and Finance) and disciplines (Media, Advertising, CRM and Sales Promotion). He is responsible for European Strategy across all Starcom Global Network Clients including Samsung, P&G, Coke, Airbnb, Novartis, Etihad, Mars. Paul holds a Bachelor in Biological Sciences, Zoology from the University of Oxford.