This promoted content is produced by a member of The Drum Network.
Data science for marketers: 5 steps to preparing your data
Every business holds a unique key to overhauling their marketing success. That key is data. Or more precisely, unique insights that could be derived from that data and applied to optimising different marketing activities – from customer retention to search engine marketing and pricing strategies.
The wrinkle, however, is that big data analytics isn’t a point-and-click solution like your typical MarTech tool. It requires certain operational changes, backed up a solid transition plan.
So if you are already sold on the benefits of big data in marketing, here are the next steps you should take.
Step 1: Get company-wide buy-in
Marketing data analytics adoption is a big step (not just because you plan to use big data). It assumes that your organisation is ready to embrace not only technical changes but undergo an organisation-wide transformation that will tackle the people and the processes as well.
According to Big Data Executive Survey 2018, 48.5% of respondents name ‘people challenges’ as the main barrier to becoming a data-driven organisation versus 19% identifying technology as an issue. Over half of respondents also state that insufficient organisational alignment and/or cultural resistance are the two issues slowing down the adoption of new technologies.
To succeed, you will need to get the critical mass of employees on-board. Education is the simplest cure for that. Show your teams that marketing analytics isn’t a threat to their careers or skill sets. On the contrary, data science and big data analytics can help your marketers do their best work without being decelerated by mundane, repetitive chores.
Step 2: Determine the types of data you will need
Your analytics will only get as good as your data. The problem? Most businesses already feel overwhelmed with what they have. And that data is typically siloed in all the wrong places.
To determine what data to operationalise first, consider the end-game of your big data programme. What are the exact outcomes you want to achieve? Data science use cases in marketing are manifold:
- Programmatic pricing optimisation and dynamic pricing systems;
- Advanced customer segmentation and user profiling for personalised marketing;
- Predictive lead scoring;
- Search marketing foresight that enables you to determine the success of different campaigns;
- ‘Intelligent’ PPC advertising campaigns that will self-adjust depending on changes in the advertising environment.
A clear end-goal will help you determine the primary data sources to prepare for departure. Start with defining the minimal data sources you will need to achieve the results. For most companies that would be a CRM system and online analytics tools. Though, you may later consider connecting additional data streams on an ad hoc basis.
Step 3: Arrange a consistent flow of data
Digital marketing assumes a constant inflow of dynamic data from multiple sources – social media, website analytics tools, email marketing tools, advertising platforms and so on. Your job is to ensure that new information can be seamlessly delivered for analysis.
Often this means that you will have to account for both structured and unstructured data. Structured data is delivered in a machine-friendly format; 99% of the time it’s readily available for further analysis.
Unstructured data, on the contrary, is everything we humans love to share – videos, text messages or documents, audio, images. Such assets need to be converted to other formats before they can be processed by the algorithms. Certainly, this presents additional challenges, but the trade-offs are significant. Unstructured data usually holds unique insights that cannot be retrieved any other way.
Step 4: Schedule data cleansing
Depending on your tech capabilities, you may either perform data cleansing in-house or outsource it to a data-science vendor.
Before being sent for analysis, all your proprietary and external data will have to be converted to the same format, suitable for the algorithms to comb through. Duplicate and ‘fuzzy’ data should be also removed. Data cleansing is a crucial step to ensure that your analytics will work properly and deliver objective insights. This is a time-consuming step as well: 60% of data scientists say that they spend the majority of their time on data cleansing and quality assurance.
Step 5: Launch the data consolidation process
Now all your crispy-clean data, along with the data sources should be directed towards a single destination – a data lake. Data consolidation can seem like a major technical investment, but it pays off in multiple ways:
- Reduced total cost of ownership. A single storage unit means that you no longer have to pay for multiple software licenses used by different teams; and for storage space to accommodate ever-growing databases.
- Increased efficiency and productivity. A single entry point means that your data becomes readily available for further analysis and easily accessible to employees in different departments.
- Simplified compliance. By knowing exactly what kind of data you are collecting and where it resides you can comply with new data protection laws and requirements without much hassle.
The final preparation step is to find the right talent who will transform your business data into actionable insights. You will need to consider what skill sets your organisation requires to execute every step of your data programme – from determining the best use case and estimating ROI to selecting and implementing the best technology strategy – and fill in the gaps with external expertise.
The second part of this article can be found here.
George Karapalidis, head of data science, Vertical Leap
Have your say
Do you have a strong opinion on a topical industry issue? To submit a comment piece, please send a short summary of your idea to email@example.com. Views of writers are not necessarily those of The Drum.