Analytics can be a tremendous competitive advantage for your startup, I am sure you already know this. If you get it right, it empowers business users to make the right decisions, it helps fuel growth via product and marketing ROI optimisation, it enables quick feedback cycles on product releases and the list goes on. The key question is: how do you get it right? If you have yet to leverage data to its full potential, or do not know where to even start, then this article is for you! It is the first of a two part series aiming at helping startups set up an analytics function quickly and efficiently.
The Limitations Of Out-of-the-box Analytics
If you are a startup, it is quite likely you already have some kind of analytics in place. You might have a combination of Google Analytics deployed via GTM, rely on some dashboards provided by a 3rd party such as Salesforce, Shopify, etc. Those are a great start, they can give you an overall idea of where the business is heading and do some simple user segmentation. However, you will certainly face limitations of such tools quite quickly. For example, what if you want to join data from multiple systems? What if you want to get a deeper understanding at user level? What if you want to avoid having your team spend hours every week stitching reports together? What if you want to go beyond those limitations and really unlock the true power of analytics?
All those questions can be answered by a more sophisticated analytics setup.
Setting Up An Advanced Analytics Infrastructure
If you want the aforementioned questions answered, then you need to start thinking about investing more into analytics. The first thing to do is often to build the infrastructure so you can extract and manipulate data efficiently. Such infrastructure can be represented in a simple manner through five core layers as shown on the chart below:
Below is a brief definition of what each layer is. The second part of the series will provide a more detailed overview of each layer and the tools you might want to consider for those.
- The extraction layer: automated processes to extract data from your various sources such as app & website events, replica databases, digital marketing platforms, etc. and load them into a data warehouse such as Snowflake, Redshift, BigQuery etc.
- The modelling layer: automated transformation processes that organise the data in your data warehouse so it is easily consumed by reporting tools or analysts — the modelling layer is extremely important, not doing this will most likely create high inefficiency in the reporting, analysis and data science workflows! For example, without modelling your analysts / data scientists will easily spend twice more time cleaning and processing data. Certainly not the best use of their time.
- The reporting layer: a reporting tool (e.g. Looker, Tableau, etc.) that sits on top of your data models and enables non technical users to visualise and digest the data. This covers all the dashboards you do for business users in order to automate insights.
- The analysis layer: a workflow for analysts to do in-depth analyses of the data. This layer is different than the reporting layer in the sense that it is used to investigate specific topics in much greater detail than you would do with dashboards. Jupyter notebooks are often a good option for this. If you end up repeating an analysis frequently, it should then be moved to the reporting layer with an automated dashboard.
- The data science layer: A collection of models and algorithms predicting user behaviours and running on production. Those can be refreshed at any desired frequency with results embedded into other systems for optimisation purposes. The data science layer should be entered once you have a robust structure on all other ones.
An efficient way of getting started is to build the Extraction, Modelling and Reporting layers first. This will enable you to implement quick, reliable and efficient reporting allowing users to make data-driven decisions on the fly. Reporting can easily become a highly time-consuming task in early stage startups — and, spoiler alert!, the value of your team members is not in crunching numbers, it lies in taking decisions based on those! Once you have reporting sorted, you can move on with leveraging analytics in order to improve the various business areas of your company. For example, you might want to analyse cohorts with a high level of granularity, build LTV models and use those for your digital marketing optimisation, etc. The applications are endless, really!
The Team: Who Do You Need?
The next question you might ask is “who should I hire?”. It is extremely important to choose your first analytics hire well as this person will heavily influence the direction of analytics within your company. Here are a few things to bear in mind when doing so:
- Analytics is an extremely competitive labour market. Talent is sought after so you need to react quickly and present yourself as an attractive employer.
- Analytics is an area which requires a broad spectrum of skills. You will need to ensure that your team has people with a strong business acumen (analytics’ final purpose is to improve business performance), people with strong programming skills (e.g. Python, R, SQL), people with solid statistical understanding and people with solid data processing experience (i.e. database performance optimisation, ETL management, etc.). You do not need to find one person with all those skills (and it is unlikely you will) but make sure those are covered by the overall skillset of your team .
- As a result, for your analytics team to be efficient, it is quite likely you will need various roles in your team. You will need data analysts / scientists who are more geared towards translating data into actionable insights and data engineers who are responsible for smooth data processes and building data products.
- When starting, it is a good idea to consider external help in order to ensure you implement best practices from the very beginning (disclaimer: I am the co-founder of an analytics agency called 173TECH that helps high growth companies kickstart their analytics journey). Getting the infrastructure right is extremely important — if not you will face severe inefficiencies when you scale preventing your analytics team to spend time on things that really matter.
So this is it, you should now have a good idea of what to do in order to get started with more advanced analytics. Once you have established a solid base, you can expand and start doing more sophisticated things such as data science in order to build predictive models. This is where most of the value lies and is worth the investment. For example, user segmentation and LTV prediction models can help you optimise your product and marketing towards high value customers and generate hefty returns!
If you want to know more, read the second part of this series which gives specific details on each layer. Or if you need some advice, we are always happy to chat :) Best of luck in your analytics journey!