Our Insights.

Covering strategic and technical aspects of analytics.

By subscribing, you agree to receive a monthly newsletter from 173Tech.

The Analytics Guide, Getting Started.

September 15, 2020 | By Robin Watteaux.

Over the past year, we have worked with many startups to kickstart their analytics efforts. We know that getting started with analytics can be daunting at first. There are so many options that it makes it difficult to understand what is the right approach to build a proficient analytics department.

Technical implementation is often when things start to go wrong. Any single mistake will add an increased burden over time on your analytics department. At 173tech, we have seen the highs of analytics, but we have also seen the lows. We know the traps most companies fall into. If you want to ensure your analytics team focuses on value generating tasks (i.e. finding actionable business insights) as opposed to fixing bugs and infrastructure, then read on! This guide provides clear guidelines to get started with analytics and outlines the analytics priorities at each stage along the way.

The Beginnings: I Want Numbers, Quick!

You are getting started, your efforts are certainly almost exclusively focused on getting your product right and fueling growth. You have limited time and resources to get fancy with analytics — and this is fine for now. What you need is simple: a good overview of where the business is heading, quickly and at little cost. The following will get you started with analytics whilst minimising the amount of time and resources invested. However, as you scale, you will quickly come to face the limitations of this setup and need to switch to a setup optimised for scale.

Analytics Priorities:
  • Implement a product analytics solution (e.g. Google Analytics, Amplitude, Heap, Mixpanel) in order to track key metrics (e.g. registrations, retention, etc.).
  • Use the reporting UIs of the various out-of-the-box tools you choose.
  • If you are going big on marketing, track all sources with UTMs for web and use a mobile measurement partner for mobile (e.g. Adjust, AppsFlyer).
  • You can use tagging solutions such as Google Tag Manager to relieve your development team from being the bottleneck of implementing a lot of tracking.
  • If you need raw event collection, you can also consider using a Customer Data Platform (e.g. Segment, mParticle). However, your product analytics solution should suffice for now.
Limitations at scale:
  • No / Limited data ownership. With product analytics solutions, you do not fully own your data. Some product analytics solutions will ask you to move to paying tiers to access your raw event data. In addition, you might need to have to pay for data retention over time.
  • Limited analysis possibilities (hence limited value extracted from your data). Those solutions will never be as flexible as SQL/Python based analytics where you can crunch the data as desired. In addition, if you want to do more sophisticated segmentation, you might have to move to a paying tier depending on the provider.
  • No single customer view. Because you have segmented tools for analytics, it will be almost impossible to join data from different business areas (e.g. understanding the impact of emailing on user retention).
  • Cost inefficient at scale. As your business grows, so will your data volume. Your product analytics and customer data platform tools can become really expensive!

If you want to avoid those limitations from the very beginning, you can decide to get started with the longer-term solution directly (next section). However, this will require more involvement to set up and maintain — especially as the structure of your data in the early days might be unstable.

Modern Analytics: An Agile and Scalable Analytics Solution

You are experiencing significant growth and your team now has more than 20 people. You find yourself increasingly limited by the capabilities of your out-of-the-box solutions as the need to better understand and optimise customer journey intensifies. On top of all this, your tools are becoming more and more expensive as you scale. This is when transitioning to a modern analytics tool stack becomes critical. This solution is infinitely scalable, ensures full data ownership, enables connection to all your data sources (i.e. product, marketing, CRM, etc.) and fast-tracks your business to growth intelligence.

Modern analytics stacks rely on a centralised data warehouse (the repository of the Single Customer View) where data is loaded and modelled. This is then complemented with a BI reporting tool.

Analytics Priorities:

Focus on laying robust data foundations. Everything you do at this stage will directly impact the success of your future data projects — a messy infrastructure setup will lead to many headaches!

  1. Create your “Single Customer View” in a data warehouse. This will enable you to load data from all sources in a central location and be infinitely scalable cost efficiently:

    • Choose a data warehouse (e.g. Redshift, Snowflake, BigQuery, Exasol). Stay away from operational databases such as PostgreSQL for analytics, they do not optimise and scale well.
    • Automate data extraction as much as you can (e.g. Stitch, Fivetran).
    • Use a data processing and modelling framework which supports both SQL and Python (e.g. SAYN — open source and developed with love by us @ 173tech). It is important to ensure you have the flexibility to use both languages to keep your data processes agile — especially for data science or complementing your automated data extraction tool.
    • Implement a reporting solution for self-service BI (e.g. Metabase, Looker). Metabase is great to start with as it is open source and free. Looker has more muscles as you grow.
  2. Organise data processes and workflows in your analytics team:

    • Make sure your analytics team uses version control (e.g. Git). This will facilitate analytics code maintenance and QA as it grows.
    • Implement a data QA process and code reviews. Ensure all data modelling is reconciled against its source of truth for accuracy. Code reviews minimise mistakes and facilitate knowledge sharing across the team.
    • Prioritise your tracking efforts. Make sure that you cover all core metrics across the business accurately first. More tracking means more engineering effort and you will typically find that 90% of the business questions can be covered with 10% of your data estate.
  3. Automate reporting across the business and save (a lot of!) time. Creating one master dashboard for the business overview and one core dashboard per department is a good start.

  4. Create documentation for events and metrics. Build a data dictionary which business users can refer to quickly for definitions of metrics.

Setting up such an infrastructure can take as little time as 8 weeks depending on the current state of your data ecosystem — if you want more detail on this stage you can refer to my article series “Building World-Class Analytics For Startups”.

Growth Intelligence: Mining Data Value

Now you have an agile and scalable data infrastructure to facilitate infinite growth. It is time to generate unlimited value out of your data! This is what we call “Growth Intelligence” and is all about applying data to maximise your business’ performance.

Analytics Priorities:
  • Set up a process for ad hoc analyses. These are one-off business requests aimed at quick insights and creating an automated dashboard is not necessary. Use Jupyter notebooks with Python / R for this.
  • There is a lot of value in descriptive analytics — so make sure your analysts review the data before getting fancy!
  • Identify high ROI data science projects. Some examples of our frequent and top winners:
    • Customer Lifetime Value (LTV) prediction coupled with marketing data to optimise digital spend. If you are a big marketing spender, a lot of value lies there!
    • Churn prediction to help improve retention.
    • A/B test results automation.
    • Natural Language Processing (NLP) at scale on text data such as customer reviews and tweets for instant product feedback and competitor insights.
  • Look into building robust event tracking. Define your event taxonomy and check out Snowplow for implementation. It is open source and you can deploy it on your own infrastructure — meaning no huge 3rd party bills at scale.

Where Next?

This guide hopefully brought you some clarity on how to get started with analytics. If you follow the above, you will be in a good position to have a proficient analytics department and unlock a lot of business value from your data. “Growth Intelligence” is only a few weeks away!


Subscribe to our monthly newsletter for more insights.

By subscribing, you agree to receive a monthly newsletter from 173Tech.

Get In Touch

Send us a quick message about your data challenges, we are always happy to chat!