Startups’ building blocks for BI

After a certain point, startups must build a Business Intelligence (BI) infrastructure. However, without expertise in BI, it’s difficult to know when or to what extent an organization should invest in analytics. Drafting a structured roadmap during this “analytics onboarding” process will allow business to scale their BI or Analytics team in a positive ROI manner. Below is a roadmap for bringing analytics into startups or growing SMBs.

Data

Data is, of course, the key ingredient in analytics. However, structuring data is the most difficult and time-consuming part of any analytsis. It’s even more difficult at the enterprise level — we must consistently connect and align data across many, disparate systems. However, once a company has well-managed data, they will save thousands of hours in future labor costs per year (and, in some cases, their analysts’ sanity). The goals of the “Data” step are to:

1. Connect all valuable data sources into a single database
2. Calculate dimensions, segments and metrics in a uniform way across the business
3. Ensure data and calculations are accurate
4. Automate the ETL process and provide seamlessly updated data for reporting and analysis

Reporting

Every decision maker across a business (no matter how small or significant the decision) can benefit from being better informed. This is where reporting comes into play. Building a reporting infrastructure is an iterative process — business operations give feedback on reporting and business analysts make the necessary augmentations.

From a technical perspective, tools such as Tableau, Domo and Looker allow users to easily build reports that can be automatically updated (once connected to a data warehouse). Due to their ease of use and visual appeal, these SaaS tools are replacing Excel reporting at a rapid rate. Even non-technical employees can build valuable reports with these platforms. However, no matter the end user or the reporting tool used, all quality reports share certain qualities:

1. Metrics and segments are clearly defined
2. Contains a high level or executive summary with the metrics most relevant to the end user
3. Combines charts and tables that allow end users to uncover the “data story”
4. Automatically updates (through a data visualization tool)
5. Guides and/or constraints that prevent the end users from pulling incorrect or misinterpreting data

By the time a B2C company reaches 100+ employees, they should look for a dedicated analyst. This person will work closely with back end developers and/or the IT team to manage company data and reporting. Prior to hiring this person, developers with business acumen or a business strategists with technical chops can own these responsibilities.

Exploratory Analysis

Reporting provides a foundation for Operations to ask meaningful business questions. When these questions cannot be answered through a standard reporting tool, then we move to “exploratory analysis” phase. The general goals of exploratory analysis are to:

1. Answer key business questions that are currently unknown in the simplest way possible.
2. Develop concrete findings and business recommendations that can be tested in the future
3. Document results in a way that can be transmitted to decision makers and maintained throughout the duration of the business

Bespoke analyses are naturally more time consuming to create than standard reporting; therefore, companies moving to this phase should consider hiring a small analytics team. As a team, analysts can gain efficiencies of scale by specializing in areas of expertise, building from prior knowledge and sharing templates (e.g. code, Docker images). One of the most important (but commonly overlooked) steps of ad hoc analytics is collecting an archive of findings and recommendations that can be easily shared across the business over time.

Testing

A testing framework goes hand in hand with exploratory analysis. Below is a summary of how businesses can develop effective cycles of strategic testing:

1. Business stakeholders (e.g. executives, product owners, ops teams) ask questions based on hypotheses and their existing knowledge of the business
2. Analysts explore these questions — either through A/B testing or correlative analysis — ultimately resulting in insights
3. Operations and analysts work together to develop strategic recommendations
4. Operations implement the recommendations (or MVP versions of them) and use A/B testing to determine to determine the incremental uplift from each change
5. Based on the results of the test, operations and management determine a final course of action

This continuous cycle of testing and strategic improvement is the linchpin of Eric Ries’ theory in “The Lean Startup.” Although this system was originally created for startups, it has been effectively used by Fortune 500 companies as well.

Advanced Analytics

After the analytics team has used reporting and exploratory analysis to gain a thorough understanding of their data, they can implement advanced analytics. Advanced analytics is the use of statistics to achieve better business results. Below are a few statistical methods and applications within the purview of “advanced business analytics”:

• GLM Regressions. Forecasting demand for products and categories in an e-commerce store
• Logistic Regression. Determining whether or not a certain online transaction was fraudulent
• Cluster Analysis. Determining similar types of customers based on their purchase behavior

In conclusion, this is a (very, very simplified) overview of enterprise analytics’ building blocks, tailored for startups. While the four “pillars of the pyramid” are key to scaling enterprise analytics, I believe the following concepts are just as important when building such an infrastructure.

• Start with the data; make it clean, accurate and accessible
• Use prior insights and unanswered questions to guide next steps
• Build iteratively, using feedback loops to improve reporting and analysis