Startups: 5 steps to enterprise-grade data mining

Even for the smallest, leanest startup, bigger is better when it comes to data mining. The more data sources you have and the more data you can analyze, the better your conclusions — and the more successfully you can serve your specific customer base.

The biggest lesson my company (GameSalad) learned in designing our own data-driven startup is that it’s better to wire in analytics and gather intelligence before you build — and to focus obsessively on the customer from day one. Here are five simple steps that will help you mine with the big dogs.

1. Gather customer data before you build

The first step before building is to gather data from your customers. Challenge your assumptions about who you think they are and who you want them to be. It might be as simple as surveying the people who visit your site. A question such as “What prompted you to come to our site?” can provide helpful information that you may otherwise have overlooked.

Use available tools (i.e. videos of actual user behavior) to learn how people navigate from initial visit to eventual purchase. What made them access one page but not another? Measure what users are doing and determine which key performance indicators (KPIs) need to improve in order to make users more successful. Iterate on the product and user experience to move those KPIs in the right direction.

A/B testing is popular enough to warrant mention here, but I don’t rely on it for any decisions. It takes a lot of traffic and patience to reach the statistical significance that validates assumptions. In most cases, you’re better off ignoring it and simply iterating towards KPIs.

2. Build in data from the beginning

Consider the feedback provided by your customer base when designing your product. Properly instrumenting your product upfront allows you to take advantage of data in multiple ways at startup and saves you the effort of retrofitting it at a later time. A properly instrumented product lets you engage with your customer and observe how they are interacting so that you don’t end up inundating them with survey questions.

3. Collate your data and manage customer flow

In our own company, we use multiple providers for different functions, including data routing, customer support, and marketing automation. MixPanel contains all of our raw data. It monitors user flow, performs retention analysis, and sets up conversion funnel analytics. Segment.io identifies users, tracks user activity, and routes data to the right places. Intercom triggers event-based messages and handles automated retention and engagement marketing. This allows us to determine user activity and preferences, such as where they logged in from, how they arrived at the website, and what pages they are navigating. We also use a custom routing system to ensure that the data stays clean, which becomes especially important with thousands of users generating events.

4. Use proven strategies to streamline the process

We learned early on that iterating quickly is key: Grand designs do not work. Applying the Agile approach, we are able to systematically meet the day-to-day needs of our business while still preserving time to consider new options and explore alternate strategies. We constantly measure, review, make corrections, and repeat. Monthly or quarterly plans only serve to create additional work and promote inflexibility. We hammer away at each component every single day, discovering what we know and don’t know, taking it one step at a time to solve the hardest and most critical issues and iterating on that.

When determining which user archetypes might be most appropriate to your business, use smart strategies to avoid becoming mired in archetype discovery. Finding out who is using your product and why may seem like an easy task, but it involves looking at raw data and finding correlations. These programming and data science packages exist in R and Python and can help you decide which and how many user archetypes are needed.

5. Empower your employees

A simple change from “customer support” to “customer success” can make a huge difference in employee attitude and client satisfaction. Support implies a burden, something you have to do. Success is something you share in, something you want to do. Customer success is every employee’s job, so they need have to have the power and authority to advocate on behalf of the customer. Empowered workers result in empowered customers.

In the past, we did not have the instrumentation to know what our users were doing. Now we can see what they are clicking, where they logged in from, and where they went from tool to website. This allows us to be hands-on with each customer, whether walking them through a particular process or handling a specific concern. We know who is visiting our website and, as a result, can tailor it to reach the widest audience. Even more importantly, we can continue to customize our products to suit what our users need — not just what we assume they want.

In 2016 and beyond, this will be a foundational ability for all companies, and it will separate those that thrive from those that are still trying to adapt based on guesswork.

Stephen Nichols is the CEO of Austin startup GameSalad.