Crossing Your Data Science Chasm

Published Mar 22, 2018
Crossing Your Data Science Chasm

An analytics roadmap for growth

Scenario — You’re an up-and-coming e-commerce/SaaS startup. You’ve got your site up, you’ve A/B tested your message, and you’ve got your SEO and social ad buy.

You’ve got your email drip campaign and reminders. You also have basic BI reporting telling you channel traffic and conversions. Traffic is decent, and revenue is growing.

It’s likely that you are in the initial growth phase, you’re flying on skates, and business is booming. You think you’ve found your voice, you’re expanding, you’re reinvesting for more growth, you’re going big on marketing... things seem to be going great for a while.

Then, all of a sudden, traffic slows down.

Ad buys are not as effective as they used to be. Promotions still bring a spike, but they are temporary and short-lived. Overall growth can’t seem to get over the hump and LTV is in decline, which means it would take twice as long to be profitable.

Why?

There could be many reasons, but a likely reason is that, inevitably, the same sales and marketing engines that have worked in the past have started to slow down over time, explained by famous growth-hacker-turned-VC Andrew Chen, in his post, “Law of shitty clickthroughs.”

Over time, all marketing strategies result in shitty clickthrough rates. — Andrew Chen

We’ve all heard the term “crossing the chasm:” all successful technology products eventually have to cross the chasm to leap from early adopters to mainstream consumers. Finding product/market fit is hard, and making this leap is not trivial.

Whether you’ve started from scratch with a novel idea and hustled your way to what it is today, or you had a successful transition of your offline business to online, you’ve taken your first beachhead and secured your first batch of loyal customers and early adopters.

To get to the next level, you’ve got to look more closely at optimizing internal operations as well as opportunities to expand. It’s more important than ever now to have all of the information you need to make the right decision, and make more winning bets than losing ones.

Now you’ve got to think deeper and exercise data muscles you’ve never thought to use before. You need to know who your customers are, when are they likely to purchase, and what are they going to buy. In other words, going from knowing “what happened” to “why it happened” and “what will happen,” or, in short, predictive analytics.

This is where Data Science and Machine Learning come into play.

But, wait a minute, you say, you have already looked into this but couldn’t find good out-of-the-box solutions on the market... or maybe you tried to hire a data scientist but found out it’s hard to find the right combination of talent/cost/skills?

Or maybe you think the metrics you have and your decision making are good enough, because they’ve gotten you this far, right?

This is what I’d like to call the Data Science Chasm — the gap between web analytics/BI reporting to predictive analytics. It’s like going from Google Analytics to Google Deepmind (of AlphaGo fame).

The typical choice of Buy vs. Build breaks down because of the inaccessibility of data science today. Due to a shortage of talent, insufficiently defined problem spaces, lack of intuitive, affordable products, and lopsided labor markets toward top tier companies has prevented many startups and SMBs from gaining access to early data science and insights that could make the difference.

While products like Google Analytics, Optimizely, and Mixpanel have more-or-less filled the gap with regards to web/mobile traffic analytics and BI reporting, the deeper power of advanced analytics and machine learning — and more generally, artificial intelligence — are still well beyond their reach.

We want to change that. We believe in the human element of data science. To effectively cross the chasm, you need access to expert advice to outline and explain clearly what steps to take to get there.

On the other end, you need a person to interpret and hold accountable for the results, as well as design the go-to-market that makes sense for your business.

To get you started, we present here an analytics roadmap to help you navigate exactly what you need to get to the next level. Let’s cross your #DataScienceChasm together.

analytics_chasm.png

Descriptive Analytics

analytics_chasm_descriptive.png

The hard part of using plug-and-play tools is when things don’t go as planned (e.g. a metric spikes or plummets). It requires knowledge and experience to understand why and to get back on track.

Business intelligence and reporting can tell you what happened, but to really understand why, you will need to go deeper.

Descriptive analytics sits right on the edge of the data science chasm — SaaS tools can solve pieces here and there, but many problems require a seasoned eye. Several useful analyses and techniques include:

  • Seasonality and historical trends
  • Product purchase cycle
  • Promotion and discount effectiveness
  • User segmentation and clustering
  • Funnel conversion analysis
  • Market basket analysis

Large companies will often have many in-house analysts dedicated full-time to the above areas to optimize and accelerate their businesses. At LinkedIn, every product line has extensive user segmentation and purchase funnel optimization.

For example, you are likely to see a different subscription product offered when you check out, based on your profile. For smaller companies, there are no resources for hiring full-time analysts. The chasm needs be crossed by DSaaS (Data Science-As-a-Service) services that provide analyses and algorithms that deliver immediate results, without spending big on R&D.

Predictive Analytics

analytics_chasm_predictive.png

Effectively crossing your chasm into hyper growth is all about knowing where to allocate your resources.

What if you knew exactly which users were more likely to purchase your products? What if you knew the upcoming few weeks were going to have a massively outsized effect on your sales this quarter?

The higher probability you can have of certain events or outcomes, the higher the chance of surviving another year. Some predictive analytics include:

  • User acquisition propensity models
  • Repurchase or churn models
  • User lifetime value predictions
  • Sales forecasting and promotion planning
  • User lifetime value (LTV)

Large companies have even more senior data scientists working on these problems. These techniques can be the defining factor in dominating a market — if you can use data to understand your business and know where to spend your resources, you will be in the top 1% of businesses.

Prescriptive Optimization

analytics_chasm_prescriptive.png

Prescriptive Optimization is the Holy Grail of data science and machine learning — a fully automated, self-learning, low code deployment of AI connected components that react to your business based on real time purchase behavior and action based protocols.

AI handles ordering, shipping, pricing, emailing, and customer service that delivers optimization to all corners of your business.

Some key components you would likely need are:

  • Dynamically updating product hierarchy
  • Purpose driven product categories based on purchase behavior
  • Smart pricing models
  • Autonomous AI that updates rules and flows as it learns

Wherever you may be in the analytics roadmap, there is always room for learning and improvement. The data science journey is a long one but rocketship growth and, ultimately, a unicorn business is a reward worth working towards. If there’s anything we can do to help, please drop us a line!

Special thanks to George Dy, Bruno Wong for editing!


  1. Originally published at blog.tresl.co on March 18, 2018. [1] ↩︎

Discover and read more posts from John Chao (Tresl)
get started