Small E-com, Wanna Know Why You’re Failing?
Here’s why small e-commerce marketers need to get with the predictive analytics game. In auto racing, drivers get the opportunity to walk the track before racing begins. New, inexperienced drivers may spend the walk talking or thinking about how good they’ll look coming out of a turn. (If they walk the course at all!) But the veteran drivers treat the opportunity completely differently. They don’t talk to others and may even wear headphones to block all distractions. They take their time, stopping at each turn to memorize as many details as possible.
See, the veteran drivers understand the importance of knowing what’s coming next. When they race, they know what’s after each twist and turn. This allows them to apply the perfect amount of throttle, steer in the right direction, and apply brakes precisely when needed. Without knowing what’s next, a surprise turn catches many inexperienced drivers off-guard.
You may already see where this is going. Predictive analytics is the e-commerce marketer’s equal to walking the track. Do it right and you steer your business in the right direction. Ignore predictive analytics, or underestimate its importance, and risk being left in the dust. And remember, just like walking a course, there’s a right way and a wrong way to embrace predictive analytics. Here’s everything you need to know:
The Predictive Analytics Boom and Its Impact on E-commerce Marketers Today
Steadily but surely, predictive analytics has been gaining in popularity. It’s expected to explode this year as it becomes more accessible and goes from being a competitive advantage to a necessity. To fully understand how it impacts e-commerce marketers today, first we’ve got to look back to the past.
Traditionally, predictive analytics has been reserved for top brands such as eBay and Netflix. eBay is focused on using predictive analytics to make its shopping experience better for both buyers and sellers. Indeed, eBay considered predictive analytics so important that it acquired SalesPredict to boost its AI, machine learning and data science efforts.
When a company like eBay spends millions on an acquisition for the sake of predictive analytics, e-commerce marketers must take note.
There are a couple of reasons that eBay made this move:
- Better understanding of what their customers wanted.
- Access to advanced insights for improving conversion rates and accelerating sales cycles.
- More targeted offers with relevant information for buyers.
- The ability to build-out predictive models that can define the probability of selling a given product at a given price over time.
In other words, predictive analytics answers some of the biggest challenges facing e-commerce marketers and business owners.
How to Use Predictive Analytics Like eBay Does
To achieve similar benefits to eBay, you only need to take the following four steps:
- Track e-commerce metrics and create a one-day forecast. Then identify any deviation between the forecast and the actual metrics.
- Determine the cause of deviations from your one-day predictions. Assuming the deviation can be minimized, take steps to do so; otherwise, use it to establish a confidence level in your predictions.
- Create progressively longer-term predictions, ranging from one day to 18 months.
- Use predictions to answer “what if” questions such as, “What if we increase advertising spend?" or “what if we decrease stock for this item?” and more.
Obviously, it’s easier said than done, but with the proper tools and strategy, you can be predicting the future in no time.
Overcoming Common Challenges
Predictive analytics isn’t without its challenges. For one, it can be difficult to trust the data, especially when your “gut” and past experiences may indicate otherwise. But even more challenging is doing predictive analytics right.
Ensuring Data Quality and Managing Data From Multiple Sources
Remember how the first step of using predictive analytics involved creating a one-day prediction, then testing that prediction’s accuracy? That is the key to ensuring quality data. Start small with predictive analytics and only move up to more complex/long-term predictions when you’re ready for it.
At first, your data may make a significant impact (especially if you’re still in the one-day prediction phase), but taking the time to validate predictions will ensure that your data won’t lead you astray. Validating your data is a neverending process, as well. You’ll need to put a system in place to routinely check metrics and ensure they are accurate.
One of the other common challenges is managing multiple data sources. Difficulties in multiple data sources can be minimized by utilizing a BI tool that integrates with hundreds of sources. This will ensure that all your data is kept in one place and can be easily accessed and analyzed.
Overcoming Predictive Analytics’ Specialized Nature
Predictive analytics is a highly specialized field. Because of that, it can be difficult for some marketers to see themselves using it. It helps to have the right goals in mind.
A tried-and-true method for goal-setting is the SMART (Specific, Measurable, Attainable, Relevant and Time-bound) method. This method involves establishing large goals, such as the following example:
- Increase Revenue to XXX by December 31, 2017.
Once you’ve got your big goals set, you can break them down into smaller goals, which you’ll use to fuel your predictions. These smaller goals will actually drive the business toward its larger goal.
So, for example, we’ll look at some of the smaller goals necessary to increase revenue:
- Increase customers acquired through pay-per-click to 100 a week
- Increase website traffic by 20,000
- Reduce cart abandonment by 25 percent
- Increase site conversion rate to 5 percent
Now that you’ve set your smaller goals, you can take a look into the predictive side. You may already do this with PPC, where you predict — based on your current conversion rate, cost per click and more — how many more customers you’ll get if you increase ad spending.
Related story: 98% of E-com Shoppers ‘Deterred’ By ‘Incomplete Content’