The holidays are already upon us, promising nearly $731 billion in retail sales in November and December. It's the single most critical time for brands to take advantage of seasonal shopper behavior and drive campaigns that effectively compete, capture, and convert customers at high quantities.
But many retailers miss the mark because they fail to understand a key way to secure a successful season ahead: shifting their machine-learning strategy.
Shopper behavior is fundamentally different this time of year, which means marketers can't leverage the same data and algorithms they did back in August, or even as recently as October.
Lack of preparation and adjustment could cost brands big over the holidays unless they make three changes.
1. Update your algorithms to accommodate behavioral shifts
I say this time and time again: AI at its core is just math, sometimes complex math, and it's up to analysts using the right data to make it worthwhile—especially during the holiday season. Machine-learning algorithms don't have common sense; completely without hesitation, they use and believe the data you provide them.
Customer behavior is very different during the holidays: You'll get an influx of shoppers who don't engage any other time of year, customers will look at particular brands and categories they haven't before, they'll purchase quicker than they usually do, and the list goes on. Plus, everything about their behavior is amplified right now: the number of times they visit and abandon your site, the amount of products and categories they look at, the number of items they're wish-listing and adding to their cart, etc.
Those shifts in engagement are essential to understand and cater to, yet algorithms don't know any better. The only way they can know to read those behaviors is if you tell them to. That requires a close alignment between marketing and data science teams; they must collaborate on what the models should be processing this season, and apply the updated models to the several weeks that comprise the holiday season.
2. Know how your marketing campaigns will be impacted
Since your year-round machine-learning algorithms won't apply to the holidays, you'll need to make some adjustments to your campaigns, too.
For example, your "highly likely to purchase" campaigns will look very different—because even though someone is on your site a significant amount over Black Friday and Cyber Monday weekend, it doesn't mean they are more likely to purchase. Although that type of interaction would traditionally trigger your algorithm to think those visitors are highly likely to purchase and therefore suppress certain promotional offers, you'll want to do the exact opposite during the holiday season and showcase those offers whenever those potential customers are showing an increase in engagement.
Your disengaging customer campaigns will be affected, too. Traditionally, a customer will spend a certain number of hours or days with your brand to determine what they want to buy or whether they want to buy. During the holidays, that period gets crunched down to one or two days: They want to make decisions quickly and they are price-shopping and comparing. During the holiday season, outreach aimed at disengaging customers should be triggered when they haven't been on the site in hours, not days.
3. Prepare your data for post-season and other holidays, too
After the holidays, keep the data set separate. You don't want to weave the set specific to the holidays into your regular campaigns; doing that will bias your models toward holiday-specific behaviors.
Your usual repeat and loyal shoppers, for example, will likely come to your website to buy for others, so what they will be browsing and buying during the holidays may be completely different from what they typically look at the rest of the year.
So carve out the holidays as a separate behavior period altogether, and afterward revert your models back to the more typical patterns of behavior you see from your customers throughout the year.
However, you should update and leverage these holiday models for other key shopping and gifting holidays, too. Valentine's Day, Mother's and Father's Day, Back-to-School, and, most important, Cyber Week 2020 are all time periods you should consider building and deploying specialized models for based on what you've learned this holiday season.
* * *
In general, holiday shoppers are in a hectic mood and in a time crunch, forcing them to make quicker decisions, abandon their typical browsing and buying behavior, and do more price-shopping than any other time of year. Your seasonal algorithms and campaigns should accommodate these shifts, staying hyper-focused on recent (vs. historical) engagement.
If you haven't already done so, make sure your marketing and data science teams are on the same page, and update your machine-learning strategy accordingly.
Take the first step (it's free).
You may also like:
- Chin up, Marketers: The Demise of Third-Party Cookies Isn't All Bad
- How to Marry Offline and Online Attribution Data for a 360 View in Google Analytics
- How B2B Marketers Can Absorb and Apply Data Effectively: Six Questions Answered
- Why Customers Take Brand Surveys
- 10 Tips for Automating Marketing Reports