Modern marketers are looking for better ways to identify and reach the right consumers. Many of them just don’t trust their metrics anymore, given all the business changes of the past two years. The cookies and third-party data that drove targeting innovations in the past have limited value today, and the cookieless future will be here soon, as Google, Apple, Firefox and others prepare for a phased exit of third-party cookies next year.
AI Startup News: Promoted.ai Raises $6M to Transform E-Commerce Businesses using ML Infrastructure
The good news is all this uncertainty also opens up an opportunity to uncover a much more accurate way to identify current and future high value customers – using Predictive Customer Lifetime Value scoring, which relies on sophisticated machine learning statistical algorithms to measure future value of each individual customer based on the hundreds of variables already stored in the transactional and behavioral customer data. By adopting new AI-based technologies, companies can move from the history of business intelligence to the future of predictive intelligence, and optimize their marketing spend while having a bigger impact on revenue by focusing on the right customers.
More importantly, with the constant pressure to accelerate business growth (in good and challenging economic times), there’s a competitive pressure to stay ahead of the pack when it comes to marketing technology. After all, as the adage goes, you don’t have to run faster than the bear… just faster than your competitors.
AiThority readers likely understand the potential business opportunities from more advanced AI technologies in marketing. But many companies are still relying on descriptive BI approaches to develop frameworks such as RFM (recency, frequency, monetization) to identify their best customers. Based on a small, pre-defined subset of attributes and simplistic human-based heuristics, the framework then assigns a score to each customer, and entire marketing budgets and customer journey experiences are prioritized based on these scores. This was sufficient for many years, but these scores are not nearly as predictive in nature as machine learning models are.
Part of the challenge with relying on historical data is that consumer buying patterns from 2014 or even 2019 are not really relevant for today’s market – so much has changed since then. Relying on traditional analytics approaches could send you down the wrong path and actually hurt your business because you’d be spending your limited resources on the wrong customers and paying little attention to the right ones. Chasing the wrong customers will have that bear clawing at your heels in no time.
An economic downturn also puts new pressure on shrinking marketing budgets while trying to grow customer value. By utilizing more advanced AI and predictive analytics, you can either A) reach the same number of consumers while spending less money, or B) use that same budget to reach more higher-value customers and grow revenue.
The opportunity here is that if you can deploy predictive analytics in the right way, you can pull precise, future-informed insights that the business can use, using a relatively small set of recent data from the past few years, or even weeks.
With this approach, you can make meaningful and accurate predictions even when the world is changing rapidly.
When I say predictive analytics, I mean the ability to look at specific customers and run predictive queries against their behavior, so you can see the likelihood of a customer from the last 30 days becoming a VIP-level customer within 90 days. Or you can look at similar audiences to try and determine who is most likely to churn in the next X number of days or weeks, possibly before they’ve even taken the first step of unsubscribing from your newsletter. You control the variables – days could be weeks; a VIP customer could be someone who has a certain purchasing frequency with a minimum or average dollar transaction value of Y; and a churned customer could be someone who hasn’t logged-in during the last couple of months or hasn’t made a second purchase.
For businesses with high transaction volumes, one of the big advantages of using predictive strategies is that it’s possible to get solid results in a matter of weeks. Now this isn’t instant – if you’re looking at who will buy a new car next month, you’ll need more than a month’s worth of data, given that most consumers buy a car every seven to eight years. But the advantage comes when you use the BI data you’ve collected for years, and layer on top recent customer behavior data. Then you can apply the 80/20 rule and focus your marketing efforts on the 20% of customers who drive 80% of sales. As for mobile games, only about 5% of users make in-app purchases, so focusing on those customers who’ll make repeat purchases is key.
The core advantage of individualized predictive scores is that their values – whether and when the customer would be churning, their future LTV or upsell likelihood – are then shared directly with your CRM, CDP, marketing automation or attribution solution. These scores work with your business definitions and your customer journey KPIs and deliver the most impact when they are implemented into your stack and workflows in support of a creative, customer-centric, data-driven marketing strategy. Make no mistake, marketers still need to do what they do best – create differentiated marketing messages and design campaigns promotions that entice consumers to take an action.
Then, predictive analytics can target these campaigns to deliver double-digit improvements in retention of VIPs, upsell campaigns, return on ad spend and ARPU.
AI in Cookieless:
AI Company Decide Expands in the US to Lead in the Cookieless World
Benefits of predictive analytics for future-minded, proactive marketers
There are several benefits to moving to predictive analytics for marketing. One crucial first step is to define the business problem that you’re trying to solve – is this related to CLV (also known as LTV), acquisition, churn, or other similar customer-journey metrics? Once you have a clear view of the question you’re trying to answer, you’ll have a much better sense of what data you need to use to get there.
Here’s what to expect from predictive analytics:
-
Predictive analytics are designed to give marketers precise foresight – what each consumer is likely to do within a defined period of time, instead of describing what they’ve done in the past as whole, average or a population sample. With how much consumer behavior has changed over these past years, informed foresight is better than 20/20 hindsight.
-
Finding the right message is more important than ever because consumers want marketing that speaks directly to them. McKinsey found that 71% of consumers expect personalized interactions, with more than half anticipating to be sent targeted promotions and triggers based on their behavior. No wonder 76% of consumers get frustrated when this doesn’t happen! These kinds of stats make it clear that advanced personalization is no longer optional – and aided by predictive analytics, personalization can actually become truly personal.
-
Executing predictive analytics using first-party data will prepare marketers for the inevitability of a cookie-less future. Because ready or not – the cookieless future is coming in 2023. Surprisingly, when assessing the state of readiness, 81% of companies admit that at least half of their data was third-party. In stark contrast to that, 85% of consumers want brands to only use first-party data when delivering personalized experiences. By using predictive analytics, you can capitalize on a smaller data set to derive granular consumer preferences – and you don’t even need to disclose or use any PII. You can also fairly easily incorporate multiple data sets to create a more complete picture of each customer and their likely future behavior.
-
Predictive analytics give marketers the ability to execute with precision based on hypergranular segments that are built on each user’s past and probable future behavior – something that descriptive analysis simply can’t do, even when aided by machine learning. In good times, and especially in times of uncertainty, the ability to hone-in on the right segment of customers gives marketers the much needed resilience muscle empowering them to be precise and proactive – so they can make well-informed decisions regarding their programs and spend. Now they can make truly data-driven choices about which customer segments are worth the ad spend, the promotions or discounts, or any special treatment.
-
Predictive analytics give marketers the ability to experiment and pivot spend strategically. This point may be one of the most important ones in a shifting world. Setting up a campaign, running it for weeks, then looking at the results six months later isn’t useful for today’s world. Marketers need the ability to course-correct the day after launching a campaign, so they can target the best customers without spending a fortune to see the output of the test. Learning on the fly and adjusting accordingly is key. When your campaigns are orchestrated using predictive scoring data, they will self-adjust and self-optimize based on the customer value.
Looking forward, I expect that significant changes in consumer habits are likely to continue. There isn’t one reason behind it, but numerous ones, including the lingering effects of the pandemic and supply chain disruptions, changing economic conditions, social media’s impact on consumer purchasing habits, and growth in popularity of D2C brands and subscription models.
Approaches to marketing will need to shift to meet these changing consumer demands and ever-more competitive market conditions.
Predictive analytics can give companies a chance to use yesterday’s and today’s data to predict the future. AI can identify numerous patterns not visible to the naked eye nor influenced by flawed human logic, enabling marketers to reach the customers that will drive the bulk of revenue, utilize marketing budgets in a much smarter manner, and outsmart, and outrun, your competitors.
Comments are closed.