“Big data” is the latest industry buzzword, but what do you actually do with it? Does it even apply to your firm? What it all comes down to is knowing your customer, and data just quantifies what you’ve always done by gut feel. Here’s a look at five ways firms are using data science in marketing.
It’s easy to fall into the trap of treating all prospects like potential customers. Make no mistake, you should treat everyone well, but this doesn’t mean spending time trying to win prospects who will never turn into a profitable client. You already know this and probably have some sort of weeding out process.
The problem is how long does it take to figure out whether someone will ever turn into a paying customer? If you don’t weed out early enough, you’ve wasted time and money for no return. If you weed out too soon, you’re losing out on profitable accounts.
Lead scoring gives a way to weed out customers that goes beyond experience or gut feels. It compares the characteristics of a current lead against your previous customers and failed conversions. More profitable and more likely to convert leads receive a high score, while low-profit leads with little chance of converting receive the lowest score.
For example, an auto insurance company might be targeting drivers with a high net worth who need high liability limits. They may also receive inquiries from drivers who purchased a luxury car even though it was above their means and won’t be able to afford the policy. The insurance company can use data science to score leads based on car type, occupation, zip code, and other factors to determine which leads actually fit their target demographic.
Once you have a lead score, it becomes your guide to your next steps. Leads with a high score might receive personalized follow-up calls or more detailed mail brochures. Low scoring leads might be sent into an email drip campaign with no human intervention until the lead responds in a way demonstrating a higher likelihood of converting or being more profitable than initially predicted.
Even leads who you don’t take action on have a cost. The question is, was the cost worth it? The answer isn’t an easy yes because you know that not all leads will convert. Some sources of prospective customers will provide better leads than others. Again, it becomes about making identifying and prioritizing these sources a science rather than an educated guess.
Identifying prospects is lead scoring with a different level of magnification. You’re still comparing potential customers against your past customers as well as leads who never signed on. The difference is that instead of looking at individual leads to decide how to take action, you want to look at where they’re coming from to determine if you’re allocating your resources appropriately.
Your segments might be in entirely different mediums such as radio ads versus online, or they might be narrower subsets such as specific radio stations, times of day, or how they originally entered your website. Your end goal is to identify these segments using data science so that you can focus your efforts there.
One of the biggest challenges is identifying which marketing channels prospects came through. One of the most common ways of finding out is the “How did you hear about us?” question. Of course, this doesn’t always accurately reflect your funnel since customers likely came through multiple channels.
Take a typical personal injury law firm campaign. A client might have been vaguely familiar with the firm from bus ads, seen a TV commercial after getting into an accident, gone to the website given in the TV commercial, checked out social media reviews, used an online chat, and then finally scheduled an in-person consultation.
To figure out which channels customers touched, you’ll need to get multiple points of data. This may include traditional surveys, using unique websites or phone numbers to track ads in traditional TV, radio, and print media, and a full set of analytics tools on online channels. Collecting this data is only helpful if you ensure its accuracy by auditing the attribution data step-by-step. You can do this by simply pretending to be a customer. Mimic the steps yourself to see if the data points collected reflect your journey accurately.
In addition to figuring out which channels were touched, you also need to figure out if and how each channel helped. Is it building initial awareness? Is it progressing towards the close? Is it adding an extra step that doesn’t add value or even that causes people to drop off? With this data in hand, you can choose where to prioritize your marketing spending and better optimize how prospects are handed off between channels before they make the buying decision.
When you’re looking at the data for a specific channel, it’s important to understand whether underperformance is because that’s an inefficient channel for your target audience or because you haven’t properly optimized the channel. This requires testing different approaches within each channel and comparing the result. The web is the easiest channel to do this on.
Multivariate web page testing shows different visitors different versions of the same web page. This might include things like different contact form placements, different versions of the text, or different visual designs. Each version is tagged separately within your analytics data. Once you’ve reached a statistically significant number of visitors during the test, you can see if the different versions perform at different rates. If one version outperforms the others, you know a change is in order. If all versions perform equally poorly, you might prioritize other channels instead.
The problem with experimenting and adapting, even when driven by data, is that it takes time. This includes both your time setting up and analyzing tests and the weeks you may be losing sales to the underperforming variant of the test. This process is still necessary to some extent, but artificial intelligence can take you to the next level even faster.
AI has seen increasing use in bidding for pay-per-click and other online ads. The core principles are the same as your other data models — identify the most likely and most profitable buyers and focus on them. The more valuable the prospect, the higher the bid.
AI does this in real-time through a combination of the historical customer data you feed in as well as the individual ad viewer’s current behavior. In addition to demographics, the AI looks at how a user is navigating through a search engine or website to determine if they are likely to be in buying or browsing mode.
The AI may also look at your live sales data to see if it is rising or falling and adapt accordingly. For example, rising sales might indicate a need to capitalize on a surge, while falling sales signal a need to pull back until market conditions change.
Data allows you to quantify your marketing performance in areas that were historical decided by gut feelings, anecdotal evidence, and experience. This, in turn, will enable you to optimize your marketing tactics to ensure you’re getting a high return on your marketing investment. However, marketing data science initiatives should be tailored based on the size of your organization, size of marketing efforts, and current levels of data maturity.
To learn more about how to make data work for you, contact us for a free consultation.
Sources:
3 Great Examples of Data Science in Marketing: https://www.business2community.com/marketing/3-great-examples-of-data-science-in-marketing-02052176
Data Science is the Key to Marketing ROI: https://www.forbes.com/sites/steveolenski/2018/03/06/data-science-is-the-key-to-marketing-roi-heres-how-to-nail-it/#29eb7b5131c3