Blog post

Experiment!? In this economy?

Experiment!? In this economy?
Written byNathaniel Rounds
Published16 Feb 2023

When A/B testing can’t weather the storm, automated experimentation will right the ship

While experts debate if a recession is coming, inflation remains stubbornly high, and in some industries, layoffs loom. As consumers try to stretch their thin dollars, businesses tighten budgets and prepare for a potential storm. Yet marketers have the same mandates as ever – acquire customers, retain them, and engage them. And as ever, marketers need to experiment to validate their strategies empirically. No matter how brilliant a campaign, email subject, ad, or creative seems on paper, the only way to know if an idea works is to try it out with customers. 

Traditionally, marketers have tried out their ideas through A/B testing. But A/B testing is slow and doesn’t scale. An A/B test captures a snapshot of the market at a particular moment in time. In turbulent times when customer behavior could be rapidly changing, the results of an A/B test are likely based on data that is already out of date. Savvy marketers have begun using a new application of machine learning, called reinforcement learning, to automate the process of experimentation. 

Why might uncertain economic times be the perfect time to invest in automated experimentation? Let’s look at 4 reasons.

1. Time to focus on existing customers

No business can survive without new customers, but now’s not the time to overlook retention. Customers worried about inflation or recession are looking for products to prune or subscriptions to cut. The good news is that marketers have everything they need to keep their customers from churning.

  • Comms to existing customers are cheap. Marketers own the channels (email, direct mail, SMS, push notifications, etc.) for existing customers, whereas they have to pay for ads or SEM to reach new prospects.

  • Data on existing customers is rich. Marketers have a wealth of first-party data on their identified customers – what they’ve bought, how they’ve engaged in-app or interacted with emails.

It may seem difficult to devote scarce resources to existing customers when the need for acquisition looms. But good retention marketing increases the value of each new customer. As marketers monetize the customers they have, their business case for acquisition spend grows stronger. The greater the CLV (customer lifetime value), the more valuable each marginal customer becomes. Higher CAC (cost to acquire a customer) is sustainable if marketers are confident in the value of that customer.

To put it plainly: there's no point in pouring water into a leaky bucket. That means investing in lifecycle marketing. An Automated Experimentation Platform, like OfferFit, can help marketers leverage their first-party data to optimize communications to existing customers.

2. You need to be relevant

When consumers are cautious with where they're spending their money, it's all the more important to be relevant. When the economy is running hot, marketers might get away with sending the same emails with the same offers to every customer. When consumers are more cautious, marketers need to be more targeted. And remember, “personalization” doesn’t mean putting the customer’s name in the subject line. To engage their audience, marketers need to reach each customer with the right offer, through the right channel, at the right time of day and day of the week. 

A marketer might run A/B tests to find the time of day, or day of the week, that best engages the majority of their customers. But knowing that, on average, more customers open an email on Wednesdays doesn’t tell you the best day to contact each individual. To make 1:1 decisions at the level of individual customers, marketers need automated experimentation.

3. Discounts are dangerous

As marketers fight to be relevant to cautious customers, they tend to lean on discounts and promotions to maintain revenue.  But discounting can be incredibly painful for the business, and not just in the short term. Customers become “anchored” at lower prices, and whether or not the economy dips into a recession, the business is stuck – customers have been trained to only buy when they see the discount. If you don’t have a 25%-off coupon, the customer isn’t interested.

Rather than blast discounts indiscriminately, a better approach is to target them so as to maximize margins. An Automated Experimentation Platform like OfferFit can find the best promotion for each customer, so that marketers only offer what they need to offer to secure the sale. Making 1:1 decisions means better margins now, and puts marketers in a better position exiting a potential recession.

4. To do more with less, look for measurable ROI

In a macroenvironment where businesses are cautious, marketers are facing capped budgets, hiring freezes, or even layoffs. With those headwinds, it can be hard to build a business case for new investments. Marketers are increasingly focused on financial metrics like purchases, ARPU (average revenue per user), or CLV. New initiatives, including investments in machine learning, only make sense if they help move the needle.

Marketers can clearly measure the financial impact of an Automated Experimentation Platform. Reinforcement learning algorithms, like those that power OfferFit, work by seeking to maximize a target metric for each customer, and marketers can directly measure the performance of an AI-driven campaign against their business as usual.

Today’s lifecycle marketers have rich first-party data in their data warehouses and CDPs, and are under increasing pressure to deliver value from those investments. Experimenting to discover the best 1:1 decision for every customer leverages the tech stack a marketer already has.

At OfferFit, we help marketers experiment to maximize the metrics most important to them. Don’t settle for vague initiatives with wishy-washy ROI. Ready to learn more about automated experimentation? Read our whitepaper or schedule a demo!

Experimentation unleashed

Logo

Ready to make the leap from A/B to AI?