It’s a Data Collector & Cruncher, Insights Producer, Real-Time Processer, and Channel Connector – But wait! There’s more!
Since the Dawn of Martech Times – The Goals & Principles Remain the Same
For a lifetime (mine anyways) marketers have sought the holy grail of one-to-one customer engagement: Right Customer, Right Time, Right Message/Offer, Right Channel/Place.
In pursuing that goal, those working for enterprises knew to scale beyond mom/pop audience size they needed big tech help – big data & insights about their customers, at-scale machine learning to calculate what to say and offer, a large & dynamic curated library of messages, as well as direct connections to the ever-growing channels to deliver them – both ones where customers were in channel (inbound) and ones where nudging was required (outbound).
Bringing at-scale tech to this goal started by using big databases. Those databases held customer account and transaction data and used queries achieving the first step – finding sets of customers with differences in customer behavior. And those differences proved to be great insights to train models and predict future behavior. Next, the pioneers matched messages to the predicted behavior segments. A likely buyer of a certain product just needed an offer for that product. Then, send that message, get a response, and chalk up higher conversion rates.
Take, for example, credit card marketers. They segmented their base into 2 main categories: Transactors and Revolvers. Transactors paid off their bill. Revolvers didn’t. So based on this, they offered transactors incentives to transact more so they would increase revenue (from fees charged to the merchants) and offered revolvers more credit – balance transfers / credit limit increases (where they made fees on the growing balances). The results: higher response rates, campaign lift (over controls) and more revenue.
That’s it. For the next 30 years, businesspeople in a variety of industries – from banking to telco to insurance and others – built out these systems and simply sought more data, improved prediction models, and connected their offers to more channels. Sounds simple right? Oh, but it’s not.
Why? Because big data got bigger, messier, and harder (and more important) to carefully manage. Getting the right insights from that data was (and still is) tricky. Establishing data lineage, privacy, and proper governance became crucial. Algorithms to accurately predict intent, and the right action to take next, evolved – and required controls. New AI methods sprung up along the way (and continue to do so – with GenAI another example – we’ll get to that later). And we all know about the proliferation of channels and digital devices.
Martech Stack Ecosystem – Then and Now
Ecosystem wise, not much has changed in the basic framework of a Martech stack since the early 90’s. Figure 1 shows the main components. Data is collected from a variety of sources at different velocities. Some data is distilled before it’s sent, providing insights. The stack itself produces both data & insights, and those are made available to other systems. A plan programs the content, data, and strategies employed by the decision system, and gets informed by the results. AI, both inside and outside the stack, powers predictions. Finally, ranked recommendations and messages are activated, and married (using metadata) with the appropriate content and delivered to channels. There, an orchestration layer may dip down to get actual content (e.g., digital images), and then consumers get the output (and hopefully react positively). Dashboards, reports, and analysis tools help marketers understand the results.
And the goals remain unchanged. Provide personalized experiences. Doing so generates lift (higher response and conversion rates) because more relevant offers are presented to customers who are more likely to want them. It’s not rocket science it’s just marketing science.
Figure 1: Conceptual Martech Stack and Surrounding Ecosystem
At the high-level, yup that’s pretty much it. I’ve researched and witnessed this pattern for 30 years. Attending conferences, following Martech Stackie Awards (2015, 2016, 2017, 2018, 2019, 2020, 2021, 2022, 2023.), reading countless analyst blogs, and working with hundreds of enterprise clients across the globe.
The 10 Immutable Laws of Martech Stacks
So, what have we learned? Since there is an overabundance of data, and technologies of various kinds come & go, lock in on designing the modern Martech stack so that it adheres to principles that have withstood the test of time.
- Collect the right data. You don’t need a huge number of customer behavior attributes, but instead the right ones for the business problems being solved (e.g., reduce churn) and so reflect customer intent and cause and effect with likelihood to respond to offers to solve those issues.
- Make sure collected data is accurate and, in as much as possible, feed it into your Martech stack in real-time.
- Use segments to study common traits and behaviors. Assign segment attributes to customers, not the other way around.
- Make decisions on individuals, not on segments.
- Use adaptive models to calculate “offer propensity.” Establish that these models are learning continuously on data you are collecting.
- Use one set of engagement strategies and rules for inbound & outbound decisions. Do not separate this logic and place it into channel systems.
- When making inbound decisions, send them immediately. Do not cache decisions into channels waiting for a customer to appear.
- With outbound marketing, only send permission-based relevant messages to customers on channels they opt-in and respond to, at times they prefer, and with content that is relevant.
- Use behavior triggers, not pre-set schedules, to determine the right time to send.
- Select the latest content just prior to presenting offers (e.g., versions of your offer, that include creative and language variation tests).
And here are a few more pointers:
You need a few good foundational software platforms (linchpins) that integrate, not 10k technologies (https://chiefmartec.com/2022/10/why-there-are-10000-martech-products-that-kinda-all-do-the-same-thing-but-not-really/).
Which ones? Follow this basic advice for the 4 main ones you need, and that must operate well together:
Compare your design to others that have been successful. Here is a 2023 stackie winner. Notice the biggest bubbles: Content, Execution Platform, Analytics (Insights).
And here is another, centered on using data & AI to power a brain to make decisions during the customer journey cycle (awareness, consideration, decision)
Don’t drop what you are doing to chase the latest fad. In other words, don’t fall victim again to the “shiny new object syndrome.[i] ” Stay the course and be sure to devote some of your tech budget to innovation testing (maybe 10%). Hopefully, you were already doing that prior to the GenAI hype setting in.
Speaking of GenAI, test it to see if it helps with Law #5 (finding features for models to learn on) and Law #10 (challenger tests for creative & language variations)
Old saying but it holds true: You can’t manage what you can’t measure. Measure your program effectiveness by looking at champion creative, promotions, and messaging, and then try new variations, and measure again.
The evolution of the Martech stack over the years has brought about significant advancements in data collection, insights generation, interoperability, and customer engagement. The fundamental goals and principles of delivering personalized experiences to customers remain unchanged. However, as technology and data have become more complex, it is essential to adhere to the 10 immutable laws of Martech to provide the right ingredients for success.
Collecting the right data and ensuring its accuracy in real-time is essential. Segment customers to understand behaviors but make decisions at the individual level in real-time. Use adaptive models that continuously learn from data and help calculate offer propensity accurately. Inbound and outbound decisions should follow unified strategies and use behavior triggers for timely engagement. Outbound marketing should focus on sending relevant messages on preferred channels.
Build a Martech stack with a few foundational software products that provide insights, content management, and decision management. The emergence of GenAI offers opportunities to enhance model learning and conduct challenger tests for creative and language variations. However, it is important to test and measure its effectiveness before fully adopting it into the Martech stack.
Finally, measure program effectiveness and iterate with new variations for continuous improvement. By following these principles, businesses can achieve personalized customer engagement and drive higher response and conversion rates.