Why DTC brands are turning to first-party data to combat iOS changes - Part 1
Fresh off our Series A, Richard Harris, Black Crow’s Founder and CEO, recently sat down as Interplay VC’s guest on Innovation with Mark Peter Davis, the firm’s podcast. We recap the conversation below, in which they dive deep on:
- The big-picture value of rich, substantial data
- Unlocking the impacts of AI and ML in eCom
- Applying ML to fortify marketing strategies
Applying Machine Learning to eCommerce
Black Crow is a real-time ML platform, meaning we stream or ingest massive amounts of data, process that data in real-time, and produce intelligent predictions.
Our current focus is eCom, where we can predict the future value of a customer a mere 15 milliseconds after they perform an action or event within a brand’s environment.
Reliable predictive modeling, especially in real-time, has historically been a solution with limited availability. So our mission is to make these capabilities accessible to the middle market.
Our target users? People who’ve historically not had access to Fortune 500-grade ML flows.
In order to successfully capture the middle of the market, we knew we’d have to make Black Crow unprecedentedly simple.
As such, our platform takes just one click to install — not a multi-year, multimillion-dollar project.
“While massive players like Google and Amazon can leverage the power of machine learning, most SMBs cannot. Black Crow changes that for good.”
Black Crow’s Prediction Engine: How Does It Work?
The Black Crow engine works in the background of an eCommerce site and analyzes every visitor's behavior and interactions to create a prediction model that determines how likely that visitor is to complete a certain action.
For DTC brands, that action is normally a purchase, repeat purchase, or subscription But whatever the action is, the machine will start identifying patterns and building the model — with zero work on the customer’s end.
After roughly two weeks, you’ll have a high-accuracy predictive model that tells you how likely a consumer is to purchase, subscribe, or complete any other desired action.
Underneath the hood, this process takes four core steps:
- The Black Crow engine continuously monitors about 450 signals
- It parses which of these signals can generate accurate predictions
- Each variable is weighted to predict a user’s future behaviors and value
- These predictions result in countless bespoke predictive consumer models
This is not a one size fits all model. None of the 250 brands we work with are the same, all of their data is uniquely their own, and our predictions are tailored to their customers. These predictions are incredibly valuable to brands because they are happening real time. Before now, brands haven’t been able to take advantage of these types of insights because it was simply so difficult to process.
“When you can predict what users will do and how valuable they’ll be, that changes the game for numerous elements within your eCom business.”
The Outsized Value of Intelligent Data Streams
Every time a consumer interacts with an eCom brand, that data point falls into one of three buckets and answers a unique set of questions.
1. Referral Data
This answers key attribution questions — a must-have in the aftermath of iOS 14.5 — such as:
- How has this user arrived at their current relationship with the brand?
- Where, when, and on what device did their browser arrive at this company?
- Where did they come from — an organic web hit, social media, a marketing campaign?
2. In-Session Data
Once a user arrives at a brand’s landing page, where do they go from there? Questions include:
- How is this user broadly behaving?
- Do they linger on a few images or browse many images rapidly?
- Are those images of the same product or various product types?
- Do they scroll deeply down pages or move quickly through the funnel?
3. Consumer Data Over Time
Third, we consider how the user’s behavioral decisions (and the respective data) evolve.
- Will they return to the brand?
- Do they return to view the same products?
- Are their subsequent visits bringing them further and further into the brand?
Next Steps: How to Take Action
All of these points are crunched together in 15 milliseconds to generate findings such as:
- User A is 65% likely to buy
- Meanwhile, User B is 0.02% likely to buy
Naturally, brands that have access to this type of information can absolutely transform their marketing because they are able to treat Users A and B very differently as opposed to always trying to play to an “average” customer. This optimizes everything from paid media to pricing promotions to UX design to prioritizing their customer service queue.
“You can optimize so many elements of a business if you have this knowledge of the future — and that hasn’t really been a possibility until now.”
Pushing Beyond Demos & Interest Groups
Most companies end up relying on proxies to understand their consumers, such as demographics, psychographics, and interest groups.
Of course, asking whether a buyer is an empty nester or a nomadic millennial provides a degree of context. But these guesses don't drive insights into what the team actually wants to know.
Ultimately, brands want to curate customer relationships, centered around delivering value to a user while that user delivers value back to the brand through purchases, social proof, etc.
As a result, a user’s potential value to a brand should be the primary motivating factor behind almost every decision. And, with the power of Machine Learning, those decisions can be made in real-time.
“The greatest way to add appropriate value to a customer’s experience is to have the hardcore data on how valuable your consumers will be down the line.”
Optimizing Your CAC to LTV Ratio
For our DTC eCom clientele, the number-one use case (AKA the place most users want to direct their predictive knowledge) is a brand’s marketing workflow.
In the aftermath of iOS 14.5, customer acquisition costs have outpaced even product costs as the largest line item on most commercial teams’ P&Ls.
Knowing this, bringing efficiency to your paid marketing flows and optimizing CACs becomes a critical item for your business.
Your Personal, Repurposable Data Pool
For every real-time, Black Crow-driven prediction of a user’s future value, that prediction then becomes its own first-party data unit — which belongs entirely to the company.
Then, because virtually every tool in the modern eCom stack is built to ingest a brand’s first-party data, Black Crow can push this info and these findings right onto your platforms.
For instance, our predictions show up as audiences directly in your Meta ad manager. Now, you can tell two populations yield expected conversion rates of 60% vs. 2% and plan accordingly.
By allocating your marketing spend in line with the hard data on unique cohorts and segments, Black Crow users can expect to improve their ROAS by 25–50% at scale.
Balancing CAC: A Life-or-Death Equation
Black Crow is a B2B tech company. But the next time we pitch to raise funds, we know the first question out of the VC’s mouth will be: “What’s your average CAC?”
Ultimately, if your business (like most businesses) lives or dies by how much it costs to capture and convert a shopper, you need to know how your CACs pay off over time.
Nearly every vertical is affected by these circumstances. Just a few in our network include:
- Consumer software
- Consumer financial services
Due to sheer inaccessibility, most brands have had to consider their consumers as a monolith, drawing hypothetical demographics and calculating averages like a shot in the dark.
From day one, our team has worked to shift the typical eCom team’s capabilities.
Through accurate, real-time predictions from pixel zero for a fraction of the cost and lift, Black Crow rises beyond the industry norm of soaring CACs and a brutal attribution landscape.