Tl;dr - Wondering about whether you should invest in AI or machine learning for predictive intent capability? Probably not. You've got a more powerful engine - what you know. That beats what some algorithm guesses. You probably just need the right data stack, with the right marketing technology, to build your own superior solution today!
The Power of Prediction
Everyone wants to know now.
Which deals will close and when? Which companies will become customers? Which people will join the complex buying team, and what will they each lobby for?
If you knew now, what you'll know (or think you do) later, you'd forecast more accurately, invest in marketing in the right places (with the right message at the right time for the right people) and focus sales resources where they'd make a difference.
And if you knew all this in advance, before a project became active....well, you'd be omniscient.
That's a seductive prospect. And companies use various approaches to get closer. Common techniques include:
- Ideal customer profiles (ICPs) that define the details of accounts that characterize successful customers. Beyond industry, size, geography, stack rank in their market, profitability, technographics, and funding status, this can also include more subtle details such as management pedigree, key investors, key customers, key roles staffed (or unstaffed) and others.
- Buyer personas to define the characteristics of common members of buying teams by function, role, and perspectives including Priority Initiative, Success Factors, Perceived Barriers, and Decision Criteria. (see Adele Revella's Buyer Persona model for more.)
- Lead scoring can be based on account and individual criteria of fit and activity for active leads, and used to prioritize current known prospects.
- Propensity to buy models which mine prospect lists for companies which have key similarities to closed:won deals and prioritize accounts for active outbound prospecting.
These all can provide some value to help filter out unlikely prospects.
The allure of AI & ML
Those approaches often take a lot of work and don't always demonstrate high enough correlations to inspire total confidence. When they miss, the assumption is that there are details which were overlooked and which, could they be identified, would make the models more accurate.
Those are precisely the challenges artificial intelligence and machine learning promise to ease. That's wonderful in theory, but problematic in practice for many mid-size company marketing and sales applications. Not only might the data sets be too small to really, fully massage, but the process is also subject to the biases of the algorithm. For instance, do the models consider commonalities in education and employment background of the sales person and the deal champion and decision maker? How do they account for the effectiveness of the sales person vs. other factors? And how effectively will it interpret the causation?
If companies typically purchase a specific type of software shortly after receiving Series B funding, how can we discern whether the purchase is correlated to receiving that funding or to anticipating the expectations for Series C scaling? Or maybe that's the point in growth that a certain role is added which needs certain incremental software capabilities. (You might argue that it's immaterial - that the Series B funding is the trigger to track. That's partially true, but misses the motivation / problem to solve / outcome to achieve that drives the sale and dictates successful messaging and sales approaches.)
There's no doubt that these are powerful technologies with enormous potential for sales and marketing. They are improving quickly, and some large companies have the scale of available data and resources to experiment productively. For most rapid growth companies, however, the results don't approximate what's promised. And for many, the cost and skillsets remain significant barriers.
That's the bad news.
The good news is that there's a whole range of new tactics and insights available with some common sense approaches in the meantime.
The Promise of Intent Data
Purchase intent data teases explicit predictive capability. It can help identify who, from which companies (at least in the case of contact level™ intent data) is taking action online that indicates they are researching solutions and competitors which indicate they're a net new prospect, a churn risk, or an upsell/cross sell opportunity.
It can also identify additional buying team members, highlight specific competitor engagement, correlate activity to stage in buying journey and more. Those insights can help to predict what messages would resonate most.
This information can also feed other models. For instance, if over time a company observed that a typical buying team includes members of various seniority from the marketing function, mid-level IT members, senior finance and senior sales, then it could predict that an opportunity isn't qualified unless a sales person has identified the corresponding players for a specific opportunity. Further, if a correlation could be established between success, and whether the mid-level IT contact only engaged with competitors, or worked directly with your own product marketing experts, that would be helpful.
Too often intent data is perceived as simply "leads" and the enormous value as marketing intelligence and sales intelligence is overlooked.
Combining Predictive Technology with Intent Data
To distinguish themselves amidst the ubiquity of account level surge type buyer intent data, some companies have added overlays of scoring and propensity to buy models on intent and market data.
This is an appealing option because it promises to help marketers realize the value of intent data easily.
The actual value is somewhat opaque, however, obscured by discussions of algorithms and .ai domains.
The buzz heavy lingo also distracts from the immediate challenge and opportunity.
Companies don't need a science project. Companies need to engage with the buyers who they can most likely help through the buying journey and with the best product/service fit.
Prediction is a guess.
Anticipation is far more important than prediction. And the ability to anticipate and react is much more a function of the tech stack than of neural networks and powerful machine learning. While most companies still face a long wait until AI and machine learning are practically able to extract amazing nuggets deeply hidden within data, they already have the opportunity to substantially improve marketing and sales outcomes by simply surfacing and integrating relatively straightforward insights.
Anticipation and Orchestration
What is the real goal? To incorporate intent data insights into other marketing and sales data so that resources are optimally allocated and tactics anticipate the buyer's process. For example:
- Prioritize sales targets / ABM accounts based on indications of growing, significant activity
- Identify unknown accounts that match key profile characteristics and where activity momentum is growing
- Evaluate inbound leads for likelihood of close based on holistic market wide view of activity among the buying team
- Compare pending opportunities against benchmarks for buying team participation and activity
- Identify early warning of customer churn
- Pinpoint up/cross sell opportunities
In each of these cases what we really want is to combine lots of information, efficiently focus on what's relevant, and streamline our actions in response.
That doesn't take AI or ML.
What we need is the right complimentary information, the ability to sort through large volumes according to rules which will help identify what's material, and automation to facilitate awareness and to prompt or automate action including real personalization at scale.
That's a data and marketing technology challenge - one that marketing teams can solve on their own without investing in or awaiting current or potential AI power.
How do you go about it? There are really only five basic steps.
- Decide which data sources you need - this typically includes various first-party marketing and sales data sources, transactional information, third-party intent data, and enrichment info. (more detail below)
- Unify the data - beyond simply syncing, you'll need to really consolidate it and create a robust, accurate single customer view. And if you're not sweating multi-email address and multi-jurisdictional compliance....well, let's just say you probably should be.
- Enrich and validate - Omnichannel orchestration will require insights (for instance technographic information) as well as accurate email and physical addresses (wait! you're not incorporating direct mail??!!) and other details. That means the single customer view must be "fixed." Don't waste your BDRs time or compromise your campaigns with busywork and silly mistakes (like sending email at a contacts peak delete time!)
- Analyze and decide - The point of doing all this is to discern where companies, buying teams and individuals are in their buying journey and then anticipating what tactics will help them. (Yes, ultimately you hope it will help you too - like via a sale - but if that's your starting point you'll likely flounder.) That means you need the rules and analysis engine to reach those conclusions at scale and automatically.
- Orchestrate - Then you have to make the right things happen. The right message, right time, right person, delivered in the right way. Doing that at scale and personalizing (not just tokens, but content and experiences) is a huge amount of work. Taking it a step further, and virtually coaching your sales team on how to optimize their interactions and leverage your carefully crafted enablement content - well that adds yet more complexity and work.
Here's the thing. To reach this point requires no AI. There's no machine learning necessary.
That's good news. No mystical capabilities required. All you need to do is really execute what you already know you should, incorporating a broad range of signals and insights, and do it at scale.
There's a barrier though, that you've probably already recognized. Your MarTech stack, built typically around a CRM and Marketing Automation Platform, probably can't take you beyond step one above.
Gulp...you will probably need a customer data platform (CDP.)
Let's dive deeper
Data Sources
Let's not worry about predicting. Instead let's look at what's actually happening. To do that we'll need information about activity at the account level (where there's aggregate activity), and at the individual level (to make sure the activity is actually indicative of a project with the right people taking the right kind of action.)
First we'll pull this from your first-party buyer intent data. Let's include:
- anonymous visitors to your site from accounts which you can identify using reverse IP technologies
- known user activity including page visits, downloads, conversions, return visits
- email opens, clicks, bounces, unsubs
- engagement with your social profiles (and let's include brands, corporate and key execs)
- CRM and sales activity and opportunity info
- ABM engagement and ad platforms like Engagio, DemandBase, Triblio and Terminus
- Other elements of your tech stack - Drift chat for instance
- transactional details from your finance and ecommerce systems
- customer success insights (app use, tickets, etc.)
Of course your website is only one of approximately 2 billion around the web. Would what's happening everywhere else that might be helpful to know? What companies are researching software in the category you sell, for instance, as G2 data would provide? And you'll need insights from well beyond their site alone - not to mention knowing who's taking the action - so you can effectively analyze and decide later. Therefore you'll also need third-party purchase intent data which should include:
- pertinent account level signals
- contact-level™ intent data to provide granular contextual detail around the types of engagements happening and who (including role and seniority) are taking them
- contact details to support custom audience creation and outbound sales efforts
You probably know which roles and seniorities are part of the 10.2 person buying team for products/services like yours. You'll want to be sure to surround the account as you develop a project, and not all of the key players may appear in your first and third party data. So you'll also need:
- sales contact database access (e.g. ZoomInfo or LinkedIn Sales Navigator)
That's a lot of sources! And they'll all live in different systems. In some cases you'll be charged by the contact count. In others by API calls, etc. And while many will sync with each other, none will provide a true "Golden Record."
So next....
Data Unification & Hygiene
All the data will have to be consolidated and integrated.
That sounds simple, but it's not.
Neither your CRM nor your MAP are capable of this deceptively complex task.
Creating a true, accurate single customer view which will support orchestration at scale - not to mention compliance - requires a purpose built solution. That's typically a CDP - yet one more layer in the tech stack, and one which can be comfortably managed by the marketing team.
In addition to deduplication and data accuracy, you'll need to understand and honor each individual's preferences for channel of communication, time of day and more.
Further, the multitude of emerging data and privacy regulations will demand much more granular control that many companies can manage. A simple case is the common opt-in, unsubscribe, cookie acceptance, and right to be forgotten challenges with B2B contacts who have provided both personal and work email addresses.
This is potentially daunting. It's likely impossible with your current tech stack. And it's only the second step. After you've created your single customer view you'll discover lots of gaps. Now it's time to start to fill those.
Enrich and Validate the Data
You'll definitely need to know more about the accounts and people who are involved in projects. These will include:
- technographic - do they use software that's complimentary or competitive to yours?
- firmographic - have they recently received another round of funding or opened a new office? Is there a new CFO? Has their largest competitor recently experienced a significant event? Are they aggressively hiring for roles that indicate an emerging need with which you can help? How large are they by revenue? Employees?
- demographic - you'll likely need email and physical addresses, perhaps mobile or direct dials, and job history details for each contact
That will mean access to tools like Datanyze, Clearbit, Owler, Zerobounce, and others.
Subscription access is the easy part here. Important but technically challenging (depending on your MarTech) will be the automated process to validate and augment your data. Don't cut that corner though - otherwise the other work, investment and expertise will be potentially squandered on a foundation of mediocre data quality.
That strong data foundation is going to be important as you start to crunch it in the next step!
Analyze and Decide
Here's where it gets juicy. And here's where your "machine" is going to do more, with much less, than some grandiose AI implementation.
First, define what circumstances would get you excited. Criteria that I'd suggest incorporating into your model include:
- 1st party - # of known contacts from the company taking action (considering seniority & job role)
- 1st party - types of action (form conversion, video watch, chat, email open/click, page visit, meetings)
- 1st party - stage in buying journey that can be understood from the pages/conversion/chat topics
- 1st party - number of anonymous visitors from the company
- 1st party - opportunities, pending or past
- 1st party - event/webinar attendance
- 1st party - ad engagement
- 1st party - social engagement
- 1st party - rolling velocity of engagement
- 2nd party - any sales channel partner engagement
- 3rd party - # and nature of competitor interactions
- 3rd party - rolling velocity of account engagement (surge indications, number of contacts, etc.)
- 3rd party - key buying team roles engagement (number, seniority and function vs. what you know constitutes a "proper" buying team)
- 3rd party - specific competitor engagements
- 3rd party - details on stage in buying journey, problem to be solved, outcome sought
- 3rd party - hiring signals for key roles that demonstrate resource allocation toward the areas you address
- 3rd party - social engagements with competitors and industry thought leaders
You'll likely have more, but this will get you started.
Before we move on, let's answer the question you're probably asking. Isn't this just lead scoring?
Looking just at your first-party data, it will be superficially similar. You'll quickly move beyond contact lead scoring though. The intersection of account and contact activity will be really important - much like you might be doing with ABM engagement software.
Next you'll have to decide on criteria. Examples include:
- What's a reasonable rolling time frame for observing trends? One day? week? month? quarter? It will depend on the sell cycle in your industry and the buying team observations you consider.
- What's a critical mass of total contacts engaged across all data?
- What's a critical mass of engagement activities? With content? With competitors? With industry thought leaders?
- What roles/functions/seniority must be engaged to trigger specific marketing/sales plays?
- What stage in buying journey/problem to solve/outcome to achieve will trigger specific marketing/sales plays?
- What competitive engagements will trigger specific marketing/sales plays?
- What roles and what kinds of engagement, at which stage in
There's a trade off here between simplicity and effectiveness. Think of it as an enterprise sales person. How will each additional insight change your approach? If it won't make a difference in how a great sales person would approach the account, then it's not worth worrying about. But if they'd say "Oh...now that's interesting..." then it's probably worth considering in your rules.
It's key to keep in mind that you be able to run these automatically in the background, and that you microsegment based on the cumulative data - account and contact fit and activity.
With analysis done, decisions made at scale in the background, and segments dynamically created, now...finally...it's time to get to work. Let's orchestrate amazing results!
Orchestration
Knowing is powerful. When you know more (various data sources) and can see it clearly (unified, verified, enriched data) and automate the process of analyzing, you've created an amazing knowledge asset. An asset, we should note, that's based entirely on what actually is vs. what some algorithm, created by some person, predicts might be.
So how will we take that "actually is" and turn it into forward progress that matches the prospects' journeys?
That takes the right mindset and the right MarTech. It gets built around a CDP.
If analysis helps you anticipate the buyer thoughts and needs at each stage in the buying journey (even down to preferred channel and time of day for communication) then orchestration is about making it happen across all your systems.
You'll need a central system which can execute tactics across all your different platforms. Sales acceleration cadences, paid ad custom audience campaigns, marketing email sends, phone calls, SMS, personalized web content, special segments for customized chat bot experiences, direct mail and more.
And of course each step will likely depend on the outcome of the previous step! So the linear execution of most marketing automation workflows can't manage and adapt to the variabilities of engagement - or absence of.
That's what a CDP can do on top of unification and analysis. And that's why orchestration is such an elusive goal.
Common Sense Sales > AI Predictions
Why would you guess - much less pay top dollar for opaque guesses - when you could know, anticipate and deliver with hyper-personalization at scale?
The honest answer is you wouldn't.
The real answer is that for most companies delivering hyper-personalization at scale has been impossible with typical data and MarTech.
Here's the bottom line.
Don't invest a ton of money in multi-year agreement for technology that's in it's infancy - especially when you can do better, now, using your own insights with the right data and platform combination.
