Skip to Content

Forecasting & Modeling Your Customer/Client Behavior

March 12, 2023

Dedicated to creating the blueprint + roadmap to build what you need – and to alert you when something materially deviates from the expected? The thing that enables that result is a high-quality forecast? You’ve found the right article.

But before we get into the forecasting piece, let me take a minute to illustrate why it’s critical, using an experience I’m sure most people reading this have been involved in: 

The “Great Success” Meeting

We’ve all been in this meeting – in one capacity or another. It usually takes place after a big campaign or significant initiative (or maybe just before your agency/partner/client’s annual renewal is due). And while the details of The Meeting may elude you (probably for the best), I’m sure you remember the tl;dr:

The [thing] was a great success! The evidence for this? A prodigious number of graphs where stuff that we should care about went up-and-to-the-right or “beat the benchmark.” 

After The Meeting™, there’s an (often) unasked question: if everything we’re doing is so great, then why isn’t our organization doing great? 

Over the years, I’ve found the following four things to be true: 

  • Absent a clear + attainable success criteria, any initiative will be deemed a success. 
  • Understanding the import of various levers / micro-level factors on macro outcomes is limited at best
  • That limited understanding results in poor prioritization, which leads to more iterations of The Meeting™ 
  • Tight accountability loops are the foundation on which remarkable things are built

To use the “blueprint” metaphor above: most organizations are fantastic at big-picture strategy – they know what they want to build. There’s a bold, audacious vision. And the teams tasked with bringing that vision to life tend to be tactically solid. The gap is the connection between these two strengths – the accountability check that ensures the tactical execution remains in service of the bold, audacious vision. This is the proper role of a forecast.  

The unfortunate reality: most organizations struggle to create performance + financial forecasts. 

And that is what leads to the “great success” meetings, to questions about what’s “working” (and what’s not), to stalled growth and ultimately, to poor macro-level outcomes. In fact, I’d go as far as to say that forecasting is one of the most critical skills needed to thrive as a growth marketer, C-suite leader or entrepreneur. That may seem odd – after all, aren’t there more important capabilities? Shouldn’t strategy, creative, product, execution, iteration/evolution all trump forecasting? 


For one simple reason: accountability is necessary for sustainable growth. A properly-constructed forecast is a prime source of accountability: it forces us to contrast what we believed to be vs. what is – and, if done properly – it enables us to identify and address the factors causing the divergence. 

This is not to say that a forecast is a substitute for execution. It isn’t. Rather, forecasts are a blueprint – a source of truth against which all execution (the actual building) can be evaluated against. 

Done well, this should permeate the entire organization – informing your staffing, inventory, offer construction, financials/fundraising and more. 

Unfortunately, most forecasts do not live up to this standard – and so, we throw the forecasting baby out with the proverbial bath water. The simple truth is that we need better forecasts. I have a few ideas for exactly how we can – and should – do that. 

The Unlocks

The single-most common pieces of feedback I receive at the outset of conversations about forecasting (usually with new clients) are: why do I need it? What’s the value-add? How does spending time on forecasting help me build a better business? 

While there’s no single, one-size-fits-all answer to that question, let me illustrate a few of the “unlocks” that it has enabled: 

  • Parse Signal vs. Noise – one of the fatal flaws in most growth (not just marketing) efforts is an inability to distinguish between signal (something of material import) and noise (random fluctuations + luck). The difficult part is teasing out when something has cleared the threshold for luck and is now something of material importance (and thus, needs to be addressed). The result? Most people spend way too much time + effort thinking about (and reacting to) noise – and what gets lost is the signal. We’ve all had those “need new ads up NOW – no leads today!” or “OMG! Since launching the new creative yesterday, sales are up 50%! Roll it out everywhere!” – and most of the time, this is just luck. But instead of having a source to push back and keep the focus on what really matters (finding better offers and angles, refining audiences, building more robust funnels, etc.), we find ourselves pulled away to chase noise. 
  • Course Correction vs. Wrong Destination Almost every initiative has periods when it does not perform as expected – the challenge is identifying the discrepancy + resolving it ASAP. I call this “course correction.” Unfortunately, what typically happens is the opposite of this: no one realizes the issue until the budget is spent, the campaign has run, and the final results don’t align with the expected results. This is the “wrong destination” problem – and it is absolutely avoidable IF you have an “early warning system” (in the form of your forecast). 
  • Diagnostics – Once you’ve identified the need for course correction, the next question is usually, “How do we get back on track? What’s the root cause + how do we fix it?” Answering that question is rarely easy or obvious – but a forecast can help you by pinpointing which atomic elements (more on those later) are not performing at their expected level. That, in turn, leads to productive conversations like, “We’ve noted that our actual conversion rate is 27% lower than the forecasted range site-wide, so we’ve done the following research to identify a cause AND we propose the following action steps to resolve it.” That’s the kind of productive conversation that (i) earns the trust of senior leadership AND (ii) avoids wrong destinations! 
  • Prioritization – if you could improve your site-wide conversion rate by 5% or your ad CTR by 45% (while maintaining traffic quality), which one would you choose? Is increasing lead qualification rate by 10% worth a 4% decline in conversion rate? Not sure? A properly built forecast can help you answer those (and many other) questions – and in so doing, inform your prioritization. After all, wouldn’t you rather spend more time on the things that make the biggest impact? 
  • Inventory & Operations – there is more to growth than marketing – and a proper forecast can help you understand what inventory + staffing levels you should maintain at a given point in time (so you’re not sold out or unable to take on new work AND not overstocked // overstaffed, and thus forced to discount in order to cover your costs). 
  • Accountability – Each of the previous 5 “unlocks” flow into this one: accountability. Ultimately, growth (including marketing) is an investment, and like all investments, it should be expected to produce a return. Your forecast is that accountability check that ensures you’re on track to produce an acceptable return. To that end, I highly recommend including actual vs. forecast in all your reporting moving forward. We integrate it directly into our dashboards for this reason – it keeps everyone accountable! 

I hope that convinces you that forecasting is an interesting + worthwhile exercise; and if it does, I’m sure the next question is: “How do I get started?”

The Ingredients Of A Remarkable Forecast

Atomic Data: 

The #1 issue I see with most forecasts is that they are “backed-in” from a desired result. An often-arbitrary target is set, and the inputs to the forecast are calibrated to attain that result – with little-to-no regard for the validity of those input/assumptions. The end result is a model that appeases the executives in the short-run, but ends up placing unrealistic expectations on the team tasked with executing against the targets set. When reality doesn’t align with the forecast, excuses are made, accountability is conspicuously absent and we do the entire exercise over again next year (and we throw a “great success” meeting in there for good measure). It’s a tale as old as time – and one we’re going to break. 

Instead: build your forecast from the bottom-up. This means starting with the foundational elements + processes that contribute to the ultimate outcomes. For most businesses, these atomic elements are (1) cohorts; (2) line items and (3) transition metrics. 

Let’s outline some quick definitions: 

  • Cohorts: a group of people (or customers) with a shared characteristic, usually age. These represent the “atomic unit” from a consumer standpoint – primarily because cohorts tend to be more stable – which is to say, their behavior tends to be relatively consistent over time. 
  • Line Items: the constituent parts of your P&L and Cash Flow Statement. These typically include a mix of variable costs (like your cost of goods sold, processing fees and shipping fees), along with fixed costs (like rent, salaries, etc.) 
  • Transition Metrics: these are the “converters” that indicate the rate at which one thing in your forecast transitions into another – for instance, how often an impression transitions to a site visit, or a site visit to a customer, or an active customer to a churned customer. 

A forecast built from atomic elements is likely to surprise you – and that’s a good thing. It is exceedingly difficult to grasp the full implications and relative value of initiatives/updates on your macro goals. 

Start with a Basic Model: 

The easiest way to get started is to begin with a basic model template (I love the  D2C  &  SaaS  models built by Lightspeed Ventures, then customizing it for your organization. While these models are primarily aimed at D2C & B2C, the underlying principles readily translate to the B2B sector (including SaaS) remarkably well. The templates are in excel, and you can play around with them to find a solution that meets your needs. 

You don’t need to be perfect on the first (or any future) go-round. Forecasting is not an exact science; think of it as an approximation that derives its value not from correctness, but from its ability to determine when something is wrong – giving you the opportunity to address it before it snowballs. 

Clarity On The Numbers + The Business: 

Fundamentally, forecasting is an exercise in understanding organizational operations, and using that understanding to project forward. That starts with understanding your business’ numbers, including:

  • Gross Revenues
  • Discounts + Costs of Revenue
  • Returns Allowance
  • Costs of Goods Sold
  • Costs of Delivery
  • Operating Expenses (salaries, rent, utilities, etc.) 
  • Costs of Acquisition
  • Debt Service (if applicable)
  • Bad Debt Allowance (if you use accrual accounting)  
  • Target Profit Margins

If you don’t understand your business’s (or your client’s) numbers inside and out, you’re going to have a devil of a time not just with forecasting, but with growth more generally. The simple reality is that most SMBs don’t fully understand how money moves through their organization, nor where it ends up – and the result is an overreliance on marketing to fix a fundamental business problem. 

Building a forecast requires you to get granular and put all of this on paper – and sometimes the results are scary. Sometimes they’re flat-out terrifying. But I firmly believe knowledge is power, and the first step to fixing a problem is clearly acknowledging that one exists. 

Parse Out Different Types of Customers

One of the inherent flaws in most models – for both B2B + B2C – is a tendency to treat different audience classes similarly – which leads to weird outcomes. In reality, you have 3-4 audience types, each with different needs, levels of predictability, variability and import to your macro-level success. 

A forecast must reflect those nuances. 

The simple reality is that you should be modeling your customer/client behavior differently than your potential customer behavior, which in turn is different from your prospective audience behavior. Layering these together provides a much more accurate, stable and useful forecast: 

To break this chart down: 

  • Existing Customers – for almost every business, the most stable + predictable source of revenue is the existing customer base. Using your existing customer data and your historical performance data (by month or quarter), you can predict a significant component of your future revenues – which is why your existing customer file should always form the foundation of your forecast pyramid.  
  • Owned Audiences – the next level of your pyramid is your “owned audience” – the people who you know and can communicate with directly (i.e. without paying or relying on a 3rd party, like Meta or Twitter). This includes your email, phone + SMS lists (minus your current customers), along with your lapsed customer file. The key ingredient here is that you have direct control over your ability to contact these people – which makes them more valuable AND a more stable revenue stream than anything that requires a third party’s cooperation. You can forecast these on a “revenue per send” or “revenue per call” basis (among other approaches). 
  • Rented Audiences – up one level are non-paid sources of traffic that you do NOT have direct control over – this could be organic (since you can’t control if Google ranks you), social (again, you only have control on if you post something; who that post is shown to is largely beyond your control), and referral/affiliate. Each of these exhibit different levels of variability, but each is less stable than your owned audiences. 
  • Paid Acquisition – finally, there’s your paid acquisition – which is the most volatile of the group. Unlike each of the previous three levels, there are substantial, direct costs associated with advertising AND you are subject to the volatility of ad markets, along with questions like new customer conversion rate, AOV, lead qualification rate, etc. The best way forward for these is to forecast your acquisition costs using your budget, estimated CPM and your historical action rates (CTR + CVR): 

CAC = CPM / (1000 * CTR *CVR)

Allow For Uncertainty

Most forecasts are of the “point” (or “categorical”) variety. That’s a fancy way of saying that they state “X” as the output – 15 leads, 20 net new customers, $2,000 in daily revenue – whatever. Unfortunately, the world we live in is rife with uncertainty + full of externalities just waiting to blow up your neat and tidy forecast. Compounding this is the reality that data tends to be noisy in the short-run + predictive 

These can range from the mundane (perhaps a local sports team has unexpectedly advanced to the Big Game, and no-one cares much for your in-store sale; or perhaps everyone is throwing parties for said game, and now your catering business is busier than you can handle) to the catastrophic (COVID lockdowns). The important thing is that you have a mechanism for distinguishing between normal variability / luck and material deviation. 

The solution to this is to take a probabilistic approach to your forecast – so instead of a specific target ($2,000 in daily revenue), there’s a range of “acceptable” outcomes – $1,525 – $2,597. 

Anything in that acceptable range over the short-run? Acceptable. This is critical for stomping out meddling – a staggeringly counter-productive passtime for way too many marketers + executives. Not only are you changing something that (for all intents and purposes) is working, you’re also wasting time that could have otherwise been used for more critical tasks. 

Connect the micro to the macro: 

Even in smaller organizations, almost no-one sees the full picture. A well-designed forecast links each organization function together – allowing everyone to see a 30,000’ view, as well as zoom in to assess the drivers of that overall performance at a micro level. 

Fundamentally, this comes from understanding that big picture outcomes (profitability, net customer growth, etc.) are a result of a series of smaller processes – from how efficiently we acquire customers to how adept we are at retaining them to how lucrative those customers are in the interim. Each of those intermediate processes (acquisition, retention, costs, etc.) is itself a product of underlying functions – for instance, acquisition efficacy is contingent on the cost to reach people (CPM), how effectively you convert those impressions into traffic (CTR) and how efficiently you convert traffic into customers (CVR). 

Building a forecast out from the micro (the atomic units of each core function) to the macro (the business-outcome activities that we care about) enables everyone in the organization to better understand the impacts on the big picture. 

Put simply: a well-designed forecast can be a game-changer for your organization. The challenge, at least historically, is building a model that does this. Most (that I’ve seen) suffer from a number of problems, from structural deficiencies to arbitrary assumptions to back-fitting and poor maintenance.