%20(2).svg)
.svg){kind=icon}
.svg)
.svg)
As 2022 draws to a close, Finance teams everywhere are focused on forecasting, budgeting, and tracking revenue targets for 2023. But this is always an incredibly hard exercise, and the cost of being wrong can be terrible.
Clearly, this is worth getting right. So, how do the best teams do forecasting? They make sure they are balancing forecasting based on leads, sales capacity, and sales efficiency to get the right numbers.
From talking with experts like you, we've gathered five main approaches to producing a great new revenue forecast (note: as a subscription company with recurring revenue, you know you also need to forecast existing revenue, but we'll talk about that another time).
🖐️ 5 approaches to forecasting new revenue:
- 📝 Account-by-account
- 📕 Pipeline forecast
- 🧑🏿🤝🧑🏼 Sales capacity
- ✖️ Modeling your funnel (un-cohorted)
- 📐 The holy grail: Cohort-based approach
📝 Approach 1: Account-by-account
Who uses this approach?: When a business is in an early stage (usually only one person is selling, and there are only a handful of viable opportunities), or when a business serves very few but very large enterprises, this is the best way to forecast.
What do you need?: A pipeline and a gut feel ☝🏽 .
How does it work?: In this method of forecasting, you simply look at every possible deal that could close in the future, use conversations and notes to guesstimate when they might close, and what dollar amount it might close for. Using those dates and amounts, you can aggregate the amount of revenue you forecast closing in any given month.
📕 Approach 2: Pipeline forecast
Who uses this approach?: This is an expanded form of the previous methodology, but for bigger companies with a multi-member sales team.
What do you need?: Historical data on close rates by stage + a pipeline with consistent and accurate-ish amounts.
How does it work?: You assign a probability of close for each stage (i.e., 10% for deals in discovery stage, 50% for negotiation stage, 90% for verbal yes stage) and then add together a weighted sum of each opportunity multiplied by the probability of close for the stage of that opportunity.
Watch out: If your business has outlier opportunities that are much bigger (say your sales team mostly sells $10,000 deals but there are a handful of $500,000 deals in the pipeline 🙌🏽 ), you might only use this methodology for the more "normal" deals and pull the exceptional ones out entirely for a conservative analysis (or use the previous strategy to estimate those line by line).
👆🏿 It's important to note that Approach #1 and Approach #2 above have a major constraint: They only help you estimate what's in your pipeline at the current moment. To produce a forecast for more than a few months out (like FY 2022), you'll need to estimate revenue from deals that aren't even in the pipeline yet.
👇🏿 Approaches #3, 4, and 5 below help you handle that!
🧑🏿🤝🧑🏼 Approach 3: Sales capacity
Who uses this approach?: Companies with 10+ salespeople
What do you need?: Historical data on how much revenue each sales rep can close, a hypothesis on how sales efficiency will increase/decrease, data on how sales performance looks during ramp time.
How does it work?: Selling requires "talking" to customers (whether it's an in-person meeting, a call, or an email) - something that definitionally takes time, and therefore is subject to the 40ish work hours per week we all have. So, one way to figure out how much new revenue you can close is to figure out roughly how much new revenue ramped up salespeople generate per quarter, and then overlay a headcount planning model to figure out the amount of revenue you might make.
Your headcount planning model needs to account for the ramp time - a brand new salesperson won't make as much as your average ramped salesperson in month 1, so it's critical to figure out a) how much can they produce while ramping up and b) how long it takes them to become fully ramped up.
Then, quarter-by-quarter (or month-by-month) you can figure out the amount of new revenue you can expect to have based on the number of sales reps you have available to sell.
From there, you'll also want to adjust for anticipated sales efficiency changes (examples: maybe your lead mix is changing and things will get harder, or you're doing more brand advertising and things will get easier, or you've adopted a new sales methodology and anticipate seeing some gains). You'll want to model those changes.
Watch out: The biggest drawback of this model is that it does not consider either existing opportunities you have in your sales pipeline (maybe you have tons of great ones and you will do better than expected in the coming two quarters, or maybe the other way around), nor does it consider the number of leads those salespeople need in order to be maximally productive. That can be totally fine if your salespeople get their own leads, but if they rely on marketing support for leadgen, then this model is helpful but far from sufficient.
✖️ Approach 4: Modeling your funnel (uncohorted; see cohorted approach below)
Who uses this approach?: Companies with enough volume to be able to calculate funnel stats, but not enough to do a cohorted approach.
What do you need?: Conversion rates from each step of the funnel to each subsequent step of the funnel, an estimate of how many leads by month you'll have, anticipated average deal sizes.
How does it work?: To deal with the fact that other approaches only capture current pipeline but not future pipeline, you can consider your whole lead-to-close funnel. Your funnel will be specific to your business, but as an example:
.png)
Watch out: Getting clarity on your funnel can be a difficult data problem and getting your data infrastructure in order to figure out the key details is important. The biggest downside, however, is that this approach misses the fact that each of the steps above has a delay associated with it. Leads don't become MQLs the same month, and MQLs don't become opportunities the same month, and Opportunities definitely don't close the same month that they are created. Implicit in this approach is the assumption that everything moves down the funnel instantly, which (sadly 😿) is not how the world works.
To illustrate this point, say that in December 2021 you are worried that you won't hit your number for this fiscal year which ends January 2022. So, you decide that you're going to increase marketing spending rapidly to generate more leads, and because of the funnel you modeled above, you should close enough revenue to make your number.
That won't work! If you successfully generate the leads in December, the odds of those same leads becoming MQLs, Opportunities, and Won deals all within a month won't work unless you have a very short sales cycle.
📐 Approach 5: Cohort-based approach
Note: Many B2B SaaS leaders consider the cohort-based approach the holy grail, but we know it can be an unfamiliar concept for some! Rest assured we'll be writing a separate focus piece on this in the future; stay tuned.
Who uses this approach?: Companies with robust data infrastructure (like Subscript.com) that enables granular ways of cutting data.
What do you need?: Historical conversions of leads → opportunities by months after creation, historical win rates of opportunities by months after creation, an estimate of how many leads by month you'll have, anticipated average deal size.
How does it work?: This approach uses both your funnel, but also takes time-delays into account to produce a shockingly-accurate view of the far future. The cohort-based approach to forecasting solves for the timing issue. In the cohort-based approach, you first gather historical data on the leads you've generated and when they converted to opportunities (not an average, but a distribution - how many turn to opportunities month 1, month 2, etc), then historical data on opportunities and when they converted to deals (also a distribution). Sound hard? It totally is, but it's also totally worth it (Subscript.com makes it much easier 😎).
Once you separate things out by cohorts and model in the delays, you can get eerily accurate with your forecasts - you still need to consider what your cost-per-lead will be and what your average sale price will be, but you can very precisely get a sense of how much revenue you can hope to add each month of the next year.
Watch out: The shortcoming of this method is that it does not consider the number of salespeople you have or what's already in the pipeline, so we recommend combining it with the sales capacity method (approach #3 above) to make sure your headcount planning matches up, and with the pipeline method to double check the numbers for the months that are coming up.
👆🏿 Ultimately, the methods you choose for forecasting will depend on the stage of your business, the types of deals you are closing, and the makeup of your sales team. Regardless of which you choose, we always recommend forecasting utilizing multiple approaches to have the best chances of getting an accurate number, and planning the rest of the business accordingly.
😉 Coming next: How do you choose the combination of forecasting approaches that's right for you?
Spoiler alert: If your data infrastructure can support it, we recommend using cohort-based lead forecasting (approach #5) to make sure your lead planning is solid, and a headcount planning model (approach #3) to make sure you have the right team to hit your goal.
But that's a big, loaded topic so we'll reserve that for a future article. Make sure you subscribe to our newsletter so you don't miss it!
🗣️ We want to hear from you!
Forecasting is complex, and there are so many "right" ways to do it! Write us to:
- Share your two-cents: What forecasting approach has worked well for you?
- Add your expertise: Are there other forecasting methods you use at your company that we didn't mention above?
- Influence what we write about next: Is there a forecasting approach you'd like us to dig deeper into in future articles?
❤️🙏🏽 Thank you to: Laura Del Beccaro and Justin Megahan (Sora), Laura Forth and Dan White (Propeller), and John Dodderidge (Thankview) for reviewing early versions of this article
