Forecasting What Matters
“CEOs always act on leading indicators of good news, but only act on lagging indicators of bad news.” — Andy Grove
Why do we bother forecasting the unpredictable? Having gone through the process hundreds of times over with our founders, I’ve found that forecasts deepen our connection with reality and help us manage optimism rationally.
In this blog, we will look into how to operationalize forecasts and make them as actionable as possible. We will begin with high level principles and then transition into the more tactical aspects of the forecasting process.
Forecasting principles
1. Keep it simple: commit to growth and burn targets
It is incredibly helpful to narrow down the set of targets you commit to — going from 10 to 5 to 2 will help you focus on what matters. When you forecast, make your growth and burn targets explicit in your summary output. Then below those two targets show a high level snapshot of what is going on using a limited set of metrics that you find helpful. We’ll return to this idea in a few templates below.
2. Keep it within reach: 1 year time horizon or less
The incremental value of extending forecast horizon beyond 1 year decreases quickly. I believe that even growth stage startups (series B to D) should focus on a 1 year forecast, and look beyond that timeframe primarily through sensitivity analysis — ie by stress testing top down assumptions. Eg. how much cash do I need to add $XM of revenues if it costs me twice as much as expected?
Veeva’s Peter Gassner captured this idea well. In its first year, management at Veeva purposely kept forecast horizon to one quarter — what’s the point of going beyond that early-on when you don’t know if you have product-market fit? They extended forecast horizon progressively as they matured and are now looking at 3-year plans post IPO.
3. Complexity and long forecast horizon don’t go well together
As you pull multiple variables into the future, the combined probability of all variables turning out to be true converges to zero. The point here is that you want to forecast things that matter for your business and that are within a line of sight.
4. It is better to forecast reality than pipe dreams
This one is from my colleague Dan Freedman and it never fails to make me smile. His point is that forecasts should predict reality. The intuition here is that it is far more useful to look at observed behaviours than to use formulas to forecast. When building your assumptions, try to think beyond the numbers and understand how you can make sense of them based on what you’ve seen (either with user engagement, acquisition funnels, retention, employees, etc.).
For instance, it is way more useful to look at empirically observed year-1 and year-3 contribution margin LTVs to understand CAC payback than it is to use formulaic derivations. Note that using observed patterns is also extremely useful when looking inside the company. For example, one can use observed sales rep productivity (eg. measured via ramp times and quota attainment) or the velocity at which your customer success organization can respond to customers (eg. measured as the portion of total calls answered within 2 minutes) to inform assumptions regarding these things.
5. Provide context with historical performance data
Always include previous historical performance data in your forecast to put things in perspective and comment on key trends (eg. growth, capital efficiency, S&M efficiency, gross margin, etc.). For instance, if you are preparing your 2019 forecast, try to include 2016, 2017 and 2018 actuals (or actuals since inception) *in the same format* as the forecast summary.
6. Focus on the targets you need to hit in order to raise your next round
Time in startup land doesn’t continuously flow like a river. Rather, it is divided by a number of checkpoints that need to be crossed with momentum to secure the next round of financing. This is where you can get some of the most valuable feedback from your investors — ask them what your targets should be to raise a next round.
Startups can be on one of two trajectories: (i) VC-backable, which requires high growth; (ii) not VC-backable, which requires either a path to breakeven or one to make the company VC-backable again. Each of these two states has a set of targets dictating what the forecast should look like at the end of its horizon.
7. Forecasts are best managed dynamically
The irony of forecasting is that the second you finish it, it will diverge from reality (either in a good or bad way). Beyond using empirically observed data to inform assumptions, the strength of your forecast will come from your ability to dynamically measure variances and take appropriate actions.
8. Use charts to put your scenario analysis in perspective
When you are done building your base case and have clear growth and burn targets, it is useful to model 2 other scenarios: (i) 2X your base case growth rate (or at a minimum a growth rate that is top quartile for your industry): this will force you to identify the bottlenecks in your model and where things break, (ii) breakeven plan: creates clarity around the actions needed to get to breakeven within your runway (note: simply going through the exercise (ii) will reduce the chances of you having to implement (ii)).
When done, put your 3 scenarios in a single chart and show what growth rate and cash balance look like as two separate lines under each scenario (so that you get to 6 lines, 2 for each of the 3 scenarios).
9. Benchmarking is useful but understand its limits
There are useful generic benchmarking resources for SaaS and marketplaces. However, it can be easy to “hide” behind benchmarks and fail to capture the uniqueness of a given business model or industry. This is especially true for marketplaces given how widely business models differ within this category.
For example, if you run an enterprise SaaS business and are trying to understand what “good” net retention looks like, you could shoot for the standard 110%, but then ask yourself why that is? Why would your business compare to one in a totally different industry? A great approach to this question would be to first look at your cohorts, and then calibrate with industry-specific benchmarks (or at a minimum relevant public comps).
10. Foresight, foresight & foresight
The importance of foresight cannot be overstated. As you develop reflexes for tracking variances to forecast (start with quarterly pre Series B, then move to monthly and weekly), a powerful additional step is to comment on the implications of these variances for the next quarter and rest of year. A great way to do that is via a waterfall chart showing the original forecast for the current quarter and rest of year.
Forecast output summary
A few notes on the templates we present below:
- Forecast output summaries are not meant to be used as your KPI dashboard (like this one or this one), which sheds light on underlying business health indicators. They are meant to help management and the board understand clearly what growth rate and burn you are shooting for.
- Define any metrics that you find useful to present below growth and burn formally in a tab somewhere in the forecast.
- It should be very easy to audit the assumptions supporting your forecast. Use footnotes and commentary when needed. If you feel that only you can understand how a number works, its probably a good idea to write down its definition.
1. SaaS forecast output summary example
2. Marketplace forecast output summary example
3. Consumer hardware forecast summary example
Link to Google Sheets templates.
Operationalizing your forecast
Once you’re satisfied with your base case targets, inject your forecast into your reporting machine, including monthly financial statements, KPI dashboards as well as your board reporting. Track variances, comment on implications of these variances for the next quarter and rest of year and adapt dynamically.
