Ask HN: Guides to Financial Forecasting?
8 comments
In my experience what the most helpful thing to investors and you will be bottoms up understanding of how the business looks with a somewhat reasonable set of assumptions. So make a tab full of your assumptions, number of customers, cost per customer, number of engineers, churn rate.
When do you collect cash from customers? When do you pay bills? How fast can you cycle cash? (Meaning if you pay back marketing over 2 years vs 2 months) a LTV/CAC ratio of 1.5 is not great, but if you have a 1 month payback you basically don't consume cash so that's cool (these can be signs of dads though). But I would just look online for comps for the specific inputs into your model.
If you want a template you can just look at PnLs online.
All forecasts for startups are wildly wrong. But I think of forecasting as a way of valididating a strategy. Lots of people dislike the "if we get 1% of the market we will.be huge" approach, but I like at least validating if that is the case (it is not always the case).
Forecasting can also help you understand pricing. If your market size is about 10,000 buyers what do you need to charge to have a business.
When do you collect cash from customers? When do you pay bills? How fast can you cycle cash? (Meaning if you pay back marketing over 2 years vs 2 months) a LTV/CAC ratio of 1.5 is not great, but if you have a 1 month payback you basically don't consume cash so that's cool (these can be signs of dads though). But I would just look online for comps for the specific inputs into your model.
If you want a template you can just look at PnLs online.
All forecasts for startups are wildly wrong. But I think of forecasting as a way of valididating a strategy. Lots of people dislike the "if we get 1% of the market we will.be huge" approach, but I like at least validating if that is the case (it is not always the case).
Forecasting can also help you understand pricing. If your market size is about 10,000 buyers what do you need to charge to have a business.
> When do you collect cash from customers? When do you pay bills? How fast can you cycle cash? (Meaning if you pay back marketing over 2 years vs 2 months) a LTV/CAC ratio of 1.5 is not great,
i recall seeing a talk from some saas company that was really driving home the importance of how quickly the business can generate cash to pay for customer acquisition costs. i think they talked about offering plans to sign up for 12 months in advance (with a fair discount) versus monthly, with the idea that if enough customers sign up to the 12 month plan at a 20% - 30% discount then if you still get enough cash in the short run to pay for the customer acquisition costs, and enough cash in the long run to operate at a profit, you can start to pour excess cash into customer acquisition
(apologies, dont remember the exact talk)
i recall seeing a talk from some saas company that was really driving home the importance of how quickly the business can generate cash to pay for customer acquisition costs. i think they talked about offering plans to sign up for 12 months in advance (with a fair discount) versus monthly, with the idea that if enough customers sign up to the 12 month plan at a 20% - 30% discount then if you still get enough cash in the short run to pay for the customer acquisition costs, and enough cash in the long run to operate at a profit, you can start to pour excess cash into customer acquisition
(apologies, dont remember the exact talk)
Thanks!
This is really helpful.
This is really helpful.
Are you seeking help in constructing the mechanics of a model or in taking an existing model and projecting it into the future? Or both?
Both tbh.
I probably should have mentioned we are B2B in an industry with very long sales cycles.
I probably should have mentioned we are B2B in an industry with very long sales cycles.
If B2B, one way to do it is estimate how many accounts, and what size, you may get on a yearly basis.
After you create your model, you can cross check it against similar company s-1 filing (initial ipo filing). They will usually include their P&L few years prior to the ipo.
You can reference their spent against yours to see if the model is realistic.
Feel free to reach out if you have more question.
After you create your model, you can cross check it against similar company s-1 filing (initial ipo filing). They will usually include their P&L few years prior to the ipo.
You can reference their spent against yours to see if the model is realistic.
Feel free to reach out if you have more question.
my first advice is to Approach this as if you are running experiments... you don’t know the outcome beforehand and you’re likely never to know it. A forecast is not a crystal ball and it’s very important both for your survival as well as the company’s never to treat it like one. What you’re doing is creating a set of future benchmarks and then trying to understand actual performance to those bechmarks. So the first step is a mindset. The second step is to build your process so that the forecast gets revised every month. You want a handle on revenue and cash burn.
In terms of creating the prediction, what I recommend is to have everyone involved name all the known potential deals... what will it take to close, what is the timing, what is the revenue, what is their confidence. The idea is to mimic how hurricane forecasts are built... they take everyone’s best model and then show an ensemble of all the tracks a storm could take. That’s the range of outcomes. You need someone to play the pessimist if there isn’t already someone in the group. Once you have an ensemble — ie a range of possible outcomes - you can average them into a single line.. you can weight the inputs and then average, etc...
Finally you need to reconvene and discuss risks, features, Investments, all the moving parts that you’ll need to manage to make this forecast a reality.
Even when you have historical trends to leverage in a statistically generated forecast, you still need to coalesce into action. Will power is what makes predictions a reality... whether you are one month into it or have several years behind you.
As for the mechanics, I recommend a book called the ten day mba by Steven Sillbigier. Don’t attempt to read it cover to cover. keep it as your bible... if you want to go deep into theory, get the McKinsey book on valuation.
Edit: one more thing: never ever accept an assumption because it sounds reasonable. For example, your sales guy says I can close the first two customers in six months, an then the next six months I’ll close four and then the next year I’ll do 12. A lot of people will say, that sounds reasonable, the first ones are always the hardest. I can tell you right now that forecast will fail. Same thing with looking at peer data... if you only look at velocity of the revenue line and don’t accurately assess the factors that contributed to that growth, it’s a guaranteed fail. Never use one estimate, never accept in input because it seems reasonable and never look at a trend without a regression next to it. Many people eyeball trends and it’s amazing how many can see growth in a flat or declining trend. Don’t fall for it!
(Ok that was more than one last thing...
In terms of creating the prediction, what I recommend is to have everyone involved name all the known potential deals... what will it take to close, what is the timing, what is the revenue, what is their confidence. The idea is to mimic how hurricane forecasts are built... they take everyone’s best model and then show an ensemble of all the tracks a storm could take. That’s the range of outcomes. You need someone to play the pessimist if there isn’t already someone in the group. Once you have an ensemble — ie a range of possible outcomes - you can average them into a single line.. you can weight the inputs and then average, etc...
Finally you need to reconvene and discuss risks, features, Investments, all the moving parts that you’ll need to manage to make this forecast a reality.
Even when you have historical trends to leverage in a statistically generated forecast, you still need to coalesce into action. Will power is what makes predictions a reality... whether you are one month into it or have several years behind you.
As for the mechanics, I recommend a book called the ten day mba by Steven Sillbigier. Don’t attempt to read it cover to cover. keep it as your bible... if you want to go deep into theory, get the McKinsey book on valuation.
Edit: one more thing: never ever accept an assumption because it sounds reasonable. For example, your sales guy says I can close the first two customers in six months, an then the next six months I’ll close four and then the next year I’ll do 12. A lot of people will say, that sounds reasonable, the first ones are always the hardest. I can tell you right now that forecast will fail. Same thing with looking at peer data... if you only look at velocity of the revenue line and don’t accurately assess the factors that contributed to that growth, it’s a guaranteed fail. Never use one estimate, never accept in input because it seems reasonable and never look at a trend without a regression next to it. Many people eyeball trends and it’s amazing how many can see growth in a flat or declining trend. Don’t fall for it!
(Ok that was more than one last thing...
Are there any good guides to building a financial forecast out there?