Is the forecast wrong again?
Extra cash on hand since you can’t trust it? – That’s expensive.
Didn’t hire enough resources? Now you’re late! – What a waste of sales effort.
Spending more time firefighting than managing? – That’s frustrating.
Forecasting can be very complex, so we wrote a tool that eats the complexity before breakfast
This is the kind of magic that we keep hearing should be possible with AI, but rarely see in reality.
Our forecasting AI assistant updates your monthly rolling forecast in minutes. In those minutes, it also considered 40 or more possible predictor updates for all of the accounts in your P&L and Balance Sheet. This self-learning system is always updating your models to be the best possible every time it gets new data.
That sounds like a lot of work to set up
It is… but we made that part easy too.
It takes about 10 minutes to get a 36-month cash flow forecast for a brand-new company
And it’s NOT just an Excel template, it’s actually going to have 200-300 account models that are tuned from your business data.
See, our AI uses “recipes” of prompts that are based on GAAP and IFRS accounting standards. Any business that works with accountants will generally follow those standards. We also have industry-specific recipes that improve the starting point. We can make an initial forecast in minutes, and then we can see which models, if any, need additional predictors.
Every business is unique, and we know that from years of modeling and forecasting experience. We start with the best available recipe, and then we customize a recipe just for your business. That combination of ingredients is tuned to your tastes.
Are you using a 47-tab Excel sheet to forecast your business?
I’ve done that for many years. I used to call it THE BEAST. It could answer an impressive number of questions and “what if” scenarios. We explored the limits of available calculation power and statistical methods. I had to upgrade to a gaming computer with a dedicated GPU to make Excel usable.
I got sick of it.
And there was nothing I could buy that solved this problem.
So we did something about it.
We moved all of that inferential statistics and complexity into software dedicated to “eating the complexity” of forecasting. The AI/ML kernel does the inferential statistics for you. You’ve just got to feed it your business knowledge. (Nervous about AI? This isn’t that tech… read more below).
We build a “Digital Twin” of your business financials.
We build a digital twin of your business financials. That means we have a statistical model of every account in your chart of accounts defined relative to sales funnels, market factors, leading indicators, or over 47 time-series indicators. This allows highly accurate and lightning-fast forecasting.
- Want to know what your P&L and Balance Sheet might look like in six months? – done
- Want to know what cash looks like and what months will be tight on cash? – done
- Want to know where profitability is going off the rails in time to fix it? – done
- Want to know how many people you need to hire in six months? – done
The traditional forecasting process requires an FP&A analyst (Financial Planning & Analysis), director, or CFO to spend days or weeks entering data and updating your financial forecasting models before you can start answering what-if questions or analyzing KPIs. This forces you to trade off between capability, speed, or frequency of updates. You don’t need to do that anymore!
This took what used to be nearly a full-time job for me and made it possible to do it in a few days a month. A 90% reduction in work, conservatively. That means I can actually spend the time focusing on value-added business partnering instead of fighting data in Excel.
We’ve taken the hard work of incorporating new data every month or more and made it an automated process that takes seconds instead of days or weeks. The purpose-built Artificial Intelligence and Machine Learning kernels make it possible to do the heavy thinking lift asynchronously. Your FP&A teams can focus on adding details to the model without having to worry about the time it takes to re-run and rebuild the model every time you update the forecast.
And it’s FAST… like really fast:
At the end of those 37.69 seconds, you get 339 refreshed models that incorporate the latest monthly data, have automatically re-tuned each model to accommodate new trend signals, and are production-ready to show the updated impact of future revenue, profits, cash, or whatever KPIs you want.
We’re Launching our BETA on April 5th, 2024! If you’re interested in being a Beta client, please send a note to Nate@BetterForecasting.com
Concerns about “AI”?
Worried about AI and its information security, its retention of your trade secrets, or whether it’s going to become SkyNet or Hal? This is NOT an LLM (Large Language Model). LLMs aren’t the right tool for forecasting. They’re frankly terrible at it… we’ve checked. This is a purpose-built inferential statistics machine that understands financial statement hierarchy and understands the cause-and-effect relationship inherent in business and how it’s displayed in financial statements.
Excel is NOT Dead!
Oh, and we’ve got lots of ways to keep working with Excel. I LOVE Excel. It’s the O.G. No-code/low-code platform. I’m not trying to proclaim that “Excel is Dead”, because it’ll never be, and I hope it lives a long happy life. This is built to be EASY to get forecasts back into Excel again so you can do the analysis that crazy VP asked you to do, but it should take significantly less time!
Not another modeling platform!
We use our AI to select which drivers are the most significant. Models and their factor values don’t have to be explicit… in fact, it’s better if they’re NOT explicit most of the time. For this workflow, all you have to do is suggest a group of input drivers, and the model automatically assesses which ones are the best predictors every single time we add new data.
Let’s talk about how that changes the modeling workflow:
Traditional Modeling:
Income = 0.5*AccountB + 7*AccountF – 3*AccountG
You have to explicitly set what those parameter relationships are (the “0.5”, “7”, and “3” above). This also means you need to do all of the updating of those manually as new data comes in. Which is where most of the time is wasted in the traditional FP&A workflow. Most companies we onboard are just not reviewing and updating these parameters on a regular basis! This causes degradation of quality over time.
I watched demos of the “best in class” modeling platforms recently. And in every single one of them the first step is to go re-enter all of your modeling calculations. That’s an incredible amount of work, and it’ll continue to be a lot of work to maintain. I couldn’t believe how much more work that is than what we do.
Inferential Modeling:
Look for predictors of Income in the following groups:
- CRM Accounts B, C, or D
- P&L Accounts F, G, H, or Z
- All Sub-Accounts of Balance Sheet Account Y
- 47 timeseries predictors
You DON’T explicitly set any parameters, the model automatically updates those parameter relationships based on all of the available data… and you should be getting updated data at least monthly. ALL of the maintenance is done for you. The models are always improving in accuracy every month as we gather more and more predicted vs actual history, rather than the degradation of static manual parameter updates.
We have a set of graphical reports that is designed for users to quickly review for issues. Our human brain’s superpower is pattern recognition. We designed these reports to take advantage of that.
For the occasions when parameters should be explicit and fixed. That’s OK, we have that capability too.