Forecasting with ChatGPT 

Forecasting with ChatGPT 

Insight from Emily Kramer

Take it from someone who has built (and rebuilt… and rebuilt) a lot of business forecasts:

It sucks.

Maybe there are some reading this who LOVE building forecasts. More power to you.

I can’t stand it. It’s incredibly time-consuming, every stakeholder has a thousand adjustments they want to make, and most of the time (even with the best data) it feels like guess-work.

So, I’ve learned to automate the tedious parts of forecast building and save myself a ton of time.

For an example of the workflow, let’s build a sales forecast using ChatGPT.

1. Gather and analyze historical data: get any relevant historical sales data together in a single table. Submit that table to ChatGPT, and ask it to highlight any trends that it notices in growth.

2. Build your initial forecast: Ask ChatGPT to create a new table that projects sales and revenue growth for the next year. Specify that it should base these projections on the trends it found in the initial analysis.

3. Adjust and fine-tune: Here’s where the workflow really shines. Bring in assumptions that you think ChatGPT should factor in. Expecting a seasonal increase in revenue in Q3? Boom, the table is updated with that logic. You can build and update the forecast conversationally, and fine-tune with more specific instructions.

Once you have an initial model that you’re happy with, ask ChatGPT to create a CSV for you to export and a list of the assumptions that it factored in.

And there you go. An exercise that takes hours is completed in 17 minutes (no joke, we timed it ⏱️).

Learn more: Read the full description of the process