One of the challenging things of using a machine learning system like Forecast Forge is learning how much you can trust the results. Obviously the system isn’t 100% infallible so it is important to have an idea of where it can go wrong and how badly it can go wrong when you are talking through a forecast with your boss or clients.
One ways that people familiarise themselves with Forecast Forge is to run a few backtests. A backtest is where you make a forecast for something that has already happened and then you can compare the forecast against the actual values.
For example you might make a forecast for the first six months of 2021 using data from 2020 and earlier. Then you can compare what the forecast said would happen against the real data.
Use data from this period To predict here ________________________________________~~~~~~~~~~~~~ |------------|------------|------------|------------|--------????|????????? 2016 2017 2018 2019 2020 Then use the same methodology here
Sometimes when you do this you will see the forecast do something dumb and you will think “stupid forecast. That obviously can’t be right because we launched the new range then” or something like that - the reason will probably be very specific to your industry or website. If you can convert your reasoning here into a regressor column then you can use this to make the forecast better.
On Tuesday I was kindly invited to discuss PPC Forecasting as part of the regular #PPCchat Twitter chat. There is a recap put up on the official website.
#PPCchat started just over 10 years ago and, whilst I don’t think I was involved in the very first one I joined the conversation as soon as it was moved to a more convenient time for the UK timezone. It was a very important part of my week for a number of years but I’ve drifted away from the community as my work moved away from hands on PPC management and more towards digital analytics and data science. I was very flattered to be invited to be a guest on the chat and very happy to be able to help a community that has helped me a lot.
The twitter chat format felt a bit frantic so I thought I’d expand on some of the ideas and questions raised here.
Before getting into anything too complicated here are what I think the two most important things you can do to get better at forecasting are. They have nothing to do with machine learning or fancy techniques or anything like that; you should be able to apply them regardless of your current process.
If you do only these two things then your forecasting will start to improve. And you don’t even need to pay for a Forecast Forge subscription to do them!
To help clarify things and avoid talking at cross-purposes (always a risk on Twitter) I split forecasting up into three overlapping areas:
Type one forecasts, where you forecast the impact of something new, are a very interesting challenge; the way to approach them from a Machine Learning angle is to collect data from other people who have done the new thing and then try to figure out which of the other people your client is most similar too. This is basically what people do too when they draw on their experience to make an estimate. Forecast Forge doesn’t store your data, know anything about the sector your business is in, or know exactly what metrics you are forecasting so this is not something it supports (there are a few ways you could “hack” this - ask me if you want to test them).
Forecasting doing more or less of something that you’ve already been doing is really important for paid media; this is how you estimate the impact of increasing or decreasing your budgets which is a super-important and frequently demanded forecast for everyone.
How often should you be updating forecasts or making new ones? The answer depends on what you mean by “forecast” and can range from “as often as possible” through to something much less frequent.
When you use machine learning to make a forecast there are three parts to it:
Making changes to any these could be called “forecasting”.
The three categories are a little bit fuzzy. For example it isn’t totally clear what the boundary is between parameters and model but the basic idea is that you can make very frequent updates for things near the bottom of the list and should be a bit more cautious with things at the top of the list.
This might be easier to understand with an example:
An easy way to make a forecast that includes weekly seasonality is to make the forecast for the next day a weighted sum of the previous seven days. By giving more weight to what happened seven days ago you will see a weekly pattern in the forecast.
Earlier this week I tweeted the below as part of another conversation
My hypothesis with @ForecastForge is that people can make better forecasts using a fairly simple algorithm + their own domain knowledge rather than getting someone else who can make a fancy algorithm but doesn't know the domain— Richard Fergie (@RichardFergie) March 2, 2021
I thought this would be a good opportunity to unpack my hypotheses about why Forecast Forge is a good/interesting idea. And also a good time to do a quick business update since it has been just over six months since launch.
One of the most requested feature since Forecast Forge launched has been the ability to forecast from monthly data and not just daily data.
Daily data is really easy to export from Google Analytics and other tools but at a more strategic level no one cares very much about what performance on 14th July looks like as long as July as a whole is doing OK. When planning for the year or months ahead it is much more normal to set a monthly target rather than look at things day by day.
My personal opinion is that working with daily data gives will give better results when you start adding helper columns - particularly helper columns that have an effect on a particular day (this is most of them in my experience). But it is also tedious to make a quick daily forecast and then have to aggregate the results to a monthly level before presenting them to a client or boss and I want to remove as much tedium as possible for my users.