Is there an AI tool that can predict what we need to order?
Discussion
We have a small business that's growing well but due to a lack of space we dont hold much stock. The products are manufactured locally in a process that takes around 2 weeks - When it gets busy this makes it hard to stay on top of things.
Our best selling product is ordered in batches of 20 then we re-order when we notice stock has dropped to around 10. By the time the new batch is arriving we've just about used all the original batch.
The store runs on Woo commerce but currently stock is managed by us simply looking at what we have and making a decision based on our experience and current stock levels.
There are a few products that we order in batches of 4-6pcs when current stock drops to 1-2pcs.
We've been doing this for around 3-4 years so there is some reasonable data to work with. However, we're regularly introducing new products which clouds the data a little as we never know if something is going to sell well until a few months after release.
Without going to a traditional ERP or MRP solution (which will be too costly and cumbersome) is there some way I can use AI to predict what we need to order and when?
I've done a fair amount of research on this but its hard to know what's just a fancy website and what is actually suitable for us. I suspect many of the solutions are actually no different to an Excel spreadhseet with a macro and aren't actually AI based.
www.ausmotion.com is the business
Our best selling product is ordered in batches of 20 then we re-order when we notice stock has dropped to around 10. By the time the new batch is arriving we've just about used all the original batch.
The store runs on Woo commerce but currently stock is managed by us simply looking at what we have and making a decision based on our experience and current stock levels.
There are a few products that we order in batches of 4-6pcs when current stock drops to 1-2pcs.
We've been doing this for around 3-4 years so there is some reasonable data to work with. However, we're regularly introducing new products which clouds the data a little as we never know if something is going to sell well until a few months after release.
Without going to a traditional ERP or MRP solution (which will be too costly and cumbersome) is there some way I can use AI to predict what we need to order and when?
I've done a fair amount of research on this but its hard to know what's just a fancy website and what is actually suitable for us. I suspect many of the solutions are actually no different to an Excel spreadhseet with a macro and aren't actually AI based.
www.ausmotion.com is the business
I don’t have the exact figures but there will be hundreds of tools that can do this. Most of them will be simple Machine Learning applications, probably fronting a Python back end.
Given the complexity of the problem (sounds very low) I’m not sure why it has to be AI based though. If a simple Excel Macro gets the job done would that not have the desired result?
The ML based solutions are just doing some simple statistical best fit modelling. When you know how it works it doesn’t sound like AI. So does that rule it out as well?
Given the complexity of the problem (sounds very low) I’m not sure why it has to be AI based though. If a simple Excel Macro gets the job done would that not have the desired result?
The ML based solutions are just doing some simple statistical best fit modelling. When you know how it works it doesn’t sound like AI. So does that rule it out as well?
There are 2 tools available:
1. ForecastForge
ForecastForge is a cloud-based tool designed to simplify demand forecasting for businesses. It leverages machine learning algorithms to analyze historical sales data and generate accurate sales forecasts. The platform is user-friendly, requiring minimal technical expertise, and integrates seamlessly with various data sources.
2. Inventory Planner
Inventory Planner is a comprehensive inventory forecasting and planning solution tailored for e-commerce and retail businesses. It offers features such as demand forecasting, purchasing recommendations, and reporting analytics. The platform integrates with numerous e-commerce platforms and accounting systems, providing real-time insights into stock levels, sales trends, and product performance.
If you need help deploying them, please message me.
1. ForecastForge
ForecastForge is a cloud-based tool designed to simplify demand forecasting for businesses. It leverages machine learning algorithms to analyze historical sales data and generate accurate sales forecasts. The platform is user-friendly, requiring minimal technical expertise, and integrates seamlessly with various data sources.
2. Inventory Planner
Inventory Planner is a comprehensive inventory forecasting and planning solution tailored for e-commerce and retail businesses. It offers features such as demand forecasting, purchasing recommendations, and reporting analytics. The platform integrates with numerous e-commerce platforms and accounting systems, providing real-time insights into stock levels, sales trends, and product performance.
If you need help deploying them, please message me.
Your products are quit bulky so kanban with a twist is possibly the answer.
If you have the space you can split the batches into 2 locations and once empty have a simple red green card you can flip round with the product details on, once a week you look for all the red cards and reorder the items.
Very simple and effective, for automated solutions I would have thought your e commerce tool would have a stock level (min/max) function that will alert you when stock drops below a specified level, you will need to keep it up to date though.
If you have the space you can split the batches into 2 locations and once empty have a simple red green card you can flip round with the product details on, once a week you look for all the red cards and reorder the items.
Very simple and effective, for automated solutions I would have thought your e commerce tool would have a stock level (min/max) function that will alert you when stock drops below a specified level, you will need to keep it up to date though.
Edited by Bertrum on Thursday 28th November 19:04
Fascinating subject 
I won’t offer specifics because although this is what I do for a job, it is for very very large retailers and the particular solutions are bespoke. The post from paddy1970 highlights a key concept of the importance of forecasting demand (what’s going out the door) being the backbone for good stock management and therefore what needs to come in through the door.
If you can get a good feel for demand over a future time period then you can apply different stock maintenance policies for different products to be lean but with enough buffer stock at any time. For new products the aim is to have the system move from new product management methods to methods for established products as performance data builds up. Then you can layer in promotions uplifts etc on top.
ML can help in all these areas in recognising patterns to use to pick amongst long established forecasting and replenishment techniques. ML is becoming much more pick and mix so getting easier to implement. If you can find one that is already working successfully for similar companies as yours that is a good start. The thing that is often forgotten is making sure you objectively measure how good any predictions perform against reality.
Also to note is this. The leaner you want your stock levels to be the more sophisticated you want to be. It is a trade off and sometimes carrying a slightly higher level of stock is worth the simplicity you gain.
Yes some current claimed AI is just rebranded what was there before!
I won’t offer specifics because although this is what I do for a job, it is for very very large retailers and the particular solutions are bespoke. The post from paddy1970 highlights a key concept of the importance of forecasting demand (what’s going out the door) being the backbone for good stock management and therefore what needs to come in through the door.
If you can get a good feel for demand over a future time period then you can apply different stock maintenance policies for different products to be lean but with enough buffer stock at any time. For new products the aim is to have the system move from new product management methods to methods for established products as performance data builds up. Then you can layer in promotions uplifts etc on top.
ML can help in all these areas in recognising patterns to use to pick amongst long established forecasting and replenishment techniques. ML is becoming much more pick and mix so getting easier to implement. If you can find one that is already working successfully for similar companies as yours that is a good start. The thing that is often forgotten is making sure you objectively measure how good any predictions perform against reality.
Also to note is this. The leaner you want your stock levels to be the more sophisticated you want to be. It is a trade off and sometimes carrying a slightly higher level of stock is worth the simplicity you gain.
Yes some current claimed AI is just rebranded what was there before!
Edited by C4ME on Thursday 28th November 22:19
Hi
I consult in this area for a number of FTSE 100 retailers. Simply put you could do this a number of ways but each will require you having the required inputs. For examples 3 years sales history with lost sales, a forward predicted demand forecast could be simple as a moving average or last year +/- a bit. You will also need to define you warehouse stock holding..the higher the variability of demand the higher the safety stock requirement.
I could go on and bore everyone …happy to advice for free for a fellow pistonheader.
Also what the poster above said !
I consult in this area for a number of FTSE 100 retailers. Simply put you could do this a number of ways but each will require you having the required inputs. For examples 3 years sales history with lost sales, a forward predicted demand forecast could be simple as a moving average or last year +/- a bit. You will also need to define you warehouse stock holding..the higher the variability of demand the higher the safety stock requirement.
I could go on and bore everyone …happy to advice for free for a fellow pistonheader.
Also what the poster above said !
Edited by magic360 on Saturday 4th January 19:22
C4ME said:
Fascinating subject
I won’t offer specifics because although this is what I do for a job, it is for very very large retailers and the particular solutions are bespoke. The post from paddy1970 highlights a key concept of the importance of forecasting demand (what’s going out the door) being the backbone for good stock management and therefore what needs to come in through the door.
If you can get a good feel for demand over a future time period then you can apply different stock maintenance policies for different products to be lean but with enough buffer stock at any time. For new products the aim is to have the system move from new product management methods to methods for established products as performance data builds up. Then you can layer in promotions uplifts etc on top.
ML can help in all these areas in recognising patterns to use to pick amongst long established forecasting and replenishment techniques. ML is becoming much more pick and mix so getting easier to implement. If you can find one that is already working successfully for similar companies as yours that is a good start. The thing that is often forgotten is making sure you objectively measure how good any predictions perform against reality.
Also to note is this. The leaner you want your stock levels to be the more sophisticated you want to be. It is a trade off and sometimes carrying a slightly higher level of stock is worth the simplicity you gain.
Yes some current claimed AI is just rebranded what was there before!
We should compare notes !!I won’t offer specifics because although this is what I do for a job, it is for very very large retailers and the particular solutions are bespoke. The post from paddy1970 highlights a key concept of the importance of forecasting demand (what’s going out the door) being the backbone for good stock management and therefore what needs to come in through the door.
If you can get a good feel for demand over a future time period then you can apply different stock maintenance policies for different products to be lean but with enough buffer stock at any time. For new products the aim is to have the system move from new product management methods to methods for established products as performance data builds up. Then you can layer in promotions uplifts etc on top.
ML can help in all these areas in recognising patterns to use to pick amongst long established forecasting and replenishment techniques. ML is becoming much more pick and mix so getting easier to implement. If you can find one that is already working successfully for similar companies as yours that is a good start. The thing that is often forgotten is making sure you objectively measure how good any predictions perform against reality.
Also to note is this. The leaner you want your stock levels to be the more sophisticated you want to be. It is a trade off and sometimes carrying a slightly higher level of stock is worth the simplicity you gain.
Yes some current claimed AI is just rebranded what was there before!
Edited by C4ME on Thursday 28th November 22:19
Gassing Station | Business | Top of Page | What's New | My Stuff