Based on the actuals we have access to, we are spot on 15% of the time, within 5% (or 5 units) about 70% of the time and within 10% (or 10 units) about 90% of the time. When we are off, we are typically under biased.
Estimating sales is an imperfect science. We take several inputs such as customer reviews, sales rank, category, price, seasonality, etc. to create an algorithm that is trained with actual sales data to estimate the rest of the catalog, The best way to see the margin of error is to compare your actuals to our estimates to get an idea of how far off we might be. We don’t anticipate we will ever be at 100%, but as we grow and get more inputs, we will get better. Our goal is to give you a direction assessment as to how fast an item is selling.
The best thing you can do to help train our models is to add DataSpark in your list of Walmart solution providers. Instructions for doing that are included here: https://youtu.be/PxerqzHKrUI