Sales Predictions shows you the likelihood that...
- A particular listing title
- At a particular price
- In a particular category
- With a particular item condition
- Using a particular listing format
...will sell on eBay.com in the next 30 days.
How Probabilities are Calculated
- A category is determined for the item. Sales Predictions will either use your existing category (for eBay listings in a linked seller account) or will select an appropriate eBay category for your product based on the keywords that you've supplied (for Shopify listings).
- Individual probabilities are calculated. Sales Predictions then uses the performance of other items that have already sold in the same category to measure how each of the following is correlated with sales:
- Up to eight of your keywords (beginning with the keywords that perform best in the category)
- Your item price
- Your item condition (for listings from a linked eBay account)
- Your listing format (for listings from a linked eBay account)
The combined probability is the figure shown in your Sales Predictions results.
How Accurate Sales Predictions Are
The Sales Predictions algorithm is based on Terapeak's own research and is regularly tested to ensure that it is predictive of actual sales on eBay.com.
On average, predictions are very accurate for eBay.com in its entirety. There are two important caveats, however, to consider:
- Accuracy varies by category. Though accuracy across all eBay categories together is very good, predictions are more accurate for some categories than for others. Categories that see very low listing volume or that have highly unique or varied items, such as some collectibles categories, are more difficult to accurately predict.
- The predictions are probabilities. As is the case with all probabilities, real results can in some cases vary from the most probable outcomes. Just as you might flip a coin ten times in a row and receive heads each time—even though the probability of heads is only 50 percent for each flip—it is also possible that listings might perform better or worse than was probable in real-world cases.