Inventory Counting "Product Recognition" & Demand Forecasting "Suggested Purchase"
Merchants often juggle complex tasks with pen & paper and whatever they can keep in their heads, from managing inventory to predicting sales. Intelligent tools can offload much of this labor and increase returns for informal businesses. Some examples include:
- Inventory tools: Smartphone-camera tools can identify products and track inventory movement, recording sales at higher volumes and providing a valuable input into working-capital loans
- Price prediction: Eliminates the guesswork of product pricing to optimize sales, especially for new shopkeepers and new products
- Demand forecasting: Recommends inventory to purchase based on historical data, optimizing a shopkeeper’s precious working capital
- Receipt digitization: Data input with the snap of a photo saves time, increases accuracy, and bridges to formal business practices
Needs Addressed. MSMEs need to be able to track products sold to determine which and how many products to re-order when replenishing their inventory. The ability to simply point a camera at a product and be provided with pricing and ordering quantity suggestions, would free up merchants from the tedious process of mental or manual calculations based on a historical paper ledger and tracking down stock keeping units (SKUs) for ordering.
Technology. Using computer vision and object classification, we can train a model to translate key features of a live camera image into a product identifier, then look up information on pricing and availability. The image-based search technology in Amazon's mobile application is an example of such technology already in use, and what we used for demonstration purposes.
Validation. In visiting dukas in the Kibera district of Nairobi, Kenya, to demonstrate a proof of concept of the product recognition using Amazon's app, we received generally positive feedback. While the initial reaction was, "this looks like an app for the younger generation," after a minute of playing with the app, the merchant explained that he'd love to have such an app for ordering -- without having to wait for the wholesaler to come around.
Gap. The main gaps here once again were around localized product data and quality of hardware, as described in Use Case 1 (Product Counter for Estimating Sales to Assess Credit Worthiness). The prescriptive pricing algorithms also benefit from network effects in data provided by a critical mass of merchants in sufficiently close proximity, and may be limited in effectiveness in their initial rollout.