Lovelocal
Head of Product
Lovelocal (previously m.Paani) is a platform that helps Local retail compete against big-box retail and online e-com giants.
​
This Google accelerator company started as an SMS-based loyalty program for India's local retail stores to improve their customer retention.
​
In 2019, the company pivoted to a unified hyper-local marketplace (B2C) + retail store management & marketing solution (B2B).
My Contrubution
The Challenge
The company's B2B Retailer App replies on the store staff proactively checking ordering via the Lovelocal & processing the order. The SLAs we had set for identifying an order and starting to process it had been set at 15min. Anything above 15min, our customer service team would call the retailer and relay the same order & request them to process it.
During the first month of launching the product we immediately noticed a problem, 84% of orders needed our customer service teams to intervene & for the 37% of all orders processed by the retailer, the incorrect products were being delivered.
Solution Design
Step 1. Analysing the problem.
As I spent time in a few stores to observe the operational flow on-ground, a few things were suddenly apparent.
1. All home delivery orders were over a phone call with the customer. Getting the store staff to view an app to get the orders was out of their regular operational flow.
2. They staff are extremely stretched during rush hour.
​
3. Many of the staff members either didn't know to read well, even in their local language.
Step 2 - Keys to success.
My mandate was clear:
​
1. I needed to develop a feature that fits within the current operational flow for the shopkeepers & their staff.
​
2. The solution needed to be extremely quick & needs to account for different litracy levels.
Step 3 - Leveraging Natural Language Processing
-
The solution we deviced was to use NLP (Natural Language Processing) to dictate the order as if it was a phone call.
-
The same could be translated into 4 languages - English, Hindi, Gujrati, Kannada.
-
The conversation flow of the recording used mimicked that of an actual phone call that the user
Key Results
Order Accuracy increased by 25%.
Intevention from customer success dropped by 44%