Grofers Ad bidding platform is a medium to facilitate auctions on different monetisation assets. The bidding takes place on 48 keywords on PAN-India level for three assets - Sponsored Product, Sponsored Brand and Search Display Banner. Currently we have around 65 brands associated with Grofers.
Although, we have at least one brand associated with a keyword, we noticed there are few keywords those were never bid across all assets. The goal was to find out reason behind this behaviour and how we could resolve this problem in the best possible way.

We brainstormed and listed down some hypothesis of what all could be the possible reasons, then we voted for them. The winner was "Brands did not find any value in investing on those keyword". Even if this was the main problem, we were still not sure why this was happening.

While looking at bidding numbers we noticed a pattern that most local brands although associated with a keyword avert themselves from bidding on them.
When we tried to dive down deeper into the problem and noticed that those keywords left out of bidding had only regional brands associated with them. This pointed us out that regional brands are not finding value in investing on their associated keywords, but still we were not sure why this was happening.
One obvious hypothesis that came up was that regional brands would not want to bid at PAN-India price as they know their presence is limited to a particular region.

Other important learnings from data in regional wise bidding were;
Finding #1  At least 40 brands out of 65 can potentially benefit from advertising locally in their strong zones
Finding #2 Local brands can always benefit irrespective of whether they are already top selling in their area or not
Finding #3 Local top selling products keep varying in regions. Still, two different campaign periods showed a consistency of 65%  

Data analysis clearly showed us the bigger problem we had, but at this stage we were not sure of the solution.

Till now we were sure we identified the problem correctly, but still had doubts that changing the bid from PAN-India level to regional level or city level would be beneficial overall. Also, we were not sure how national brands might react to this change.
To clear all our doubts we arranged some user calls for national and regional brands, and with our sales executives who frequently interact with brand representatives.
‍
From user insights two conclusive points came out;
Finding #1 Regional brands would prefer city-wise bidding on some keywords and PAN-India level bidding on other keywords.
Finding #2 National brands want to stick with PAN-India bidding for most keywords and can explore city-wise bidding for some keywords.

Once we better understood our users and explored our datasets, we began additional ideation for our vision. We quickly settled on the basic components and layout of our visualisation through several quick rounds of sketches. We explored interactions around our platform and how we could display the new city-level option bidding along with PAN-India one. Since the Information Architecture now became more complex, we went through it once again to be sure the new bidding structure makes sense to users.

We went a little back and fourth with our iterations, did some usability test round it as well to make sure we didn't miss on any point.

Some revelations that came forward were;
#1 Time is money
Our bidding gets very intense during last hours, to make sure we didn't kill user's precious time we decided to keep our city list open and removed accordion which we had in our first iteration so as to save clicks.
‍
#2 Excel sheet lovers
Our users are very handy with excel sheets and most of the time they get bidding values in excel sheets. so we added an option to bid through excel sheets as well.

#3 Keyword first or City first
We were initially not sure about our approach on city first or keyword first, we did a few iterations around it and tested with people and had our clear winner.

Based on our explorations and study, we felt very confident in our direction going into the development of our final platform design.

What we did well
As a group (me and my Monetisation teammates) we feel that we were successful in developing and implementing a bidding system that allows users to explore city-level as well as PAN-India level bidding. We used the similar visualisation we already had to make user experience easier.

We considered minute things that could create a larger impact on users. For example, we knew our users don't have much time while bidding and every minute counts for them so we kept the city list open and also removed some extra redundant information. We also kept an option to bid through csv files as we knew our users are very comfortable with excel sheets. All these things came out while user testing, we feel that our process was able to define the issue and give us time to address it properly, which would label this change a success.


What could be done differently
An area where we feel we were not entirely successful was helping users guide the city to bid in. While exploring other platforms we realised their systems have algorithms to take care of city-wise bidding internally. Additional exploration in this direction was deprioritised due to time constraints. But, this could have been a very interesting addition.