Skip to content
Non Compliant VS -
Spike driven by 2 PIDs contributing to 40% illegal VS
LbyS trends - LbyS Drop primarily dominant for Sarees and Lehengas, Run-rate expected to come back in Dec 1st week. Expected seasonal drop post Diwali.
Listing funnel view - 
Go live rates dropped on Nov 11th - This was majorly due to:6 suppliers uploaded unexpectedly high number of duplicate products which were deactivated upon listing. These suppliers contributed to 89% of the spike. All of these suppliers were from Surat.
p70 go live times have spiked to ~1.8 hrs in Diwali week. RCA - Pending to update
P0 and search Attribute fill-rateImage fill-rateSearch views per catalogPanel activity Net Orders / ViObyVTotal number of products
Attribute fill rate: 0.1% increaseP0 attribute fill rate increased by 0.11%Search attribute fill rate increased by 0.07%Total Orders/ Views: +0.9%GMV/Views: increased by 0.24%~35% of the test suppliers have visited the feature. 60% of them have interacted with the tool.
Seller legal policy update on seller panelP0 ask from legal from a hygiene perspective. No product support needed, FE bandwidth driven.Awaiting updated legal policy doc from Views dipstick module DBE load optimisation - BE fix needed to streamline DBE load from the new views dipstick serviceRepeat offender updated screens - False positive QA on admin - PRD
FSSAI license repo - Deprioritised this R2R
Self serve bank account updateGSTIN OTP provider update - Cleartax vs SurepassCost - Surepass - Rs 3 per requestCleartax - Rs 1.5 per requestImplementation caveats - Cleartax integration would require significant dev effort since the flows are different vs SurepassSurepass - Returns GSTIN linked phone number on passing GSTIN as input / OTP to the number triggered via in-house communication engineCleartax - Returns a binary flag in case phone number doesn’t match GSTIN linked number and then triggers OTP to GSTIN linked number through a separate API (GST advanced ownership API - Rs 1 per hit / OTP API - 0.5 per hit) - Method claimed to be more compliant as per Cleartax team Delaying go-live for Lehengas and Blouses - Need more details from Capability exists and can be handled w/o dev with ‘s supportReverse weight update CRON setupMinor dev to read weight reference in the existing CRON as either forward or reverse to make the subsequent update - POC Temp or permanent delete cases update - Logging to be picked by GunitDev picked by The flow affected cases where seller was temp or permanently deleted due to Bank / GST KYC issues and lead to PSMs marked as permanently blocked, Lead to issues in reversal of FP identification / EFDRemoving free shipping tile from seller landing pageNeed to check whether this will need BE /FE dev support or can be handled by Biz using Webflow toolExtending edit of validation failed mobile uploads in seller web panelGeneral upload guidelines revamp - HygieneWAAS build solutioning support
v1 Experiment - QC coverage / readinessIncreasing agent coverage - Need to increase agent allotment here by reducing view dipstick coverage temporarily - POC Dev updates - E2E dev testing in place, Exp start date 27th NovFirst experiment - Incremental images from duplicates - To be kickstarted 29th Nov
Google draping POC - Cost approval for 3D model / Obj procurement post internal alignmentTeam to share flat / raw images of Sarees with complex designs / prints
Almost ~2.7K PIDs were not flagged by the DS model Keywords/Alternate keywords for these PIDs are already fed to the DS models, but they were still not flagged. We have shared the list of ~2.7K PIDs with the DS team to identify the reason for the miss in the DS modelsWe are awaiting DS analysis on this. Will update on the next steps post alignment with DS team. Majority of the leakage in agent miss is coming from profane images.80% of these PIDs were flagged by DS in text based model These PIDs should have been deactivated in the first QC but were identified in view dipstick process Re-training agents on guidelines for QC for profane images will be implemented this weekBasis feedback from respective category teams in case of escalations, legal alignment (in case of subjective calls wrt profane images) will be takenThese PIDs are already flagged for clean up activity that is to be performed for ~12L PIDs (which includes both brand infringement as well as illegal listings)These agents are currently working on auditing image enhancement experiment PIDsThese agents will begin the clean up activity for these 12L PIDs by early next week once the QC for image enhancement is completed This will get roughly 1 week to complete. These PIDs belonged to the Profane images cohort which were wrongly deactivated in the first QC process Agent training for profane images guidelines These PIDs will be re-activated as per guidelines These PIDs were flagged for deactivation but there was a time lag, hence have been caught in view dipstick process. These PIDs have now been deactivated
Mentor sync
Key metrics -
Note - There was a Redis timeout issue on Friday (10th Nov), just before the RnR break. This issue caused a spike in drafts and validation failures. However, it was resolved quickly and the metrics returned to stable within a few hours.
Experiment updates -
Listing optimiser update
Listing optimiser was launched for ~35k sellers on 7th Nov.
Homogenisation criteria - Sellers were homogenised on -
Early reads-
GTM for the experiment was started on 23rd Nov and we expect these numbers to mature by 1st week of Dec.
Compliance roadmap -
Cataloging KTLO pendencies -
@Amitoj Singh
Merchandising updates -
Step 1 : QC / Audit completed for ~1L / 3% VC PIDs
Currently the team is able to deliver 700 PIDs per agent in 8 hours.(35 Resource utilised here)
In order to QC each PID the team is spending 40 Sec (18 Images to be compared against each PID)
Also will be pausing the view dipstick twice a week and those resource will be utilised to this excise and by doing this will be able to get additional uplift of 50K PIDs per week.
Step 2:
Based on the initial validated results (Where ever the team has mentioned to use the image will be re-qced by the in-house audit agents)
This exercise is started since 22 Nov and post completing the second level of QC the data can be used for the experiment. (Should consists => 7 Images per PID, Including the original Images)
Additional 10 resource from the taxonomy team will be used here from 24 Nov (Overall 17 Resource working on re-audit process and expected to deliver 17K PIDs per day)
Step 3:
If the Validated PID image size is less then 7 then the re-audit will be done on against the images where the team has mentioned not to use.
@Prashant Kumar
Illegal VS RCA:
In the last 3 weeks, we have seen a spike in the view share for illegal listings on the platform.
Additionally, we have also seen an increase in #PIDs flagged in the view-dipstick process by agents.
Note : ~1200 PIDs every week in the last 3 weeks are being flagged as compared to the BAU ~800 PIDs.
Roughly ~30 PIDs on an average contribute to 50% of the illegal view share every week.
Of the ~4.6K illegal PIDs flagged in the view dipstick process in the last 30 days, we have collated the reason for the miss in deactivating these PIDs earlier :
View dipstick analysis
Reason
%
Details
Next steps
PIDs missed by DS model
~58%
Agent Miss
~22%
Yet to picked up in the cleaning activity
~11%
Wrongly deactivated
~5%
BAU deactivation
~4%
There are no rows in this table
As next steps, we will set the view dipstick analysis as a recurring process at a weekly level to identify potential actionable/improvements in the DS model and operational process.
Want to print your doc?
This is not the way.
This is not the way.
Try clicking the ··· in the right corner or using a keyboard shortcut (
CtrlP
) instead.