Back to portfolio
recykalEPR Next · Recycler Portal
The outcome
Same recycler team, before and afterLess of everything that slowed recyclers down.
Days to process one brand, per year133 days back
19152
Before152 days
After19 days
Time per delivery37 min saved on every truck
8 min45 min
Fields to fill
1236
24 fields removed or derived
Documents per truck11 fewer, uploaded in any order
415
01 · Context
Every truckload is a compliance record.
Brands owe recycling targets under India's Extended Producer Responsibility rules. Recyclers prove each tonne, truck by truck.
Step 1BrandOwes a recycling target for the plastic it sells.
Step 2Purchase orderBrand asks a recycler to fulfil part of it.
Step 3DeliveryEach truck of waste becomes one record.
Step 44 documentsProof the material really moved.
Step 5CertificateApproved tonnes count toward the brand target.
EPR Next is where steps three to five happen, for plastic, e-waste, tyres and metal.
02 · The problem
36 fields. 15 documents. Per truck. Thousands of trucks a month.
Typing the same data twiceRepeat entry across forms caused errors.
Too many requirements36 data points and 15 documents per truck.
Documents everywhereShared over email and zip folders.
Excel ignoredTeams lived in spreadsheets. The system did not.
No view of demandHard to see what each brand still needed.
No e-way bill linkTransport details were copied by hand.
The math behind the wall
One brand. 152 days of data entry.
A back of the envelope count of the old workflow made the brief obvious. The system, not the team, was the bottleneck.
100,000 MTAvg. brand demand in a year
÷
6 to 8 MTAverage truck load
=
12,500Deliveries to track
×
45 to 60 minTime per delivery
=
12,500 hrsTotal effort
÷
10 × 8 hrsPeople × hours a day
≈
152 daysOf data entry per brand
Before · per brand
152
After · same team19working days to process a brand's yearly demand
Every square is one working day19 needed now133 given back
10 people · 8 hours a day · about 100,000 MT of brand demand a year
03 · Research
We watched the work before redrawing it.
12+In-depth interviewsRecyclers from small yards to large plants.
On siteShadowing ops teamsPurchase orders, trucks and paperwork, live.
End to endWorkflow mappingEvery tool and handoff, from purchase order to certificate.
Insight 01
“Our whole day already runs on Excel.”
So bulk upload starts from the sheet they already keep.Insight 02
“Bills come first, weighbridge slips come days later.”
So documents upload in any order, never in a fixed sequence.Insight 03
“If the system guesses, I want to see where.”
So automation shows its confidence and stays editable.04 · First attempt
Our first version just moved the pain online.
One delivery at a time. One document at a time. Same 36 fields. Workload barely moved.
The lesson
Digitizing is not simplifying.
The system had to remove work, not host it.
So we split it in two. Cut first, automate second.
Phase 1Reduce the work
Fields per delivery36 12
Documents per delivery15 4
Deliveries per upload1 Whole sheet
Phase 2Automate what's left
E-way bill integrationTransport details fetched with the recycler's consent.
OCR document taggingUploads are read, typed and matched to the right delivery.
Operational dashboardTargets, supply and blockers in one view.
05 · Key flows
Five flows carry the whole product.
Shown as they ship. Read the screens, skim the notes.
01
Purchase DataOne spreadsheet in. Dozens of deliveries out.
Bulk upload turns each invoice row into a delivery.
1
Drop several files at onceProgress per file, so nothing feels stuck.
2
See what each file createdInvoices found, deliveries created, documents pending.
3
Fix only the rows that failedEach row says what went wrong and where.
02
Delivery detailDrop four documents. We fill the form.
OCR reads each document, tags it and fills the delivery.
1
Upload in any orderFiles are classified and matched as they land.
2
Check against the sourceRead values sit beside the document itself.
3
Recover when a scan is badReplace the file or type the value yourself.
03
Create deliveryNo spreadsheet? Add a delivery by hand.
A short form that saves as a draft and checks itself.
1
Errors that name the fixFormat, missing values and date order, inline.
2
Lands in the list as NewA toast confirms it and points to the next step.
3
Documents can come laterCreate now, upload the 4 files when they arrive.
04
Manage DeliveriesEvery delivery, sorted by what it needs next.
Status tabs and document filters replace a long flat list.
1
Unallocated, In review, AllocatedTabs follow the life of a delivery.
2
Pending or uploadedSegments and filters show what is blocking review.
3
One clear next action per rowUpload docs, Continue or Send for review.
05
AllocationTurn approved tonnes into client credit.
Allocate reviewed quantity to brand purchase orders.
1
Supply by categoryApproved quantity grouped the way brands buy it.
2
Preview before you commitRemaining supply and PO balance update live.
3
Certificate requestedAllocation and certificate start in one step.
06 · Design choices
Small decisions that did the heavy lifting.
07 · System
Built on a system, not a pile of screens.
One type family, one blue, four semantic tones. Every screen and every state is assembled from the parts below.
Type · Figtree
Page title30 / 38 SemiBold
Card title18 / 28 SemiBold
Input value16 / 24 Regular
Label and table text14 / 20 Medium
Table header, badge12 / 18 Medium
Color · meaning first
BrandActions, focus
Gray 900Headings
Gray 600Body text
SuccessMatched, issued
WarningCheck this
ErrorBlocked
Shape and depth
8 px · inputs, buttons
12 px · cards
Overlay · toasts, panels
4 px spacing scale
Buttons
Badges
MatchedCertificate issuedLow confidenceDocs pendingCouldn't readIn reviewNewMissingCat 14
Inputs · five states
Check it against the document.
Enter a 6 digit pin code.
With unit
Select, date and search
Cat 2 · Flexible plastic
Cat 1 · Rigid plastic
Cat 2 · Flexible plastic
12/12/2024
Search⌘K
Navigation
Unallocated8In review3Allocated3
AllDocs pendingDocs uploaded
Manage DeliveriesDL#12281
Deliveries8Allocation
Selection
Deliveries that need changesIn app
File processing finishedEmail
Feedback
Ready to create. Upload its 4 documents later.
2 fields need review. Read with low confidence.
6 fields need attention. Fix them to create this delivery.
DL#12288 createdUpload its 4 documents to send it for review.
Progress and files
2/4 documents
520 MT
XLS
Telangana_DLs.xls70%
AD
Aditya DesaiAdmin
Table · pinned ID and action
Delivery IDAggregatorVehicle no.Docs uploadedActionDL#12288NewGreen Loop TradersTG08 TE22310/4Upload docs
08 · Impact
What changed on the recycler's floor.
09 · Learnings
What I'm taking to the next product.
01
Reduce before you automate.Cutting 36 fields to 12 did more than any model.
02
Automation should assist, not replace.People trust it when they can see and correct it.
03
Design around the existing habit.Recyclers already lived in Excel, so we started there.
04
Confidence is a UI element.Showing doubt tells people exactly where to look.