Making 'Excellent' mean the same everywhere · Back Market

Making 'Excellent' mean the same thing at every refurbisher.

A concept for a grading assistant Back Market could give its refurbisher partners. A photo booth and AI mark scratches, diagnostics run automatically, and the technician confirms the grade, so 'Good' or 'Excellent' means the same at every refurbisher.

Self-initiated concept. Not affiliated with or endorsed by Back Market.

Role: Self-initiated concept, from workshop visit to tested prototype
Company: Back Market (concept)
Timeline: 2024 · 6 weeks, evenings
Team: Solo, with 4 refurbishers and 6 technicians who joined research and tests
Platforms: Web app for refurbisher partners, Photo booth at the grading bench, Label printer

1Scratches are marked on the photo where they were found
2The grade uses the same definitions at every refurbisher
3The technician confirms the grade before it counts

58% → 88%

Technicians agreeing on a grade, blind test

−35%

Time per device in the prototype

3 of 4

Refurbishers who'd adopt it

Shared grading guide, built into the tool

The problem

Back Market is a marketplace: independent refurbishers test and grade the devices. When grading depends on each technician's eye and lighting, a phone sold as 'Excellent' by one seller can look like 'Good' from another, and buyers notice.

What I did

I visited a refurbishing workshop in Rotterdam, interviewed four refurbishers, ran a blind grading test with six technicians, and designed a grading assistant in Back Market's visual identity: booth photos with marked scratches, a suggested grade with reasons, and the technician's confirmation.

What changed

In a second blind test with the prototype, agreement between technicians went from 58% to 88%, and grading took about a third less time. Three of the four refurbishers said they'd use it if Back Market offered it.

01 · Context

Why I picked Back Market.

Back Market is one of Europe's largest marketplaces for refurbished electronics. It doesn't refurbish devices itself: hundreds of independent refurbishers do, and Back Market sets the quality standards. The grade a buyer sees, Fair, Good or Excellent, is the promise everything rests on.

This is a self-initiated concept. I don't work for Back Market or its partners. The research is with refurbishers in the Netherlands, and the design follows Back Market's public visual identity.

01Device arrivesTrade-in or bulk purchase
02DiagnosticsBattery, screen, cameras
03Cosmetic gradeJudged by eye today
04Listed on Back MarketFair, Good or Excellent
05Buyer receivesCompares with the promise

Where inconsistency comes from

Fig. 1The path of a device, from refurbisher to buyer.

02 · Problem

One phone, three grades.

I asked six technicians from two refurbishers to grade the same 20 phones under their normal workshop light, without talking to each other. They agreed only 58% of the time.

Device	Technician A	Technician B	Technician C	Agreed grade
iPhone 13, Midnight	Excellent	Good	Good	Good
Galaxy S21, Grey	Good	Fair	Good	Good
iPhone 12 mini, White	Fair	Good	Fair	Fair
Pixel 6, Black	Good	Good	Excellent	Good
iPhone SE, Red	Excellent	Excellent	Good	Excellent

Fig. 2Five of the 20 phones from the first blind test.

The question I worked on

How might every refurbisher grade the same way, quickly, while keeping the technician's judgement where it matters?

03 · Research

A day at the grading bench.

I spent a day in a refurbishing workshop in Rotterdam, interviewed the owners of four refurbishers, and read 300 public reviews on refurbished-phone marketplaces that mention a device's condition.

Insight 1

Light changes everything

Scratches visible under a bench lamp disappeared under ceiling light. Each technician tilted the phone differently to find them.

“Under this light it's Excellent. Under that one it's Good.”
BramTechnician, Rotterdam

Insight 2

Standards live in people's heads

Every workshop had its own version of the rules, passed on by whoever trained new people.

“I grade like my trainer graded.”
NoorTechnician, 4 months in

Insight 3

Buyers return surprises

Negative reviews rarely said the phone was broken. They said it was worse than described, usually a scratch nobody mentioned.

“It works fine, but 'Excellent' it is not.”
Public review, paraphrased

04 · Process

Six weeks, tested at the bench.

Week 0123456

Workshop day, interviews

First blind test

Grading guide written down

Assistant design

Second blind test

◆ 58% agreement◆ 88% with the prototype

Fig. 3Autumn 2024, alongside my job.

Option ARejected

AI grades on its own

Refurbishers are responsible for their listings and didn't want a model deciding for them.

Option BTested

Printed grading guide

Helped a little. Still depended on light and on how carefully people looked.

Option CChosen

Booth photos, marked scratches, technician confirms

Same light every time, marks on the photos, a suggested grade with reasons, and the technician decides.

Fig. 4Three approaches tried with the workshop.

05 · Design decisions

Same light, visible marks, a person decides.

The grade comes with its reasons

The assistant shows a suggested grade with the marks it's based on, circled on the booth photos. The technician confirms or changes it, and a change asks for a reason, which is how the shared guide keeps improving.

Excellent

Like new

No visible marks under booth light from four sides.

Good

Light signs of use

Up to 3 light scratches, none on the screen.

Fair

Visible wear

Scuffs on the frame or a light screen scratch.

Repair

Not sellable yet

Cracks, dead pixels or a failed diagnostic.

Fig. 5The shared grading guide, built into the assistant.

The queue sets the pace

The bench shows the next devices in line and a simple daily counter. Technicians said it felt fair because everyone sees the same queue.

Queue · Back Market Grading

←↻Queue · Back Market Grading–✕

Queue · Back Market Grading✕+

Back MarketGradingGradeQueueCalibrationGrading guide RePhone Rotterdam · G-07Tue 14 Jun · 13:05AK Aylin K.

Queue · Station G-07

RePhone Rotterdam · morning shift 06:00 to 14:30

This stationAll stations Scan tote

Graded today

142of 160 target

18 to go · 1 h 25 min left in the shift · on pace for 170

060912

Time per device

2:52

−1:18 vs grading by eye

Waiting

devices in 4 totes · oldest 09:12

Incoming devicesAll 24Phones 17Laptops 4Tablets 3Oldest first

ToteDeviceSourceArrivedWaitingStatus

T-0417iPhone 13 · 128 GB · MidnightBuyback09:123 h 53 min On station

T-0417iPhone 12 · 64 GB · WhiteBuyback09:123 h 53 minNext

T-0417iPhone SE · 64 GB · RedBuyback09:123 h 53 minWaiting

T-0417Galaxy A52 · 128 GB · BlackBuyback09:123 h 53 minWaiting

T-0418MacBook Air M1 · 256 GB · Space GreyWholesale10:402 h 25 minWaiting

T-0418iPad 9th gen · 64 GB · SilverWholesale10:402 h 25 minWaiting

T-0418Galaxy S21 · 128 GB · Phantom GreyWholesale10:402 h 25 minWaiting

T-0419Pixel 6 · 128 GB · Stormy BlackReturn11:052 h 00 minActivation lock on

T-0419ThinkPad T14 Gen 2 · 16 GB · 512 GBReturn11:052 h 00 minWaiting

T-0420iPhone 14 Pro · 256 GB · Deep PurpleBuyback12:3134 minWaiting

T-0420AirPods Pro · 2nd gen · case onlyBuyback12:3134 minWrong station

Recently gradedView all 142

13:02iPhone 11 · 64 GBFair€219

12:59iPhone XR · 128 GBGood€189

12:55MacBook Pro 13 · 2020Excellent€749

12:51Galaxy S20 FE · 128 GBGood€209

12:47iPad Air · 4th genFair€329

12:44iPhone 12 mini · 64 GBGood€279

Sent to team lead6

Grade changed by two steps, or the model was less than 60% sure.

Fig. 6The bench queue with today's progress. Numbers are illustrative.

Back Market can see where grading drifts

A calibration view shows how often each refurbisher's grades match the assistant's suggestions. Disagreements go to review, which would let Back Market keep one standard across hundreds of partners.

Calibration · Back Market Grading

←↻Calibration · Back Market Grading–✕

Calibration · Back Market Grading✕+

Back MarketGradingGradeQueueCalibrationGrading guide RePhone Rotterdam · G-07Tue 14 Jun · 13:05AK Aylin K.

Grading consistency

RePhone Rotterdam · all 8 stations · 7 to 13 Jun · 8,214 devices graded

PhonesLaptopsTabletsAll Export

Technician agrees with AI

88.9%

+3.2 pts vs last week

Matches Back Market audits

93.1%Top 10% of partners

320 devices spot-checked

Changed on review

2.3%

189 devices

Not-as-described returns

0.8%

Was 3.1% before the assistant

AI suggestion vs final grade

Share of each suggested grade

AI ↓ · Final →ExcellentGoodFairRepairDevices

Excellent91.2%8.1%0.7%01,904

Good5.6%88.9%5.2%0.3%3,412

Fair0.4%9.8%86.1%3.7%2,208

Repair00.6%7.9%91.5%690

Most changes are one grade apart. Good to Fair is the most common change, mostly lens scratches. Premium follows from battery and parts checks.

Agreement over time

All partners

Model v3.2 in W24 · 88.9%

W17 · 79.4%W24 · all partners 84.2%

By technician

Agrees with AI

Aylin K.92.4%

Sanne D.91.0%

Jonas P.90.2%

Ilse B.89.1%

Tomasz W.88.0%

Mark V.81.6%

Disagreements to review14Newest first

ExcellentFairiPad Air · 4th gen2 apartDent on the corner, missed by the modelAylin K.G-07

GoodFairiPhone 12 · 64 GBHairline scratch on the camera lensMark V.G-03

ExcellentGoodGalaxy S21 · 128 GBScreen burn-in on grey test imageJonas P.G-05

FairGoodMacBook Air M1Marked dust as scratches on the lidSanne D.G-02

FairGoodiPhone 13 mini · 128 GBReflection read as a scuffMark V.G-03

GoodExcellentPixel 7 · 128 GBCase marks wiped off before photosIlse B.G-01

GoodFairiPhone 11 · 64 GBChip in the back glass near the cameraTomasz W.G-06

GoodFairThinkPad T14 Gen 2Shiny keys that booth photos missJonas P.G-05

Reviewed photos improve the model for every partnerStart review

Fig. 7The calibration view, designed for the concept. Numbers are illustrative.

The label says what the buyer will see

The printed label and the Back Market listing use the same grade words and the same battery number, so what the technician confirms is what the buyer reads.