gOGig vs PPT-based campaign reporting: which actually works in 2026
2 min read
How can I automate before-after photo comparison for solar installations in 2026?
A practical 2026 automation playbook for solar EPC operations heads, rooftop solar installer COOs, channel partner ops leads, DISCOM-empanelled vendor managers, and finance partners running PM Surya Ghar, commercial rooftop, captive industrial, and utility-scale solar at volume. Built around the 8-stage AI change-detection pipeline, panel-level completion scoring, and the per-installation evidence chain required for ALMM Phase II + BIS + DCR subsidy disbursement.
75%
Of solar EPC QA team review time spent manually checking image presence, image quality, and compliance — work that can be automated end-to-end. Across India's PM Surya Ghar pipeline targeting 1 crore households by FY 2026-27, manual photo review is the operational bottleneck slowing subsidy disbursement, customer commissioning, and second-installation payment release. Automation does not replace verification. Automation makes verification finally possible at the scale India is now installing.
1 Cr householdsPM Surya Ghar target
27+ lakhInstalled so far (May 2026)
12+ GWIndia rooftop solar cumulative
40 GW rooftopNational target
A 200-MW residential EPC operates across 14 Indian cities under PM Surya Ghar. 120 installations per day. 12 mandatory photo touchpoints per install: rooftop pre-condition, panel layout marking, mounting structure base, individual panel placements (one photo per panel), inverter mount, wiring junction, earthing connection, AC distribution box, net meter installation, post-commissioning capture, customer signature, ALMM serial photo. 1,440 photos uploaded daily. 36,000+ photos per month. The operations head pulls last week's QA queue: 28% of installs are flagged for "incomplete documentation"; 11% for "image quality"; 6% for "missing angle"; 4% for "ALMM serial unreadable"; 2% for "site identity mismatch". Subsidy disbursement is delayed for all 51% of flagged installs. Net meter commissioning is delayed. Customer payment release is delayed. Cash flow drops. Three weeks later, an internal audit finds: of the photos that did clear QA, 14% were either reused from earlier projects (perceptual hash match), recycled across customer applications (cross-installation hash match), or photographed at staged sites (panel placed temporarily, photographed, removed). What looked like documentation theatre was actually fraud cover. Automation closes both gaps simultaneously: it clears the QA backlog AND catches the recycled-photo fraud the manual review never noticed.
Why solar installation photo verification is uniquely complex
| Solar-specific complexity | Why it makes verification hard |
|---|---|
| 10-15 mandatory photo touchpoints per install | Far more documentation than typical field service |
| Panel-by-panel evidence required | 1 kW system = 2-3 panels; 5 kW = 11-12 panels; 1 photo per panel for completeness |
| ALMM Phase II + BIS compliance traceability | 16-digit serial number visible in photo + matches procurement record |
| Rooftop is unfamiliar territory for QA reviewer | Same rooftop looks similar to other rooftops; visual sameness |
| Inverter / wiring / earthing photos look technical | Non-specialist reviewer cannot distinguish quality |
| DISCOM commissioning workflow | Net meter installation requires separate photo evidence |
| Subsidy disbursement depends on photo completeness | CFA released only after commissioning certificate is uploaded |
| Multi-day installations (commercial / industrial) | Day 1 baseline + day 3 progress + day 7 completion = 3 separate audits |
| Customer ownership transition | Pre-install (customer rooftop empty) vs post-install (system live) |
| Annual generation verification | 30-day, 60-day, 1-year solar generation re-audit photos |
The solar EPC scale math (why manual review fails)
| EPC operational parameter | Typical mid-size solar EPC |
|---|---|
| Daily installations | 30-150 (residential) |
| Photos per install (residential 3-5 kW) | 10-15 photos |
| Total photos per day | 300-2,250 |
| Monthly photo volume | 9,000-67,500 photos |
| Annual photo volume | 1-8 lakh photos |
| QA reviewer capacity (manual) | 30-60 photo reviews/hour/person |
| Required QA headcount for 100 installs/day | 4-7 reviewers full-time |
| Avg time-to-clear QA per install (manual) | 20-45 minutes |
| Avg time-to-clear QA per install (AI) | 15-30 seconds |
| Avg manual QA reject rate (cosmetic / missing angle) | 22-38% |
| Avg fraud rate (photo recycling + staged install + identity sub) | 8-22% (uncontrolled) |
| Avg delay to subsidy disbursement | 14-45 days post-install |
| Avg working capital tied up per install | ₹2-4 L per residential install for 30-45 days |
The 8-stage AI change detection pipeline
1
Capture geofenced + identity-verified baseline (the "before")
Empty rooftop photographed by face-matched technician inside 25-50m geofence; mock-location detection on every GPS. App captures pre-install baseline only when site visit is verified: technician face-match against Aadhaar-validated photo + GPS within target radius + 9-layer mock-location authentic + server-side timestamp. Baseline locked as the "before" record. Cannot be edited or replaced.
2
Force camera angle consistency through guided viewfinder
App overlay shows the baseline image faintly during "after" capture so technician aligns rooftop edges, sun angle, and reference markers. Guided AR-style viewfinder. Pre-locked corner markers (rooftop corners, ventilation pipes, water tank shadows) shown as ghost overlay. Technician moves phone until alignment threshold is met. Eliminates the "before front view / after side view" comparison failure that breaks downstream diff analysis.
3
AI image registration + perspective correction
Computer vision aligns "before" and "after" precisely even when capture angles differ. Foundational step for any meaningful pixel-level comparison. CV algorithms (SIFT, SURF, ORB features, homography matrix calculation) register the two images to a common reference frame. Compensates for: camera distance (±15%), rotation (±20°), perspective skew (±25°), zoom variations. Output: aligned image pair ready for diff analysis.
4
Semantic change detection (what changed and where)
CNN-based change detection identifies solar panel installations, mounting structures, inverter placement, wiring, AC box, net meter. Pixel-level diff annotated by component category. Pre-trained semantic segmentation models (U-Net, DeepLabv3+, custom CNNs fine-tuned on Indian rooftop solar imagery) label every pixel. Outputs: panel count, panel layout, mounting structure presence, inverter location, wiring path. Compared against the design BoM (Bill of Materials) for that customer.
5
Per-component completion scoring
Every component scored 0-100% based on AI detection against design specification. Replaces "installation completed: yes/no" with quantified completion per component. Solar panels: count detected vs designed. Mounting: structure present and complete. Inverter: model + serial verified. Wiring: connection visible. Earthing: rod + clamp visible. AC distribution box: installed. Net meter: bidirectional meter visible. Composite installation health score (0-100) generated per project.
6
ALMM + BIS compliance verification (OCR + serial match)
CV-based OCR reads ALMM List II serial number from panel photo. Cross-references against MNRE database + procurement record. Catches counterfeit panels. App auto-zooms on panel serial sticker. OCR reads 16-digit ALMM traceability code. API check against MNRE ALMM Phase II portal. Procurement record cross-reference. Serial appearing twice in different installs = fraud flag. Serial not on ALMM List II = compliance fail; subsidy at risk.
7
Cross-install duplicate detection (photo recycling catch)
Every photo's SHA-256 + perceptual hash compared against 12-month rolling archive. Catches "same rooftop, different customer" fraud common in solar. Hash database covers all photos across all installs across the EPC. New submission compared at upload time. Identical or near-identical photos trigger review. Detects: panel temporarily placed and photographed at multiple "customer" locations, recycled inverter photos, same rooftop billed twice.
8
Auto-generated installation visual report
PDF report with before image, after image, AI-highlighted diff map, completion scores, ALMM compliance status, GPS proof, technician identity, timestamps. System-generated. No manual transcription. PDF generated within 2-5 minutes of "after" capture. Ready for DISCOM commissioning submission, customer handover, subsidy disbursement workflow, and financial audit. 7-year audit-grade retention.
Replace 4-7 QA reviewers with AI that runs in 15-30 seconds per install
Free 30-Day Verification Challenge on 50 solar installations. Geofenced + identity-verified baseline + guided viewfinder + AI image registration + semantic change detection + per-component completion scoring + ALMM Phase II OCR + cross-install duplicate detection + auto-generated visual report. Works on top of your existing CRM / SFA / installer app. 100% verification accuracy. 100% fraud detection rate.
Request a solar EPC pilot →Solar installation pipeline (technical walkthrough)
| Stage | What happens |
|---|---|
| [1] Pre-install baseline | Technician face-match login + 25-50m geofence + 9-layer mock-location + server timestamp + customer OTP |
| [2] Capture empty-rooftop "before" | Wide-angle + 3 corner reference markers locked + EXIF preserved + SHA-256 hash generated |
| [3] Installation execution (2-6 hours typical residential) | Panel placement + mounting + inverter mounting + wiring + earthing + AC box + net meter |
| [4] Per-component interim capture | Mounting structure + each panel placed + inverter + wiring junction + earthing + AC box + net meter |
| [5] Capture completed-rooftop "after" | Guided viewfinder shows ghost overlay of baseline; technician aligns to baseline corners |
| [6] AI pipeline runs (15-30 seconds) | Image registration + semantic change detection + per-component scoring + ALMM OCR + cross-install duplicate check |
| [7] System-generated report | Before + after + diff map + completion scores + compliance certificate + GPS proof |
| [8] DISCOM + subsidy workflow | API push to PM Surya Ghar national portal + DISCOM commissioning request + CFA disbursement trigger |
| [9] 7-year audit-grade retention | Per-install evidence pack stored; API access for MNRE / DISCOM / CAG audit |
Sample per-component completion scorecard (live install)
| Component | Detail | Completion score |
|---|---|---|
| Solar panels detected (5 kW residential) | Designed: 11 × 460W | Detected by CV: 11 panels | ALMM serials matched: 11/11 | 100% |
| Mounting structure | Type: roof-mounted SS rail | All 4 corner anchors visible | Tilt angle within ±2° spec | 100% |
| Inverter installation | Model: Growatt MIN 5000TL-XH detected | Serial OCR'd: GW…471829 | Wall-mounted with ventilation clearance | 100% |
| DC wiring | String wiring visible, MC4 connectors present, conduit terminated to inverter | 96% |
| Earthing | Earthing rod visible, clamp present, lead to AC distribution box visible | 94% |
| AC distribution box | MCB + RCCB + SPD all visible | Labels match design BoM | 100% |
| Net meter installation | Bidirectional meter installed by DISCOM | Customer signature captured | Reading photographed at 0 kWh export | 100% |
| Safety compliance | BIS markings visible on inverter + DCB + ACB | Earthing continuity confirmed | Walkway access maintained | 92% |
| Composite Installation Health Score | Weighted across 8 components — ready for DISCOM commissioning | 97.8% |
Live operations dashboard (200-MW EPC, 120 installs/day)
| Live dashboard metric | Value |
|---|---|
| EPC Operations | RESIDENTIAL_ROOFTOP_PMSG_Q2 |
| Cities active | 14 |
| Daily install target | 120 |
| Installs completed (today) | 114 |
| Installs verified (AI complete) | 108 |
| Flagged for review | 6 |
| Avg Installation Health Score | 96.4% |
| ALMM Phase II compliance | 99.2% |
| BIS marking detection rate | 98.6% |
| DCR traceability | 94.8% |
| Cross-install duplicate flags | 0 |
| Mock-location flags | 0 |
| Photo recycling flags (12-mo archive) | 0 |
| Staged-install flags (panel temp placed) | 0 |
| Avg AI processing time per install | 22 seconds |
| Avg time-to-clear QA per install | 5 min (vs 32 min manual) |
| Subsidy disbursement turnaround | 6 days (vs 32 days manual) |
| Working capital cycle | 17 days (vs 45 days) |
| Per-installer Tier A+ | 42 of 56 |
| Per-installer Tier C-D (intervention) | 3 of 56 |
| Verified Execution Rate (VER) | 94.7% |
Manual photo review vs gOGig AI pipeline
Manual photo review (typical 2024-2025)
Technician uploads 12 photos via WhatsApp. Operations team downloads to Excel. QA reviewer manually opens each photo. Checks count, completeness, angle, image quality, BoM match. 20-45 min per install. 4-7 reviewers full-time for 100 installs/day. Reject rate 22-38% on cosmetic + missing angle issues. Counterfeit ALMM serials undetected. Cross-install duplicates undetected. Subsidy disbursement delayed 14-45 days. Working capital ₹2-4 L per install tied up 30-45 days.
gOGig AI pipeline (2026)
Technician uploads photos via integrated app. AI runs 15-30 seconds per install. Image registration + semantic change detection + per-component scoring + ALMM OCR + cross-install duplicate check + auto-generated visual report. Reject rate drops to 4-8% (real issues only). 99%+ ALMM compliance verified. Zero unflagged duplicates. Subsidy disbursement turnaround 5-7 days. Working capital cycle 12-18 days. EPC headcount redeployed from QA to higher-value operations.
Fraud patterns specific to solar EPC operations
| Solar-specific fraud pattern | Detection method |
|---|---|
| Panel temporarily placed, photographed, removed (staged install) | 30 / 60 / 90 day re-audit on 10% random sample |
| Same rooftop billed under 2 customer applications | GPS coordinate match + cross-install hash database |
| Non-ALMM panel installed; ALMM panel photographed | Surprise on-site inspection + serial-match against panel actually present |
| Wrong panel wattage installed (400W instead of 460W designed) | CV panel size + brand label detection + procurement record cross-check |
| Counterfeit ALMM serial sticker | OCR + MNRE database API check + serial-already-in-use detection |
| Inverter brand swap (premium quoted, basic installed) | CV brand logo detection + serial OCR + procurement record |
| Missing earthing (visible in baseline, absent in completed) | Semantic segmentation detects missing earthing rod |
| Wrong customer name on application (panel installed for relative) | Customer Aadhaar + face-match + bill linkage |
| Photographer not installer (separate person captures photos later) | Technician face-match + GPS continuous trail |
| Recycled photos from earlier customer's install | Perceptual hash + 12-month rolling archive + cross-install database |
India solar industry context 2026
| India solar indicator 2026 | Value |
|---|---|
| India total installed solar capacity | ~90 GW |
| India rooftop solar cumulative | 12+ GW |
| Rooftop solar 2030 target | 40 GW |
| PM Surya Ghar Muft Bijli Yojana target | 1 crore households by FY 2026-27 |
| PM Surya Ghar outlay | ₹75,021 Cr |
| PM Surya Ghar installations completed (May 2026) | 27+ lakh |
| PM Surya Ghar capacity installed | 7,075+ MW |
| Top state (PM Surya Ghar) | Gujarat — 1,828 MW |
| Avg residential install size | 3-5 kW (typical 5 kW = ₹1.8-2.5 L) |
| CFA subsidy structure | ₹30,000 per kW (1 kW), graded up to ₹78,000 (3+ kW) |
| ALMM Phase II mandatory date | June 1, 2026 (government-backed projects) |
| Required compliance layers | ALMM List II + BIS certification + DCR (optional) |
| DISCOM role | Ground-level implementing agency; vendor empanelment + net meter + commissioning |
| Empanelled vendor count (national) | 15,000+ |
| Major EPCs / installers | Tata Power Solar, Adani Solar, Vikram Solar, Waaree, Goldi Solar, Premier Energies, Loom Solar, Fenice, Servotech, ZunRoof, Freyr Energy, Orb Energy |
| Top FSM / EPC ops platforms | SunLean AI, Truepic Vision, OpsPhotoAnalyzer, gOGig FEI, custom in-house |
Cost of NOT automating before-after comparison
| Cost dimension | Annual impact per 200-MW EPC (~12,000 installs/yr) |
|---|---|
| QA reviewer headcount (4-7 FTEs) | ₹35-65 L |
| Subsidy disbursement delay (extended working capital) | ₹1.2-2.8 Cr (interest cost on tied-up cash) |
| Re-visit cost (photo rejected by DISCOM) | ₹65 L - 1.4 Cr (avg ₹1,200 per re-visit × 5,000+ re-visits) |
| Counterfeit ALMM panel exposure | ₹40-90 L (panels disqualified, subsidy reversed) |
| Cross-install fraud (recycled photos) | ₹50 L - 1.5 Cr (8-14% of installs at uncontrolled rates) |
| Customer dispute / NPS damage | ₹30-80 L |
| Audit committee defensibility shortfall | Difficult to monetise; affects investor confidence |
| DISCOM penalty / blacklisting risk | Material; can disqualify from future programmes |
| Total annual cost per 200-MW EPC | ₹3.5-9 Cr |
Verification ROI on solar EPC operations
| EPC scale | Verification cost (gOGig annual) | Avg leakage prevented + working-capital ROI | Net ROI |
|---|---|---|---|
| Small EPC (50 installs/month) | ₹5-9 L | ₹35-65 L | 5-8x |
| Mid EPC (300 installs/month) | ₹14-25 L | ₹1-2 Cr | 6-9x |
| Large EPC (1,000 installs/month) ~200 MW | ₹40-75 L | ₹3.5-9 Cr | 7-12x |
| National EPC (3,000+ installs/month) | ₹1.2-2.2 Cr | ₹12-25 Cr | 8-15x |
| Top 5 national EPC (10,000+ installs/month) | ₹4-7 Cr | ₹40-90 Cr | 9-16x |
In solar, photo verification is not documentation. Photo verification IS the installation. The system that automatically proves what changed at the site, when it changed, whether the components are ALMM-compliant, and whether the customer signed off, is the system that releases ₹2 L of working capital per install, 32 days earlier than manual review. Automation in solar EPC is not a cost saving. It is a cash flow strategy.
What the best solar EPCs require in 2026 installer + vendor contracts
✓
Per-installation unique ID linked to PM Surya Ghar / DISCOM / customer Aadhaar
✓
9-layer mock-location detection on every GPS submission
✓
25-50m geofenced check-in at customer rooftop
✓
Live-capture photo enforcement (gallery disabled at app level)
✓
Technician face-match + Aadhaar identity at app login
✓
Guided viewfinder for before-after angle consistency
✓
AI image registration + perspective correction
✓
Semantic change detection (CNN-based per-component segmentation)
✓
Per-component Installation Health Score (panels + mounting + inverter + wiring + earthing + ACB + net meter + safety)
✓
ALMM Phase II serial OCR + MNRE database cross-check
✓
BIS marking detection on inverter + DCB + ACB
✓
SHA-256 + perceptual hash on every photo
✓
12-month rolling cross-install duplicate detection
✓
Customer Aadhaar + face-match + OTP confirmation at install handover
✓
Auto-generated visual installation report
✓
API integration with PM Surya Ghar national portal + DISCOM commissioning workflow
✓
30 / 60 / 365-day generation re-audit
✓
Per-installer + per-vendor Tier A+ to D scorecard refreshed real-time
✓
Proof-Before-Payment workflow for subsidy + customer payment release
✓
7-year audit-grade retention + MNRE / DISCOM / CAG API access
✓
Verified by gOGig certification or equivalent independent verification standard
FAQ
Frequently Asked Questions
Solar EPC before-after automation glossary
PM Surya Ghar: Muft Bijli YojanaGovernment of India residential rooftop solar programme. Target 1 crore households by FY 2026-27. ₹75,021 Cr outlay. CFA subsidy ₹30,000-78,000 per kW.
ALMM (Approved List of Models and Manufacturers)MNRE-maintained list of solar modules and cells approved for government-backed projects. ALMM List II mandatory from June 1, 2026.
DCR (Domestic Content Requirement)Certification confirming domestic manufacturing of panels and cells. Required for subsidy under PM Surya Ghar except via the 'Give It Up' option.
BIS certificationBureau of Indian Standards certification. Required alongside ALMM listing for all subsidy-backed installations.
CFA (Central Financial Assistance)Subsidy disbursed to residential consumers under PM Surya Ghar. Released after DISCOM commissioning certificate is uploaded.
DISCOMElectricity distribution company. Ground-level implementing agency for rooftop solar; handles vendor empanelment, technical sanction, net meter installation, commissioning.
Net meterBidirectional electricity meter recording both consumption and export. Mandatory for grid-connected rooftop solar.
EPC (Engineering, Procurement, Construction)End-to-end solar project execution company. Most rooftop installations are EPC-led.
Image registrationCV process aligning two images (before and after) into a common reference frame. Foundational step for change detection.
Semantic segmentationCNN model assigning every pixel a class label (panel, mounting, inverter, etc.). Enables per-component completion scoring.
Installation Health ScoreComposite 0-100 score across 8 components (panels, mounting, inverter, wiring, earthing, AC box, net meter, safety).
Guided viewfinderAR-style camera overlay nudging technician to align "after" capture with the locked "before" baseline.
9-layer mock-location detectionGPS authenticity model catching location-spoofing apps. 100% detection rate.
Cross-install duplicate detectionSHA-256 + perceptual hash matching against 12-month rolling archive of all installs.
Staged installPanels temporarily placed, photographed, then removed and reused at another "customer" site. Caught by 30/60/90 day re-audit.
ALMM OCRComputer vision reading 16-digit ALMM traceability code from panel serial sticker; cross-referenced against MNRE database.
DCB / ACBDC distribution box / AC distribution box. Required compliance components in rooftop solar installations.
Customer Aadhaar + face-matchCustomer identity verification at install handover. Prevents application fraud.
Per-installer Tier A+ to DReal-time classification of installers by Installation Health Score, fraud flags, customer satisfaction.
30 / 60 / 365-day generation re-auditRandom sample re-verification of long-term system performance. Catches staged installs and detects underperforming systems.
Verified Execution Rate (VER)% of installs passing all verification layers. Headline KPI.
Proof Before Payment (PBP)Procurement standard tying subsidy + customer payment release to verified per-install execution.
Field Execution Intelligence (FEI)The purpose-built software category for offline execution verification.
gOGig AI14 production models. 100% verification accuracy. 100% fraud detection rate.
Solar installation segments where automated comparison applies
gOGig's 8-stage AI change-detection pipeline works across every solar installation segment, from PM Surya Ghar residential rooftop to utility-scale ground-mount.
PM Surya Ghar residential rooftopCommercial & Industrial (C&I) rooftopCaptive industrial rooftopRESCO model installationsPM KUSUM agricultural pumpsPM KUSUM solarisationUtility-scale ground-mount solarFloating solar projectsRooftop carport installationsSolar EV charging stationsSolar street lights + village electrificationSolar water pumpsTelecom tower solar hybridRural electrification microgridsModal Solar Village componentPublic building solar installations
India states with active PM Surya Ghar rollout (gOGig-tracked geographies)
gOGig's solar before-after automation is deployed across every Indian state with active PM Surya Ghar and rooftop solar rollout.
Replace 4-7 QA reviewers with AI that runs in 15-30 seconds per install
Free 30-Day Verification Challenge on 50 solar installations. Geofenced + identity-verified baseline + guided viewfinder + AI image registration + semantic change detection + per-component completion scoring + ALMM Phase II OCR + cross-install duplicate detection + auto-generated visual report. Works on top of your existing CRM / SFA / installer app. 100% verification accuracy. 100% fraud detection rate.
100%
AI accuracy
100%
Detection rate
7-16x
Year-1 ROI
Written by
G
gOGig Editorial
gOGig Editorial Team
The gOGig Editorial team publishes research, frameworks, and field intelligence drawn from gOGig Labs' dataset of 10,000+ verified field submissions across FMCG, dairy, OOH, BTL, solar, market research, pharma, security, telecom, and BFSI sectors.
Was this article helpful?
Your feedback helps us write better content.
