SASA Intelligence Lab / Overview
Andhra Pradesh sanitation overview
Explore reported progress, compare places and inspect the evidence.
Rural collection changeCommon GP-day cohort · May → August
A modest statewide shift, and 5 districts that decline at every step.
85,769 panchayat-days carrying a valid measurement in every one of the four retained months.
Matched GP-day cohort · May–August 2026 · all seven days counted-1.29percentage points, from 74.71% in May to 73.42% in August
Evidence coverage & scoring limits
- Panchayat-days read
- 12,85,814
- Historical exports
- 44
- Authorized routes
- 42
- Overall scoring
- Not assigned
Historical exports and authorized routes count different things; these figures do not measure current ingestion completeness. Subject comparisons retain their own evidence rules.
Inspect source coverageHow the day selection affects these rates
Each seven-day window contains one Sunday, and those Sundays report 1.85%–3.03% across the four months. They are recurring low-reporting days, and the cause is not confirmed. Every month carries exactly one, so the periods stay comparable, but the level sits below the rate on the other days.
All matched districts · May–August 202685,769 matched GP-day observations · first seven days · all seven days
Yes ÷ valid Yes/No collection reports, all seven days. Full 0–100% scale; the same GP and day-of-month observations are used in every month.
Locate the evidence
May → August change by district
Loading district boundaries…
−21.821.8 percentage points
Choose a district28 districts
Selection focuses the accompanying figures or review list. Hatching indicates missing or unmatched evidence.
2022 boundary reference
2022 district boundaries provide a location reference. Matching names do not certify that current administrative extents are unchanged. New, ambiguous or unmatched districts remain in the selector; no value is assigned to a guessed polygon. Colour uses a linear scale.
85,769 common GP-day observations5 districts decrease at every step2 increase at every step4 validated retained months
Follow the district trajectories
Four reported first-week rates · May, June, July, August
| District | May | Jun | Jul | Aug | Changepercentage points | Pattern |
|---|---|---|---|---|---|---|
| 76.58% | 69.74% | 64.40% | 54.82% | -21.77 | ↓ Every step | |
| 84.30% | 79.57% | 75.40% | 65.95% | -18.35 | ↓ Every step | |
| 58.33% | 61.86% | 71.01% | 76.25% | +17.92 | ↑ Every step | |
| 60.89% | 64.15% | 64.22% | 70.02% | +9.13 | ↑ Every step | |
| 81.19% | 79.33% | 77.04% | 74.37% | -6.82 | ↓ Every step | |
| 70.91% | 69.27% | 66.36% | 64.34% | -6.57 | ↓ Every step | |
| 78.63% | 73.85% | 72.59% | 72.26% | -6.37 | ↓ Every step |
These are reported first-week trends, not full-month service ratings or established programme effects. Day-of-month pairing holds the weekday mix roughly constant rather than eliminating it, and the basis without Sundays and second Saturdays excludes them by calendar rather than by any published holiday list. Stored zero/No responses may reflect reporting behaviour.
Inspect these districts in Gap Radar ↗Common cohort, source validation & full-period context
A panchayat-day is in the cohort only where every month in the series carries a valid collection measurement for it. days 1-7 of every month, set by the shortest retained month. Full-period rates use different denominators and are not used in this comparison.
2026-05: 12,41,643 raw rows → 3,99,270 distinct GP-days; 31 days retained; 96.47% observed-grid coverage. Full-period context: 70.28%. Conflicting keys: 0.
2026-06: 12,01,590 raw rows → 3,93,279 distinct GP-days; 30 days retained; 98.19% observed-grid coverage. Full-period context: 70.25%. Conflicting keys: 0.
2026-07: 12,41,643 raw rows → 4,01,838 distinct GP-days; 31 days retained; 97.09% observed-grid coverage. Full-period context: 73.70%. Conflicting keys: 0.
2026-08: 2,80,371 raw rows → 91,427 distinct GP-days; 7 days retained; 97.83% observed-grid coverage. Full-period context: 73.30%. Conflicting keys: 0.
Cross-subject signal
Overall rank gatedToo little overlap for an overall score.
49 ULBs carry the same name and district in all three July delivery measures, but only 19 can be rated in all three, and 18 of those report zero for both household toilets and vehicles. That leaves 1 with a nonzero toilet or vehicle rate, so the measures stay side by side and are never ranked as one score.
Separate June → July movementsVehicles suppliedJuly 2026 is identical to June 2026 in every field for 82 of 82 ULBsIHHLs completedJuly 2026 is identical to June 2026 in every field for 123 of 123 ULBsLegacy waste balance27 lower · 0 higher · 95 unchanged
Ready for descriptive review
Evidence investigationsOPEN GAP RADARInspect changes, cohorts & works →44 historical exports · 42 current catalogue routes · 6,509 raw retained rows
123 observed ULB-name candidates provide a working reference. 6,173 bundled rows after excluding repeated endpoint copies; internal row duplicates are a separate check.Source quality findings and revision checks
Found by review, not declared by the source
What the evidence does to itself.
44 retained datasets · 6,509 rows. These conditions change how every figure above should be read.
6district labels
The supplied identity mapping contradicts itself
Six retained datasets now carry LGD codes beside their departmental labels. 6 labels map to two different LGD districts inside a single source, and 82 entities carry codes two sources disagree about. The mapping reaches 20% of observed entities and is not adopted as identity.
Inspect the mapping →1endpoint
Endpoints serving rows another endpoint already returned
soak_pits_api repeats compost_pits_api. 336 rows would be counted twice if each key were treated as its own programme. Detected by comparing content, not names.
See the excluded copy →117of 117 ULBs agree
One programme, two independent returns, no disagreement
Identification of new IHHLs is returned by both the August source and its September LGD reissue. Every one of the 117 ULBs present in both reports identical approved, identified, under-construction and completed figures. 6 candidates appear in only one vintage and are not compared. Agreement means the September reissue did not alter the measures. It is not evidence that either figure is correct: both vintages could repeat one upstream error.
Open the sanitation cohort →Supporting evidenceProcessing facilities & source statusesHistorical context / 2024Swachh outcomes & reported ranks
Methods & additional retained history Source rules and community programmes
For the urban review, 123 ULB candidates have the fields needed for this issue. Missing or conflicting measurements are held out; genuine zeros stay in the cohort. Shortfalls are calculated per ULB before summing, so excess delivery in one ULB never cancels another ULB’s shortfall.
The map and ranking use the same eligible records. Concentration means the top five’s share of the selected source’s reported shortfall. Map matching does not approve a cross-source identity.
Reported Operations Monitor
Five retained reporting periods · one column per reported value · hover any column for its source figures
REAL RETAINED HISTORY
0.5% to 23.4%Reported coverage across 5 months and 3 programmes. All 3 reach their lowest point in April, which describes how that month was reported rather than three separate collapses. 849 records state a percentage that disagrees with their own target and achievement; those are held out.