Method · 17 of 18
Where this could be wrong
Fifteen assumptions hold this console up. Five of them, if wrong, change a conclusion rather than a decimal place. They are listed first, because a number nobody can attack is a number nobody has checked.
Assumptions logged
15
Every input we do not measure
Material if wrong
7
Changes a conclusion, not a decimal
Most sensitive input
ARPD
Realisation per desk. Not disclosed anywhere.
Not yet back-tested
2
Intent prediction and forecast confidence
Assumption ledger
Sorted by materiality — high first
| Assumption | Value used | Where it comes from | If wrong by | Then what moves | Materiality |
|---|---|---|---|---|---|
| Realisation per desk per month | ₹17,900 | Assumed. Not disclosed by the company. | ±10% | The value of 100 bps moves by ±₹2.9 cr a year. Every rupee figure on this console moves with it. | Material |
| Connect rate on worked enquiries | 58% | DeployOne operating average across verticals, not measured on this book. | −15 pts | Tours booked fall by roughly a quarter and cost per tour rises to about ₹540. | Material |
| Voice-verified intent predicts outcome | Assumed directional | Not back-tested. No renewal cohort has run through the model yet. | Weak correlation | The renewal gap becomes a prompt to call rather than a forecast input. Screen 05 loses one of its five components. | Material |
| Retention against replacement cost | 1 : 4.6 | Assumed. Placeholder until your own acquisition cost is loaded. | 1 : 2 | The financial case for the renewal module weakens by more than half. The early-warning case does not. | Material |
| Verified closure sample extrapolation | Sample to centre | Callbacks run on a sample, then extrapolated. | Sample skews to complainants | The closure gap is overstated. This is the most likely direction of error on screen 13. | Material |
| Bureau coverage of small companies | Assumed partial | Commercial bureau data covers companies that have borrowed. Many small members have not. | Thinner than assumed | Risk bands understate risk for the smallest members. This is the most likely blind spot on screen 09. | Material |
| MCA filing lag | 6 to 9 months | Statutory filing timetable. A structural property of the source, not an estimate. | Longer for late filers | A company that failed in the last two quarters can still read clean. The register narrows uncertainty rather than removing it. | Material |
| City desk split | 8 cities, 133,600 desks | Constructed to sum to the disclosed total. The split itself is invented. | Materially | City rankings on screen 02 would change. Portfolio arithmetic would not. | Moderate |
| Expansion signal to filled desks | 0.6 realisation | Assumed. Stated need is not booked need. | 0.3 instead | The +2,310 desk component on screen 05 halves to about +1,150. | Moderate |
| Growth centre fill rate | 280 bps per month | Modelled from the shape of a typical flex ramp, not from your centre history. | 180 bps | Brigade Rubix reaches target in five months rather than three. Ramp drag on the portfolio increases. | Moderate |
| Attach conversion rates | 6% to 42% | Modelled from comparable nudge campaigns in hospitality, not from flex workspace. | Half | Screen 14 opportunity falls from ₹45.6 L to about ₹23 L a month. Still the fastest module to prove. | Moderate |
| Entity match accuracy | Assumed 94% | Name to CIN matching on the statutory register. Not yet measured on your enquiry file. | 85% | Roughly one in seven records would need a manual match. The register would still run, with a queue in front of it. | Moderate |
| Portfolio occupancy | 85.1% | Q1 FY27 disclosure, weighted across the city split used here. | ±30 bps | Empty desk count moves by roughly 400 desks. Immaterial to the argument. | Presentational |
| Square feet per person | 90 sqft | Industry convention, used only to convert lease-expiry square footage into a desk range. | 70 or 110 | Trigger desk estimates swing by roughly a quarter in either direction. Ranges are shown for this reason. | Presentational |
| Micro-market node positions | Schematic | Approximately correct relative to each other. Not survey data and not to scale. | n/a | Presentational only. No figure depends on it. | Presentational |
The one that would hurt most
Realisation per desk is assumed, and every rupee on this console rests on it. The company does not disclose it, so ₹17,900 is a back-solve from revenue and desk count, and it will be wrong at the centre level by a wide margin because a Prestige Central desk and an Electronic City desk are not the same product. The correct fix is not a better assumption, it is your number, loaded on day one of the pilot. Until then, treat every rupee figure here as an order of magnitude and every desk figure as the real claim.
Why this screen exists at all
A vendor dashboard that presents twenty confident figures and no error bars is asking to be believed rather than checked, and the first analyst question on a results call will find the softest one. Publishing the soft ones in advance costs nothing if the argument is sound, and it changes what the room is doing: instead of testing whether the numbers are real, they start arguing about which assumption to replace first. That is a considerably better conversation to be having.
What the pilot replaces
Six of the fifteen stop being assumptions in the first fortnight
Six of the fifteen stop being assumptions in the first fortnight.
Realisation, connect rate, cost per tour, city split, acquisition cost and attach conversion are all measured directly the moment the layer runs against a live file. The remaining nine need a renewal cohort and two quarters of outcomes, and they stay labelled as modelled until they have them. Nothing on this console will quietly graduate from modelled to measured without the back-test that earns it.