The Outside-View Forecast: Reference-Class Modeling for a New Market Launch

Marketing Academy · Field Work●Mental Models

All projects·From: Base Rates and Reference-Class Forecasting

CoreForecast· 45 minutes

Ola Electric

Objective: Construct an outside-view forecast model in Google Sheets for a major regional market expansion, establishing the reference class distribution, calculating base rate quartiles, and constraining inside-view optimism to justifiable adjustments.

You are a senior marketing analyst at Ola Electric tasked with forecasting 90-day test-ride booking volumes and unit deliveries for an expansion into 25 Tier-2/Tier-3 cities across India. The regional sales director built an inside-view model projecting 12,000 bookings per city based on local showroom hype. You must build a reference-class forecast based on the actual distribution of Ola Electric's previous 80 city rollouts.

Using the lesson's 4-step playbook, establish the reference class of historical city rollouts, calculate the base rate (median, 25th percentile, 75th percentile), test whether the inside-view claims survive empirical benchmarks, and produce a calibrated range forecast.

Before you start

What you'll need

Free path (everything below is enough to finish)

Google Sheets

FreeForecasting model workbook containing Reference Class data, Base Rate calculations, and Adjusted Forecast

Standard spreadsheet environment for statistical distribution modeling and scenario analysis.

No access? Excel or any spreadsheet editor

The process

3 steps

Step 01 of 03

The Playbook: Four Steps

Step 1 of reference-class forecasting requires selecting a reference class that is broad enough to provide statistical significance (at least 10–20 comparable cases) and narrow enough to share structural characteristics.

What criteria define the reference class of comparable city rollouts, and what is the distribution of historical 90-day booking volumes?

Google Sheets— Google Sheets / Tab: Reference Class Data

Procedure

Filter the historical city rollout dataset for Tier-2/Tier-3 demographic and charging infrastructure parity (minimum 15 comparable launches).
Record 90-day booking volume, showroom footfall, and lead-to-delivery conversion for each city in the class.
Calculate the 10th percentile, 25th percentile, median (50th), 75th percentile, and 90th percentile of the reference class using PERCENTILE.INC.
Compare the internal sales forecast of 12,000 against the historical distribution.

Sample output

REFERENCE CLASS DISTRIBUTION (Historical Tier-2/3 City Rollouts, n=24):
- Minimum (P0): 1,850 bookings
- 25th Percentile (P25): 3,200 bookings
- Median (P50 Base Rate): 4,650 bookings
- 75th Percentile (P75): 6,400 bookings
- 90th Percentile (P90): 7,900 bookings
- Maximum (P100): 8,850 bookings
- Inside-View Proposal: 12,000 bookings (Exceeds P100 historical record by +35%)

Healthy

Reference class contains 15+ comparable historical launches with calculated distribution percentiles (P25, P50, P75).

Unhealthy

Forecast relies on a single 'best case' launch (reference class of one) or unweighted averages distorted by extreme outliers.

What this means

If the proposed inside-view projection sits above the 90th percentile of historical rollouts, it requires extraordinary empirical justification rather than narrative enthusiasm.

So what do I do about it?

Symptom	Action	Effort
Proposed target exceeds the 90th percentile of the reference class	Anchor the baseline budget to the 50th percentile (median) and require proven leading indicators before unlocking Tier-3 growth budget	30 min

YouYou can do this yourself, no engineering access required.

Step 02 of 03

What It Actually Is

The outside view replaces bottom-up storytelling with the actual historical hit rate and conversion benchmarks across similar cohorts.

What is the empirical base rate for top-of-funnel conversion from digital lead to showroom test ride in comparable expansion markets?

Google Sheets— Google Sheets / Tab: Conversion Base Rates

Procedure

Calculate historical lead-to-test-ride conversion rate across the reference class: Median = 4.2%, Interquartile Range = 3.1% – 5.8%.
Calculate test-ride-to-booking conversion rate: Median = 18.5%, Interquartile Range = 14.0% – 22.0%.
Multiply the baseline lead volume by the median conversion base rates to establish the unadjusted outside-view delivery forecast.
Identify the discrepancy between the unadjusted outside-view model and the inside-view target.

Sample output

| Funnel Stage | Inside-View Assumption | Base Rate Median (P50) | Base Rate IQR (P25–P75) | Discrepancy |
|---|---|---|---|---|
| Digital Lead -> Test Ride | 8.5% | 4.2% | 3.1% – 5.8% | +102% over baseline |
| Test Ride -> Paid Booking | 32.0% | 18.5% | 14.0% – 22.0% | +73% over baseline |
| End-to-End Funnel Yield | 2.72% | 0.78% | 0.43% – 1.28% | 3.5x higher than historical median |

Healthy

Conversion assumptions are explicitly bounded by the historical P25–P75 interquartile range.

Unhealthy

Model assumes compounding top-decile performance at every stage of the funnel simultaneously.

What this means

Assuming 90th percentile performance across three consecutive funnel stages implies a 1-in-1,000 event (0.10^3 = 0.001 probability).

So what do I do about it?

Symptom	Action	Effort
Inside-view model assumes 8.5% lead-to-ride conversion (2x historical median)	Reset planning baseline to 4.2% median and build sensitivity tables for 3.1% (downside) and 5.8% (upside)	30 min

YouYou can do this yourself, no engineering access required.

Step 03 of 03

The Playbook: Four Steps

Step 3 permits adjustments to the base rate only for documented, measurable structural advantages, capping subjective optimism.

Which specific differentiators justify an upward or downward adjustment from the base rate median, and by what quantified percentage?

Google Sheets— Google Sheets / Tab: Calibrated Forecast

Procedure

Audit each claimed advantage from the inside-view pitch (brand awareness, charging network density, state subsidies).
Reject non-measurable narrative claims ('higher team excitement', 'better creative vibe').
Quantify valid structural differences: +15% for pre-established Hypercharger network density vs class average; -10% for lower state EV subsidy in target region; +5% for local festive timing.
Compute net adjusted forecast (+10% over median base rate = 5,115 bookings) and establish an 80% confidence interval (P10 to P90: 2,600 to 7,200 bookings).

Sample output

| Adjustment Factor | Inside-View Claim | Verified Evidence | Justified Modifier | Calibrated Impact |
|---|---|---|---|---|
| Charging Infrastructure | 'Every buyer has charging' | 3 Hyperchargers per city (vs 1.8 avg) | +15% | +698 bookings |
| Regional Subsidy Headwind | Ignored | State EV subsidy 40% lower than Tier-1 | -10% | -465 bookings |
| Brand Search Index | 'Everyone knows us' | Search volume index matches class avg | 0% (No adjustment) | 0 bookings |
| Festive Season Timing | 'Huge holiday boost' | Historic Q4 lift across past launches | +5% | +233 bookings |
| NET ADJUSTMENT | +158% (Story-based) | Quantified structural factors | +10% over Base Rate | 5,115 bookings (80% CI: 2,600 – 7,200) |

Healthy

Adjustments are backed by quantified structural differences with explicit downward modifiers where headwinds exist.

Unhealthy

Every single adjustment is positive with zero acknowledgment of regional friction or operational bottlenecks.

What this means

A disciplined reference-class forecast typically adjusts the base rate by no more than ±20% to ±30% unless a foundational business model shift is proven.

So what do I do about it?

Symptom	Action	Effort
Total net adjustment exceeds +50% over the base rate	Cap initial manufacturing and showroom inventory at the +20% threshold with a 30-day flex replenishment contract	half day

YouYou can do this yourself, no engineering access required.

Final deliverable

A complete 3-tab Reference-Class Forecasting Model in Google Sheets with a historical reference class distribution, conversion base rates, calibrated adjustments, and an 80% confidence interval.

See a reference example

Sample output

Worked example for Freshworks (B2B SaaS product add-on launch):
The product team forecasted 1,500 enterprise add-on purchases in Q1 ($3.6M ARR) based on bottom-up TAM estimates. Rebuilding with Reference-Class Forecasting:
1. Reference class: 12 previous product module launches across Freshdesk and Freshservice between 2018–2023.
2. Base rate: Median 90-day attach rate was 3.4% of existing customer base (IQR: 2.1%–4.6%), translating to 410 accounts ($984k ARR).
3. Calibrated adjustments: +20% for automated in-app upsell prompts (tested in pilot), +10% for sales incentive spiff, -15% for higher compliance review cycle in enterprise tier. Net adjustment: +15% over median.
4. Final calibrated forecast: 472 accounts ($1.13M ARR) with an 80% confidence band of 320 to 650 accounts. Actual day-90 result landed at 495 accounts, validating the base-rate anchor.

Success criteria

You're done when you can:

Defines a statistically sound reference class with at least 10 comparable historical initiatives
Calculates accurate distribution percentiles (P25, Median, P75) rather than relying on a simple arithmetic mean
Audits and filters subjective inside-view claims, applying adjustments only to verified structural differences
Produces an 80% confidence range (P10 to P90) rather than a single point prediction