Climate tech needs better products.

Hi! I'm Shubha, a Product Manager who bridges the gap between vision and execution. I've got a bone to pick with the climate crisis, and that's where I put the work, using data-driven user research, strategic roadmapping, and hands-on delivery to build products that scale.

View case studies Get in touch →
Who this is for
Case studies

Real problems, working prototypes.

Case studies so far, with more industries in the pipeline.

Energy Access · P2P · Research next

Barrio-Grid

Coming soon
Carbon Accounting · B2B · Research next

Refractor

Coming soon
Climate Adaptation · B2G · Research next

Firstwatch

Coming soon
Playground

A hands-on PM toolkit.

5 micro-tools you can use right now to make better product decisions, with logic and benchmarks pulled directly from Lenny's newsletter and podcast archive.

Defaults and sample data below are set for an early-stage team finding its first users.

A product-led growth (PLG) vs. enterprise sales motion trade-off calculator, grounded in growth benchmarks.

$1,000$100,000
Selected: $15,000
Sales Cycle Length
Target Persona

Score = ACV weight (0-40, scaled across $1K-$100K) + deal-cycle weight (0-30) + buyer-seniority weight (0-30), out of 100. Very low ACV with a self-serve buyer, or high ACV with a long cycle, overrides that raw score to force Pure PLG or a hybrid motion.

Recommended motion

 

Metrics to watch

Sources

A dynamic RICE feature matrix that calculates strategic blindspot penalties and retention risk.

Feature Reachusers/quarter Impact0.25–3 Confidence% Effortperson-months High Tech
Debt Risk		 Missing
Retention Loop		 RICE Score Remove
RICE Score = (Reach × Impact × Confidence) ÷ Effort, then reduced by whatever severity you pick per risk flag (mild/moderate/severe). Stress-test each row DRICE-style: dollarize the impact and estimate engineering days to get a real ROI-per-eng-week.
Sources

Both risk flags and their mild/moderate/severe penalty tiers are my own addition for this demo.

These are static, hand-written examples. Pick a feature idea to see it broken down into edge cases, a metric tree, and engineering-ready acceptance criteria.

Pick a feature idea above to see a worked example of the full spec.

Sources

Edge cases and the metric tree are standard PM/eng practice I wrote myself, separate from the sourced material above.

Plan a real A/B test: how many visitors you need and how long to run it, based on your baseline conversion rate and the lift you'd actually care about. This doesn't predict which copy wins, no tool can do that from text alone, it tells you what a valid test of your own copy would require.

Sample size per variant = (1.96 + 0.84)² × [p1×(1−p1) + p2×(1−p2)] ÷ (p1−p2)², where p1 is your current rate and p2 is p1 plus the lift you're testing for.

Sources

Assumes the standard 95% confidence / 80% power used by most experimentation platforms. This tells you what a valid test of your own copy requires, it can't tell you which variant will actually win.

How long it takes to earn back what you spent acquiring a customer, benchmarked against a survey of 16 growth operators. A shorter payback period isn't automatically better, see the note below, but it's the number that tells you how fast you can reinvest in growth.

Payback period (months) = CAC ÷ (monthly revenue × gross margin).

Sources

These benchmarks reflect one survey's consensus among 16 growth operators. Business stage and business model both change what a good target looks like, see the note below the result.