Video summary
Setting KPIs and Goals | Startup School
Main summary
Key takeaways
Goal of the talk
Help early-stage founders set KPIs and use them for prioritization to reach product-market fit faster—avoiding “busy work” and vanity metrics that don’t move the business.
Core definitions and why they matter
- KPI (Key Performance Indicator): Metrics you track internally/externally to determine if what you’re doing is working.
- Prioritization: Ordering work by impact—given finite time, focus on the few items that move your top KPIs toward product-market fit.
- Key warning: Don’t optimize for vanity metrics (metrics that feel good / are brag-worthy but don’t drive revenue or growth).
Framework: KPI → Bottleneck → Task prioritization (execution loop)
- Identify your top KPI(s)
- Post-launch: Revenue growth is usually the primary KPI.
- Pre-launch: Use short-term proxies like:
- # of user conversations
- # of conversations until launch
- Set a KPI goal for the week
- Example: “10 more paying customers by next week.”
- Weekly goals create urgency because early growth compounds.
- Example playbook (YC founders): weekly KPI goals were made highly visible (e.g., Airbnb founders used a bathroom mirror).
- Find the biggest bottleneck blocking that KPI
- Example: Super Daily (India grocery subscription; sold to Swiggy in 2018)
- After launching V1, growth was bottlenecked by conversion → churn of high-intent users.
- Root cause: users wanted a specific milk brand that wasn’t carried.
- Action: onboarded the milk brand.
- Result: +50% increase in conversion rate (by fixing the highest-impact drop-off).
- Example: Super Daily (India grocery subscription; sold to Swiggy in 2018)
- Rank tasks and test
- Process:
- Write ideas
- Rank by probability of success
- Sub-rank by complexity/time
- Pick a couple to try
- If the KPI isn’t moving: ask “why” multiple times until you find the real cause.
- Operationalize learning:
- Run weekly retros: Did you do the tasks you planned? If not, break them down, reduce context switching, and improve sprint execution.
- If experiments fail: don’t repeat the same thing—fail fast and switch.
- If growth isn’t happening: talk to users fast to recycle toward better ideas.
- Process:
“Do” vs “Don’t” list for early-stage task selection
Should be on the list (usually):
- Talk to users and iterate based on feedback (directly supports revenue growth)
- Build/iterate on what users actually need
Should generally NOT be primary focus (examples of “fake progress”):
- Passive fundraising / coffees / meetings without active fundraising
- Conference attendance (except a few targeted industries)
- Arbitrary technical milestones (e.g., launching an Android app) unless users say it’s a burning pain
- Perfectionism / over-optimizing features users aren’t using
- Paperwork/administrative optimization (incorporation/equity structure) beyond standard compliance
- “Cool hard features” that haven’t been validated by user needs
Mental traps to avoid (execution risks)
- Low-leverage busywork: check-box tasks that don’t move product-market fit
- Self-deception about what’s working: mistaking slow growth for true PMF
- Perfectionism/indecision: slowing down because decisions feel high-stakes—make “pretty good” decisions quickly, then iterate
- Downside protection over upside exploration: too much fixing small problems vs chasing upside via creative risk-taking
- Chipping away at small problems while a large existential issue is present
- Example: 150 users requesting a small feature while churn/signups are failing
Choosing the right KPIs: primary vs secondary
Primary KPI (most startups)
- Usually Revenue growth
- Exceptions:
- Marketplace: may use signups or GMV as primary
- Early Enterprise with long cycles: may use letters of intent (LoIs)
Secondary KPIs (keep small: ~3–5)
- Purpose: avoid “cheating” the primary KPI and provide early signals.
- Examples:
- Retention / churn
- Unit economics (are you making money per user?)
- CAC / payback period (especially early; often more reliable than complex LTV)
Vanity metrics (when they creep in)
- If a metric isn’t directly on the path to revenue growth—or isn’t the biggest bottleneck—it’s likely vanity. Cut it.
Growth targets / KPI goal setting (metrics & timelines)
- YC-associated reference:
- 5–7% week-over-week growth = good
- ~10% week-over-week growth = exceptional
- Targets should be compounded weekly, especially early.
- Growth rate factors:
- Latent demand (can boost early growth, but may be harder later)
- Sales cycle length (Enterprise may require different process metrics)
- Organic vs paid acquisition:
- Prefer organic early
- Run paid acquisition tests later once payback period is understood
Retention vs acquisition (how to balance)
- You must do both, but first ensure churn isn’t destroying payback.
- Focus on:
- Bringing new users into a product that can sustain them long enough to pay back CAC
- Ensuring users can share the product (word-of-mouth)
- Case example (Super Daily metric change):
- Switched from tracking signups (too gameable via promos) to tracking customers who placed 5+ orders
- Why: a leading indicator of long-term revenue users; better aligned with top-line goals.
Two KPI target-setting approaches (playbooks)
- Top-down goals
- Start from a future milestone, then work backward to weekly growth needs
- Example: $5,000 MRR by end of Startup School → convert into weekly growth required
- Bottoms-up goals
- Determine what’s realistic next week given current constraints
- Thought experiment: “What could we do with unlimited funding/resources?” then adapt with limited resources
- Recommendation: Use both periodically to avoid “underwhelming but consistent growth” (the “no man’s land” risk).
Non-revenue KPI guidance (what to use now vs later)
- CAC → LTV ratios
- Mostly a concern post–product-market fit when scaling reliably
- Early-stage guidance: focus on payback period
- Aim for ~0 CAC spend so customers are profitable on day one
- Free signups / DAU
- Paying users behave differently; free feedback may mislead
- Exception: marketplaces/network effects where volume is required for utility (e.g., Uber—drivers needed for rider experience)
- Exceptions where revenue is hard to measure early
- Hardware, biotech, Enterprise long sales cycles:
- Use LoIs, contracts, or select process/technical indicators only if they truly predict growth—audit frequently.
- Hardware, biotech, Enterprise long sales cycles:
Concrete example: Scribd (timing of charging)
- Scribd grew ~4 years with a free product, afraid of losing users.
- After switching to charging:
- Lost >90% of customers but revenue grew dramatically (“Infinity percent” in subtitles).
- Takeaway: charging later can delay learning about paid user needs; revenue metrics can be necessary to understand real product value.
Actionable “homework” / requested founder outputs
- Write down:
- Your primary and secondary KPIs
- Ambitious targets
- Then:
- Audit your task list for the week
- Ensure tasks are laser-focused on moving those KPI targets
Presenter / sources
- Presenter: Divya (YC visiting group partner; two-time YC founder)
- Named external references/examples:
- Y Combinator (YC)
- Paul Graham (growth-rate guidance; “growth” essay)
- Airbnb (weekly KPI mirror anecdote; growth-rate context)
- DoorDash and Rickshaw (order volume comparison; Rickshaw acquired by DoorDash)
- Super Daily (India grocery subscription case)
- Swiggy (referenced via acquisition)
- Scribd (charging-timing case)
- Uber (network effects exception)