← All insights
Guide · 8 min · August 28, 2026

Why realistic KPIs make or break your growth model

A growth model is only as useful as the assumptions inside it. And the fastest way to make a model worthless is to fill it with KPIs no business has ever actually hit — “10× user growth every month, forever,” or “CAC payback in one month with no marketing spend.”

Numbers like these feel ambitious. In practice they quietly destroy the model.

What unrealistic KPIs actually cost you

The damage isn’t just an embarrassing forecast. Targets that sit far outside observed reality do four specific things:

  • They kill credibility. Investors, boards, and even your own team distrust a projection the moment the numbers stop looking possible.
  • They misdirect spending. A fantasy top line leads to over-hiring, over-spending, or starving the parts of the business that actually needed the money.
  • They hide the real levers. When the headline number is fiction, it masks the metrics that genuinely compound — retention, conversion, unit economics.
  • They remove your baseline. When reality diverges from the plan, you have nothing believable to diagnose against. You can’t tell why you missed, because you were never going to hit it.

Realistic KPIs do the opposite. Grounded in what comparable companies actually achieve, they force you to confront trade-offs, set milestones you can reach, and focus on the metrics that build on themselves over time.

Start from the growth equation, not the metrics

Before you pick a single KPI, write down the core equation of your business — the handful of input levers you can actually influence:

  • Marketplace: Supply × Demand × Match rate × Monetization
  • SaaS: New customers + Expansion − Churn
  • Consumer app: Acquisition × Activation × Retention × Monetization
  • Ecommerce: Traffic × Conversion × AOV × Repeat purchase rate

Everything else in the model should trace back to these. If a metric doesn’t feed the equation, it’s probably a vanity number.

Choose 5–8 KPIs — no more

The temptation is to track everything. Resist it. A model with fifty metrics has no metrics. Limit yourself to a small set of leading and lagging indicators that matter for your motion:

  • Acquisition: CAC, channel mix, stage-by-stage conversion
  • Activation: share of new users who reach the “aha” moment within X days
  • Retention: D1 / D7 / D30, cohort retention curves, DAU/MAU
  • Monetization: ARPU, conversion to paid, expansion revenue
  • Unit economics: LTV, LTV:CAC, payback period, contribution margin

Ground every assumption in a benchmark

This is the step most models skip — and it’s the one that separates a decision tool from a hopeful guess. For each key assumption, find the range that comparable companies actually see, and stay inside it unless you have a data-backed reason not to.

Useful starting points depend on your sector: OpenView, Bessemer, and ChartMogul for SaaS retention and payback; AppsFlyer and Sensor Tower for consumer retention curves and CPI; Shopify-style reports for ecommerce conversion and repeat rates; and a16z, YC, and First Round for stage-specific startup metrics.

One caveat worth repeating: always match the benchmark to your stage, vertical, and go-to-market motion. A Series B B2B SaaS company and a pre-product consumer app live in completely different number ranges.

Build it in cohorts, and stress-test it

A few structural choices keep the model honest:

  • Track cohorts by acquisition month, so retention and LTV reflect reality rather than an average that hides the truth.
  • Model three scenarios — base, upside, downside — but vary only a few key assumptions, not every cell.
  • Separate inputs from outputs clearly: what you assume vs. what the model calculates.
  • Add a unit-economics layer. Growth that burns cash faster than it creates value isn’t growth.

Then pressure-test it. What happens if retention comes in 20% worse, or CAC rises 30%? If a small, plausible change breaks the whole plan, that’s not a reason to ignore it — it’s the most important thing the model just told you.

The realism rules of thumb

A few habits keep a model believable:

  • Start from what you already observe — current retention, current CAC by channel — then apply modest improvement, not miracles.
  • Cap your growth rates. Very few companies sustain 15–20%+ month-over-month growth for long once they’re past the earliest stage.
  • Model churn and expansion explicitly. Early models often ignore one or overstate the other.
  • Document every major assumption and its source — the benchmark report and date, or the internal experiment behind it.

The takeaway

When you present a model, lead with the realistic KPIs and the benchmarks that justify them — not the hockey-stick headline. Operators and investors care far more about whether your retention curves and unit economics are believable than about how big the top-line number gets.

That’s the whole point of modeling growth: not to predict a perfect future, but to build one you could actually defend.