Why TTV Averages Lie: The Case for Percentiles (P50, P90)

If you walk into a neighborhood dive bar, the average net worth of the patrons might be $50,000. If Bill Gates walks into that same bar, the "average" net worth suddenly jumps to $100 million.

Does that mean everyone in the bar is suddenly a multi-millionaire? No. It means the Average (Mean) is a liar. It is easily corrupted by a single outlier.

Time Data is Not a Bell Curve

In product analytics, we often assume data follows a Normal Distribution (the Bell Curve): most people are in the middle, with a few fast and a few slow.

But Time To Value (TTV) almost never follows a bell curve. It follows a Log-Normal Power Law.

The Head: A large cluster of users activate quickly (e.g., 5-10 minutes).
The Long Tail: A smaller group of users get stuck, go to lunch, ask IT for permission, wait for a data import, and activate 3 days later.

The "Long Tail" is where the most dangerous friction lives.

The Math of Misleading Means

Let's look at a simple scenario with 5 users:

User A: 2 minutes
User B: 3 minutes
User C: 4 minutes
User D: 5 minutes
User E (The struggler): 4,320 minutes (3 days)

The Average TTV: ~866 minutes (14.5 hours).

If you report to your VP that "Our TTV is 14 hours," you are practically hallucinating.

For Users A-D, 14 hours is an eternity. They were done in minutes.
For User E, 14 hours is irrelevant. They took 3 days.

The "Average" describes nobody.

Enter Percentiles: The Professional Standard

To measure TTV accurately, you must use Percentiles.

P50 (The Median)

Value: 4 minutes. This tells you: "50% of our users succeed in 4 minutes or less." This is your Typical Experience. It filters out the noise of the outliers. If P50 increases, you broke the happy path.

P90 (The Struggle / The Long Tail)

Value: ~3,000+ minutes (depending on interpolation). This tells you: "How long does it take for almost everyone (90%) to succeed?"

If your P50 is 4 minutes but your P90 is 3 days, you have a massive Consistency Problem. It means your product is Russian Roulette: usually fast, but occasionally a black hole of wasted time.

The "TTV Spread"

At Tivalio, we encourage teams to track the TTV Spread:

Spread = P90 - P10

This metric represents Uncertainty.

Low Spread: Your onboarding is a factory line. Predictable. Reliable.
High Spread: Your onboarding is a maze. Some find the exit instantly; others get lost for days.

Conclusion

Stop optimizing for the average user. They don't exist.

Optimize P50 to improve your UI/UX efficiency.
Optimize P90 to fix deep-rooted friction and edge cases.

When you present data, ban the word "Average." Your engineering team will thank you.