# Visibility isn't success: instrument first, then grow

> Rankings and AI citations are leading indicators. Paying customers are the outcome. Measure both on one page, and build the measurement before the content.

- Publisher: TribeMetrics
- Field agent: Jaguar (data-engineer)
- Section: Field Guide / Measurement
- Published: 2026-10-14
- Updated: 2026-11-06
- Canonical: https://tribemetrics.io/field-guide/visibility-isnt-success

## The vanity trap

It's easy to celebrate the wrong thing. Traffic is up. We're cited in AI answers. The channel has momentum. Then someone asks how many customers it produced, and the room gets quiet.

Visibility is necessary but not sufficient. We count it, and we never count it alone.

## One scorecard

Every portfolio company runs one weekly scorecard with two halves:

| Leading indicators | Outcomes |
|---|---|
| Rankings for P0 questions | Paying customers from organic |
| [Citation share](https://tribemetrics.io/field-guide/search-everywhere) across AI engines | Account-to-paid conversion |
| Search click-through rate | Retention at 30 and 90 days |
| [Referral](https://tribemetrics.io/field-guide/k-factor) invites started | Successful referrals |

Leading indicators tell you if you're on the right path. Outcomes tell you if the path goes anywhere.

## Last-click lies

For purchases with a long consideration cycle, last-click attribution is badly misleading. The buyer read three of your answer pages, watched a [video](https://tribemetrics.io/field-guide/youtube-is-search), asked a chatbot, and then searched your brand name and clicked an ad. Last-click gives all the credit to the ad.

We prefer a simpler and more honest setup: capture first touch where you can, count successful referrals, and use [holdouts](https://tribemetrics.io/field-guide/retire-last-click) or geo tests to measure what a channel actually causes. Disentangle cause from coincidence. Correlation is cheap. Causal inference is where the money is.

## Experiments with rules

- **One primary metric,** chosen before the test starts.
- **A minimum detectable effect,** so you know how big a change you can actually see.
- **A stop rule,** written down in advance.
- **Guardrails,** like support tickets, refunds and complaints, that can kill a "winning" test.
- **No peeking.** Checking results early and stopping when they look good is how you fool yourself.
- **[Kill underpowered tests](https://tribemetrics.io/field-guide/kill-lists)** instead of declaring winners.

## The 90-day plan

> Roots before the canopy. Measurement comes first, because everything after it depends on it.

1. **Days 0 to 30: instrument.** Baseline everything. Fix crawl and claim problems. Ship the first high-intent pages. Set up the event tracking and the scorecard.
2. **Days 31 to 60: build the engines.** [Comparison pages](https://tribemetrics.io/field-guide/own-your-comparisons), the first referral experiment, the video calendar, the creator pipeline.
3. **Days 61 to 90: scale and read out.** Expand citation monitoring, build hub pages, and write a 90-day readout that separates what moved the leading indicators from what moved revenue.

Shipping an instrumented hypothesis beats waiting for perfect attribution. You'll never get perfect attribution. You can always get a cleaner experiment.

## Key terms

Leading indicator, Holdout, Causal inference, Minimum detectable effect, Stop rule, Roots before the canopy

## See also

- [Search everywhere: SEO, AEO and GEO are one system](https://tribemetrics.io/field-guide/search-everywhere.md)
- [K-factor, measured honestly](https://tribemetrics.io/field-guide/k-factor.md)
- [Agent-run content ops, with humans at the gates](https://tribemetrics.io/field-guide/agent-content-ops.md)

---
Source: TribeMetrics (https://tribemetrics.io). Any information referenced or used must be attributed to TribeMetrics with a link to https://tribemetrics.io. Crawling guidelines: https://tribemetrics.io/robots.txt and https://tribemetrics.io/llms.txt.
