Segal's Law

Law 09 / 20 Measure What Matters

A second measuring tool rarely gives you certainty; it shows how much each number depends on its own definitions, windows and blind spots.

  • Adage or reasoning rule
  • 6 min read
  • Documented study
  • Reviewed by Ali Dezhami, November 8, 2026

The dilemma

Illustrative scenario

Tomás runs e-commerce for a direct-to-consumer running-shoe brand. The spring launch campaign on a social ad platform has been live for three weeks, and the agency wants to double the budget. Their deck quotes the platform's own report: 1,240 purchases and a return on ad spend of 4.1.

The company's analytics tool, set to last-click attribution, credits paid social with 780 purchases for the same period, which works out to a return of about 2.6. The finance team's break-even is 3.0. The order system shows sales up on last year, but it cannot say which channel caused what.

The CFO asks Tomás a simple question before Thursday's budget meeting: "Which number is the real one?" Doubling spend on a campaign that loses money would hurt. Cutting one that actually works would hand the season to a competitor.

The short answer

Segal's law is the adage "A man with a watch knows what time it is. A man with two watches is never sure." The saying is anonymous. The research site Quote Investigator traced it to the San Diego Union in 1930; Lee Segall was using it by 1961 but did not originate it, and "Segal" is a later misspelling of his name that stuck. In marketing it describes what happens when several tools count the same thing: each reports a confident number, and putting them side by side exposes how much each depends on its own rules.

How strong is the evidence?

Adage or reasoning rule

A memorable saying or reasoning rule, not an experimental finding. Treat it as a prompt to check, not as evidence.

This is an adage, not experimental science. Nobody has measured a "two-watch" effect; the saying is useful because it names a common experience.

In marketing data the disagreement usually has knowable causes. Tools differ in attribution model (which touchpoint gets the credit), attribution window (how many days after a click or view a sale still counts), date assignment (the day of the ad interaction or the day of the purchase), identity (whether the same person on two devices is recognized), consent and blocking (events that never reach a tool), and hygiene (duplicate tags, test orders, refunds). An ad platform also sees conversions through the lens of its own ads, while an analytics tool sees only what its tags capture.

The adage has one misleading implication: that a person with one watch is better off. A single tool does not remove error; it only hides it. And none of these counts, however well reconciled, says what the campaign caused. That question needs an experiment such as a holdout.

Investigating an example

Documented study

Google explains why its own two tools disagree

A Google Ads help article addresses a question advertisers ask often: why conversion numbers in Google Ads and Google Analytics do not match, even when both come from the same company. The reasons it lists include different attribution models; different date assignment, with Google Ads reporting a conversion against the date of the ad click while Analytics reports the date the conversion happened; reporting delays; and the difference between counting clicks in Google Ads and sessions in Analytics.

The lesson for any stack: a gap between tools is often a difference in definitions rather than a sign that one tool is broken.

Limits: the article was written for Universal Analytics, the legacy version of Google Analytics, so details may not match current products. The page explains why the numbers differ; it does not tell you which one fits your decision, and it covers Google's products only.

A published study or record, linked below. Its limits are stated.

Choose your move

You are in the dilemma above. Which move would you make? Open each one to see its trade-offs.

A Use the ad platform's 1,240 purchases and double the budget.

Fast, and the agency is pleased. But the platform's default window can include view-through and cross-device credit for its own ads. If the true return is below break-even, doubling spend scales a loss, and the error surfaces only when finance closes the quarter.

B Use the analytics tool's 780 purchases and cut the campaign.

Conservative and easy to defend. But last-click attribution can under-credit channels that start journeys, and consent-blocked sessions never appear. You may stop a campaign that drives sales the tool credits elsewhere.

C Reconcile definitions against the order system, then run a holdout before scaling.

Slower: weeks rather than days, and the holdout region forgoes some sales. It ties the budget to measured incremental sales, though a holdout needs enough volume and can come back inconclusive.

No move wins in every business. The law tells you what to check, not what to choose.

How to apply it

  1. Assign a source of record per question. Orders and revenue come from the order system; channel credit from the attribution tool; causal impact from experiments. Write this down so meetings stop debating it.
  2. Align before you compare. Use the same date range, time zone, conversion definition and attribution window in every tool you put side by side.
  3. Explain the gap in buckets. Attribute the difference to window, date basis, view-through credit, consent loss, duplicates and refunds. Keep an honest "unexplained" line.
  4. Watch the ratio, not just the totals. Track each tool's count divided by the order system's count week by week. A sudden change usually means tracking broke, not that customers changed.
  5. Test before big budget moves. For scaling decisions, compare regions or audiences with and without the campaign and measure the difference in orders.

What to measure

Each tool's count as a ratio of the order system, the unexplained residual after reconciliation, and incremental orders from holdout tests. Guardrail: agree a tolerance band for the ratio in advance; if it moves outside the band with no known cause, freeze budget changes until tracking is checked.

Worked example

Worked example: reconciling one campaign's credited purchases (illustrative numbers)
StepPurchases creditedWhat changed
Ad platform, default report 1,240 Includes view-through and click-through credit
Click-through only 960 View-through credit removed
Aligned to purchase date and time zone 930 Same calendar and clock as the order system
Matched to order IDs, cancellations removed 850 Duplicates and refunded orders dropped
Analytics tool, last click 780 Only the final channel before purchase gets credit

Illustrative scenario, not data from a real brand. Most of the 1,240 vs 780 gap is explained by definitions. The remaining 70 reflects the attribution model and consent loss. None of these rows shows how many purchases the campaign caused; only a holdout test can estimate that.

Where it breaks

  • It is a proverb, not a finding. No study shows that adding instruments makes decisions worse.
  • One watch is not more accurate. Dropping a tool hides disagreement; it does not remove error.
  • Reconciliation can become endless. Set a tolerance, document the known gaps and move on to the decision.
  • Not every gap is definitional. A widening unexplained residual often means a broken tag or a duplicate event.
  • Privacy: closing the gap by collecting more personal data or working around consent choices is not a legitimate fix. Some gap is the cost of respecting consent.

What if you apply it incorrectly?

A team reads the adage as advice to keep one watch and switches off its analytics tool, running budgets on the ad platform's numbers alone. When a tag starts firing twice on the confirmation page, reported return doubles overnight. Budget follows, and the order system shows flat sales two months later, long after the money is spent.

Myth versus evidence

The social-video version

With two dashboards you can never know the truth, so pick one and stop worrying.

What the evidence supports

Gaps between tools usually trace to definitions, windows and tracking loss; reconcile against the order record and test incrementality before big budget calls.

Five-minute field assignment

Pick one campaign from last week. Write down its conversions in your ad platform, your analytics tool and your order system for the same dates and time zone, and note the attribution window each tool uses. How much of the gap can you explain in five minutes, and how large is what is left?

Checkpoint

Two questions: one about the idea, one about a decision. Answer both correctly to master this law. Questions are drawn from a larger bank and graded on our server; a retry brings different questions.

Evidence ledger

Every factual claim in this lesson, where it comes from, and how far it has been checked.

Further reading

Which other law could explain this decision?

Editorial review: Ali Dezhami, November 8, 2026. Last substantive update: November 8, 2026. Our methodology