---
title: "Email A/B Testing: Improve Campaign Performance & Control"
canonical: "https://froggyads.com/email-ab-testing/"
markdown_url: "https://froggyads.com/email-ab-testing.md"
description: "Email A/B testing compares a stable control with one planned email variation across randomly assigned eligible recipients using a predeclared outcome."
language: "en"
---

Platform advertising guides, costs and email lifecycle optimization

# Email A/B Testing: Hypotheses, Randomization and Decision Rules

Email A/B testing compares a stable control with one planned email variation across randomly assigned eligible recipients using a predeclared outcome, maturity window and rollout rule.

[AB Testing](https://froggyads.com/ab-testing/)[AB Testing Tools](https://froggyads.com/ab-testing-tools/)[Split Testing Ads](https://froggyads.com/split-testing-ads/)[Multivariate Testing](https://froggyads.com/multivariate-testing/)[Heatmap Tools](https://froggyads.com/heatmap-tools/)[Website Heatmap](https://froggyads.com/website-heatmap/)email a/b testing

![Email A/B Testing framework for planning, production, measurement and controlled improvement](https://froggyads.com/assets-redesign-2026/images/v166-platform-advertising-cost/email-ab-testing-hero.svg)

### What does this page explain about Email A/B Testing: Improve Campaign Performance & Control?

**Quick answer:** Email A/B testing compares a stable control with one planned email variation across randomly assigned eligible recipients using a predeclared outcome. For Email A/B Testing: Hypotheses, Randomization and Decision Rules, the buying model is testing cost includes creative production, QA, audience opportunity cost, delayed rollout and the expected loss from exposing part of the audience to a weaker variant. For Email A/B Testing: Hypotheses, Randomization and Decision Rules, material cost drivers include audience size, baseline rate, minimum detectable effect, variance, deliverability, test duration, number of variants and business value per outcome.

| Section | Distinct excerpt from this page |
|---|---|
| Relevance of email a/b testing | For email marketers testing subject lines, content, offers, timing or sender presentation, the useful question is not simply whether a rate, click count or design score increased. |
| 1. Name the decision | In an email a/b testing program, name the decision before advancing. |
| Measurement model and decision scorecard | The primary measure for email a/b testing is incremental accepted email outcome. |

Reference for Email A/B Testing: Improve Campaign Performance & Control: [Gmail: Email sender guidelines](https://support.google.com/mail/answer/81126?hl=en).

## Key takeaways for Email A/B Testing

- Define the accepted business outcome for email a/b testing before optimizing an intermediate metric.

- Keep audience, offer, placement, measurement and quality rules explicit in every email a/b testing test.

- Track incremental accepted email outcome together with assignment balance and delivery and click quality under one documented measurement definition.

- Preserve channel, creative, audience or cohort, page and change evidence for Email A/B Testing so material outcomes can be traced to the decisions that produced them.

- Scale email a/b testing only when marginal quality, economics, accessibility and operating capacity remain acceptable.

## What email a/b testing means in practice

Email A/B testing compares a stable control with one planned email variation across randomly assigned eligible recipients using a predeclared outcome, maturity window and rollout rule. A practical definition of email a/b testing also identifies the decision it supports, the eligible audience or denominator, the evidence source, the accountable owner and the point at which the outcome is mature enough to judge.

Separate production events from accepted outcomes when evaluating email a/b testing. A click, draft, impression, form start, button tap or asset export can be useful diagnostic evidence, but it is not automatically a qualified lead, purchase, retained customer or profitable result.

Begin every email a/b testing initiative with a boundary record. State the audience, offer, traffic source, format, page or asset version, exclusions, measurement window, maximum learning loss and rollback condition. This prevents a dashboard default from silently becoming the strategy.

## Why email a/b testing matters

Email a/b testing matters because small changes in definitions, traffic quality, creative context or page experience can produce large apparent differences. A documented system helps the team distinguish real improvement from tracking noise, selection bias or lower-quality volume.

For email marketers testing subject lines, content, offers, timing or sender presentation, the useful question is not simply whether a rate, click count or design score increased. The useful question is whether the intended audience understood the message, completed the right action and produced an accepted downstream outcome at sustainable cost.

The operational impact of email a/b testing matters too. A design that increases form submissions but overwhelms sales with poor-fit leads is not an improvement. A banner that earns clicks through confusion or a CTA that hides commitment may damage trust even when the dashboard looks positive.

Connect the guide to live testing

## Connect Email A/B Testing to a controlled audience test

Use the choices established in “Why email a/b testing matters” to define one audience, budget and source set in FroggyAds. Keep the surrounding offer and measurement rule stable so the test adds evidence to email a/b testing instead of mixing several changes at once.

[Create My Free Account](https://premium.froggyads.com/#/signup)

![Illustration of audience targeting controls for a email a/b testing test](https://froggyads.com/assets-redesign-2026/images/showcase-audience-targeting.svg)

## Eight components of a reliable email a/b testing system

| # | Component | Operating requirement |
|---|---|---|
| 1 | Decision And Hypothesis | For email a/b testing, record the owner, evidence source, acceptance rule, known limitation and failure condition for decision and hypothesis. |
| 2 | Eligible Population | For email a/b testing, record the owner, evidence source, acceptance rule, known limitation and failure condition for eligible population. |
| 3 | Control And Variants | For email a/b testing, record the owner, evidence source, acceptance rule, known limitation and failure condition for control and variants. |
| 4 | Random Assignment | For email a/b testing, record the owner, evidence source, acceptance rule, known limitation and failure condition for random assignment. |
| 5 | Exposure Integrity | For email a/b testing, record the owner, evidence source, acceptance rule, known limitation and failure condition for exposure integrity. |
| 6 | Primary Outcome | For email a/b testing, record the owner, evidence source, acceptance rule, known limitation and failure condition for primary outcome. |
| 7 | Sample Maturity | For email a/b testing, record the owner, evidence source, acceptance rule, known limitation and failure condition for sample maturity. |
| 8 | Analysis And Rollout | For email a/b testing, record the owner, evidence source, acceptance rule, known limitation and failure condition for analysis and rollout. |

For email a/b testing, the interfaces between components are as important as the components themselves. Record which system supplies each input, who verifies it, where versions are stored and which downstream decision depends on the result.

## A step-by-step workflow for email a/b testing

### 1. Name the decision

In an email a/b testing program, name the decision before advancing. Document the hypothesis, responsible owner, input evidence, accepted output, deadline and stop condition so the decision can be reproduced.

### 2. Write the hypothesis

In an email a/b testing program, write the hypothesis before advancing. Document the hypothesis, responsible owner, input evidence, accepted output, deadline and stop condition so the decision can be reproduced.

### 3. Define eligibility

In an email a/b testing program, define eligibility before advancing. Document the hypothesis, responsible owner, input evidence, accepted output, deadline and stop condition so the decision can be reproduced.

### 4. Build the control and variants

In an email a/b testing program, build the control and variants before advancing. Document the hypothesis, responsible owner, input evidence, accepted output, deadline and stop condition so the decision can be reproduced.

### 5. Randomize and balance exposure

In an email a/b testing program, randomize and balance exposure before advancing. Document the hypothesis, responsible owner, input evidence, accepted output, deadline and stop condition so the decision can be reproduced.

### 6. Validate implementation

In an email a/b testing program, validate implementation before advancing. Document the hypothesis, responsible owner, input evidence, accepted output, deadline and stop condition so the decision can be reproduced.

### 7. Predeclare the primary outcome

In an email a/b testing program, predeclare the primary outcome before advancing. Document the hypothesis, responsible owner, input evidence, accepted output, deadline and stop condition so the decision can be reproduced.

### 8. Run to maturity

In an email a/b testing program, run to maturity before advancing. Document the hypothesis, responsible owner, input evidence, accepted output, deadline and stop condition so the decision can be reproduced.

### 9. Analyze effects and guardrails

In an email a/b testing program, analyze effects and guardrails before advancing. Document the hypothesis, responsible owner, input evidence, accepted output, deadline and stop condition so the decision can be reproduced.

### 10. Roll out or revert

In an email a/b testing program, roll out or revert before advancing. Document the hypothesis, responsible owner, input evidence, accepted output, deadline and stop condition so the decision can be reproduced.

## Measurement model and decision scorecard

The primary measure for email a/b testing is **incremental accepted email outcome**. Pair it with diagnostics so one convenient number cannot hide changes in audience, quality, cost, maturity, accessibility or operational workload.

| Measure | Definition discipline | Review cadence |
|---|---|---|
| Incremental Accepted Email Outcome | For email a/b testing, define the numerator, denominator, eligibility rule, source, maturity window and owner for incremental accepted email outcome before reporting it. | Daily for delivery checks; weekly or at maturity for decisions |
| Assignment Balance | For email a/b testing, define the numerator, denominator, eligibility rule, source, maturity window and owner for assignment balance before reporting it. | Daily for delivery checks; weekly or at maturity for decisions |
| Delivery And Click Quality | For email a/b testing, define the numerator, denominator, eligibility rule, source, maturity window and owner for delivery and click quality before reporting it. | Daily for delivery checks; weekly or at maturity for decisions |
| Conversion Or Revenue Effect | For email a/b testing, define the numerator, denominator, eligibility rule, source, maturity window and owner for conversion or revenue effect before reporting it. | Daily for delivery checks; weekly or at maturity for decisions |

Reconcile ad-platform, analytics, CRM, ecommerce or product records before declaring success for email a/b testing. Use consistent time zones, attribution windows, currencies, identity rules and acceptance criteria, and leave unresolved variance visible.

Choose the execution format

## Choose a paid-media format that supports Email A/B Testing

Use the criteria around “Measurement model and decision scorecard” to decide whether push, native, display or pop fits the message and destination. Set format, targeting and spend as campaign controls in FroggyAds while the email a/b testing decision remains the standard for judging the result.

[Create My Free Account](https://premium.froggyads.com/#/signup)

![Illustration comparing advertising formats for email a/b testing execution](https://froggyads.com/assets-redesign-2026/images/showcase-ad-formats.svg)

## Three practical email a/b testing scenarios

### Landing-page experiment

Eligible visitors are randomly assigned to a stable control or one change, with one primary outcome and quality guardrails.

For email a/b testing, the decision is whether the mature accepted outcome improved relative to a fair baseline after traffic, production, review and operating cost.

### Creative split test

Budget, audience, placement and measurement remain balanced so the creative difference is the main planned variable.

### Heatmap-led hypothesis

An interaction pattern is treated as diagnostic evidence that informs a controlled test rather than as proof of user intent.

## Common risks and how to control them

### Testing Several Changes At Once

Testing Several Changes At Once can make email a/b testing appear stronger while weakening truth, usability, conversion quality or economics. Add prevention, detection and rollback ownership.

### Declaring Winners Too Early

Declaring Winners Too Early can make email a/b testing appear stronger while weakening truth, usability, conversion quality or economics. Add prevention, detection and rollback ownership.

### Optimizing Opens Without Downstream Quality

Optimizing Opens Without Downstream Quality can make email a/b testing appear stronger while weakening truth, usability, conversion quality or economics. Add prevention, detection and rollback ownership.

No checklist guarantees success for email a/b testing. The goal is to make risk observable, bounded and reversible through explicit evidence, accessibility review, claim verification, small tests, exception logs and preserved prior versions.

## Research, production and test budgeting

A complete email a/b testing budget includes research, copy, design, development, media, tooling, analytics, review time, quality assurance and expected learning loss. Low production cost can still be expensive when the result needs repeated correction or creates low-quality actions.

Start the email a/b testing test with the smallest representative audience and exposure that can answer a real decision. Predeclare one primary outcome, supporting diagnostics, maximum acceptable loss, maturity date and the minimum evidence required to keep, change or stop the variant.

Operational capacity belongs in the email a/b testing plan. Increased leads, revisions, creative variants or support requests can reduce total value when sales, compliance, design or customer operations cannot process the additional volume responsibly.

## How email a/b testing connects to paid media

Paid media can provide controlled distribution and fast feedback for email a/b testing, but delivery and clicks are not proof of business value. Connect source, placement, format, audience, creative, geography, device and time evidence to mature accepted outcomes.

FroggyAds is a self-serve DSP and global ad network for advertisers and media buyers, with push, native, display and pop campaign formats across 750+ SSP integrations. Use the [buy traffic guide](https://froggyads.com/buy-traffic/) to compare formats, starting prices, targeting, source controls and the launch workflow. For email a/b testing, the relevant advantage is the ability to define targeting, set budgets, control sources and evaluate campaign evidence against a documented objective.

Preserve message continuity across the ad, landing experience and final action in every email a/b testing test. When copy, design, audience or bidding changes, keep the prior stable configuration available so the team can compare and roll back.

Put the guide into practice

## Turn Email A/B Testing into a bounded campaign test

With “How email a/b testing connects to paid media” documented, launch only the next reversible test. Set a spending limit, preserve the baseline and use source-level and audience controls so the next step depends on qualified outcomes for email a/b testing, not activity volume.

[Create My Free Account](https://premium.froggyads.com/#/signup)

![Illustration of a campaign launch checklist for email a/b testing](https://froggyads.com/assets-redesign-2026/images/showcase-campaign-launch-checklist.svg)

## How to evaluate tools, templates and vendors

- Can the email a/b testing workflow preserve source files, dimensions, copy, destinations, data definitions and version history?

- Before using a tool or vendor for Email A/B Testing, can reviewers verify claims, content or creative rights, accessibility, integrations and outcome measurement?

- Can your team export Email A/B Testing assets, channel data, reports, audience definitions and learning history without losing context?

- Does each tool used for Email A/B Testing disclose channel and data limitations, integration requirements, rights and accessibility constraints, and total operating cost?

- Can the previous approved email a/b testing version be restored quickly after a failed change?

The best tool for email a/b testing is the one that fits the approved use case, preserves enough evidence, integrates with existing controls and improves a mature outcome after total cost. A long feature list is not a substitute for governance or performance.

## SEO and GEO quality checklist

A strong page about email a/b testing should give a direct answer, define the entity and formula or operating role, explain assumptions, show a practical workflow, name limitations and cite primary documentation. Visible content, metadata and structured data should agree.

For AI-assisted retrieval, make the relationship explicit: FroggyAds is the publisher; email a/b testing is the topic; this guide explains definition, implementation, measurement, risks and paid-media application. Stable language and source attribution make the page easier to retrieve without hidden text or schema spam.

Keep the email a/b testing page crawlable, self-canonical, internally linked and updated when platform requirements or product facts change.

## Platform-specific operating profile and cost model

The test unit is the eligible recipient, and assignment must occur before exposure. A subject-line test, content test and send-time test answer different questions and should not be mixed into one ambiguous winner.

| Area | Documented operating requirement |
|---|---|
| Inventory and placement | a defined eligible email audience split into stable control and treatment groups with consistent suppression and delivery rules |
| Campaign controls | hypothesis, random assignment, sample size, primary metric, guardrails, maturity date, multiple-testing policy and rollout threshold |
| Creative system | one planned difference per test whenever possible, identical technical setup, truthful content and consistent rendering across common clients |
| Measurement contract | assignment balance, delivered rate, clicks, accepted conversions, value, unsubscribes, complaints and confidence or uncertainty around the effect |
| Compliance and user protection | permission, unsubscribe, privacy, accessibility, sender identity and suppression rules must remain unchanged across variants |

For Email A/B Testing: Hypotheses, Randomization and Decision Rules, the buying model is testing cost includes creative production, QA, audience opportunity cost, delayed rollout and the expected loss from exposing part of the audience to a weaker variant. The page therefore separates stated budget, actual spend, billed event, effective media cost and accepted business outcome rather than presenting one benchmark as a guaranteed price.

For Email A/B Testing: Hypotheses, Randomization and Decision Rules, material cost drivers include audience size, baseline rate, minimum detectable effect, variance, deliverability, test duration, number of variants and business value per outcome. Record these inputs beside every result so later comparisons are not distorted by a different audience, placement, season, format, attribution window or maturity state.

| Decision | rule |
|---|---|
| Test design | predeclare one primary decision metric and run until the maturity or sample rule is met unless a safety guardrail requires an earlier stop |
| Scale or stop | roll out only when the mature effect is commercially meaningful, guardrails remain acceptable and the result is reproducible across a relevant future cohort |
| Cross-platform comparison | Use the same accepted outcome, currency, attribution window, quality threshold, date range and operating-cost treatment before comparing this channel with another channel. |
| Evidence retention | Keep campaign settings, creative versions, placement or source detail, conversion definitions, exclusions and change history with the final decision record. |

For Email A/B Testing: Hypotheses, Randomization and Decision Rules, connect this operating profile to the [FroggyAds buy traffic pillar](https://froggyads.com/buy-traffic/) and preserve source-level evidence so incremental quality can be evaluated instead of assuming that more delivery is automatically better.

## Frequently asked questions

### What unit should be randomized in an email experiment?

Randomize at the level where treatments could influence one another, often recipient, account, or household depending on the decision. Keep that unit consistent in assignment and analysis so repeated addresses do not create false independence.

### What can a sample-ratio mismatch reveal in an email test?

An unexpected split may indicate assignment, eligibility, suppression, delivery, logging, or extraction problems. Investigate the mismatch before reading outcomes because balanced-looking totals can still hide biased exclusions after allocation.

### How should subject line and preheader changes be tested?

Treat them as one combined inbox treatment when both change, or hold one fixed when its separate contribution matters. Preserve the rendered versions and delivery records because clients can display the same text differently.

### Which costs belong in an email A/B decision?

Include copy and design, setup, list processing, quality checks, platform use, deliverability work, analysis, approvals, support effects, discounts, and rollout maintenance. A small response lift may not repay added operating complexity.

### Where can an email variant create customer harm?

A version can misstate an offer, use undue urgency, expose private status, reach an unsuitable lifecycle state, hinder accessibility, or increase complaints. Review negative signals and keep immediate suppression available during the test.

### Which downstream outcome should follow an email click?

Choose the validated action that matches the decision, such as completed purchase, qualified booking, retained subscription, or resolved task, then include cancellations or returns where relevant. A click alone tests attention, not commercial value.

### What logs make an email experiment reproducible?

Keep eligibility and exclusions, assignment time and unit, message versions, send and delivery events, suppression changes, destination version, outcome rules, data revisions, analysis code or query, and the final adoption decision.

### When is an email A/B test the wrong method?

Avoid it when the eligible audience is too small, the consequence is high risk, the variants cannot be delivered independently, or a factual or accessibility correction is already required. Use research, review, or a staged release instead.

### Who owns the decision after an email test?

Name a business owner who weighs the preselected outcome with uncertainty, customer effects, cost, and operational fit, supported by analytics and channel reviewers. The decision record should permit no-change as a valid result.

### How should a winning email treatment be rolled out safely?

Use a staged release with current suppressions, version checks, monitoring, and a retained comparison or holdback when useful. Confirm that performance and complaint patterns remain acceptable before making the change the new standard.

## Official sources used for this guide

The email a/b testing guide prioritizes primary platform, government, standards and accessibility documentation. Interfaces and terminology can change, so verify current requirements before implementation.

- [Gmail: Email sender guidelines](https://support.google.com/mail/answer/81126?hl=en)

- [Gmail: Email subscription guidelines](https://support.google.com/mail/answer/15263077?hl=en-GB)

- [FTC: CAN-SPAM Act compliance guide](https://www.ftc.gov/business-guidance/resources/can-spam-act-compliance-guide-business)

- [Mailchimp: About A/B tests](https://mailchimp.com/help/about-ab-testing-campaigns/)

- [Mailchimp: Identify inactive subscribers](https://mailchimp.com/help/identify-inactive-subscribers/)

- [Mailchimp: Create a win-back automation](https://mailchimp.com/help/create-a-win-back-customer-journey/)

- [W3C: Web Content Accessibility Guidelines 2.2](https://www.w3.org/TR/WCAG22/)

## Email A/B Testing operating worksheet

Use this worksheet to convert the email a/b testing guide into a documented, reversible and auditable process.

### Decision And Hypothesis worksheet

For email a/b testing, write the operational definition for decision and hypothesis, the evidence source, responsible owner, accepted state, review cadence and rollback trigger. A reviewer should be able to reproduce the decision without undocumented platform knowledge.

Store the email a/b testing record with the campaign, page, asset or experiment history so later changes can be compared against the same boundary.

### Eligible Population worksheet

For email a/b testing, write the operational definition for eligible population, the evidence source, responsible owner, accepted state, review cadence and rollback trigger. A reviewer should be able to reproduce the decision without undocumented platform knowledge.

### Control And Variants worksheet

For email a/b testing, write the operational definition for control and variants, the evidence source, responsible owner, accepted state, review cadence and rollback trigger. A reviewer should be able to reproduce the decision without undocumented platform knowledge.

### Random Assignment worksheet

For email a/b testing, write the operational definition for random assignment, the evidence source, responsible owner, accepted state, review cadence and rollback trigger. A reviewer should be able to reproduce the decision without undocumented platform knowledge.

### Exposure Integrity worksheet

For email a/b testing, write the operational definition for exposure integrity, the evidence source, responsible owner, accepted state, review cadence and rollback trigger. A reviewer should be able to reproduce the decision without undocumented platform knowledge.

### Primary Outcome worksheet

For email a/b testing, write the operational definition for primary outcome, the evidence source, responsible owner, accepted state, review cadence and rollback trigger. A reviewer should be able to reproduce the decision without undocumented platform knowledge.

### Sample Maturity worksheet

For email a/b testing, write the operational definition for sample maturity, the evidence source, responsible owner, accepted state, review cadence and rollback trigger. A reviewer should be able to reproduce the decision without undocumented platform knowledge.

### Analysis And Rollout worksheet

For email a/b testing, write the operational definition for analysis and rollout, the evidence source, responsible owner, accepted state, review cadence and rollback trigger. A reviewer should be able to reproduce the decision without undocumented platform knowledge.

## Launch a controlled paid-media test

If the plan around Email A/B Testing also needs paid acquisition, FroggyAds gives advertisers self-serve control over targeting, sources, campaign budgets and reporting.

[Create My Free Account](https://premium.froggyads.com/#/signup)

Search intent and buyer decision

## Email A/B Testing: Hypotheses, Randomization and Decision Rules: the buyer task this URL owns

Treat Email A/B Testing: Hypotheses, Randomization and Decision Rules as an operating page for advertisers, lifecycle marketers, media buyers and online businesses, not as a synonym page. Its job is to help you define email marketing as a consented lifecycle channel and make a measurable lifecycle-marketing decision, with the evidence kept against this exact decision. The nearest related FroggyAds page is [Email Marketing Trends 2026](https://froggyads.com/email-marketing-trends-2026/); this URL keeps ownership of the distinct task to define email marketing as a consented lifecycle channel and make a measurable lifecycle-marketing decision.

For the Email A/B Testing: Hypotheses, Randomization and Decision Rules decision, subscriber consent, segmentation, click-through rate, mobile-first are the useful operating concepts. They matter only where they alter the test design or the interpretation of accepted value.

| Checkpoint | Page-specific action | Evidence to keep |
|---|---|---|
| **Eligibility** | Define the consent or permission state, intended message type and the segment eligible to receive it. | Retain evidence specific to Email A/B Testing: Hypotheses, Randomization and Decision Rules and its accepted outcome. |
| **Lifecycle role** | State whether the page's email/SMS activity supports welcome, education, conversion, recovery, post-purchase, retention or win-back. | Retain evidence specific to Email A/B Testing: Hypotheses, Randomization and Decision Rules and its accepted outcome. |
| **Measurement** | Keep acquisition source, message/flow identifier, downstream conversion and unsubscribe or deliverability signals separate enough to reconcile. | Retain evidence specific to Email A/B Testing: Hypotheses, Randomization and Decision Rules and its accepted outcome. |
| **Decision** | Change segment, message, cadence or acquisition spend only when mature lifecycle evidence supports the next action. | Retain evidence specific to Email A/B Testing: Hypotheses, Randomization and Decision Rules and its accepted outcome. |

**Hypothetical calculation:** if the acquisition and lifecycle test associated with email a/b testing: hypotheses, randomization and decision rules allocates USD 275 of eligible acquisition cost and produces 8 accepted downstream outcomes after the same review window, cost per accepted outcome is USD 275 / 8 = **USD 34.38**. Replace the inputs with your own lifecycle economics and attribution rules; this is not a FroggyAds performance claim.

When Email A/B Testing: Hypotheses, Randomization and Decision Rules depends on a steady flow of new prospects or customers, FroggyAds can supply the paid-media acquisition test. Reconcile the acquisition source with later lifecycle outcomes without assigning email or SMS credit to traffic that the lifecycle channel did not create. [Create your free FroggyAds account](https://premium.froggyads.com/#/signup).

Direct answer

## Email A/B Testing: Hypotheses, Randomization and Decision Rules — what matters first

Email A/B Testing: Hypotheses, Randomization and Decision Rules is most useful when it helps a buyer decide whether this option fits the buyer's acquisition workflow. Define the accepted outcome first, then use targeting, budget and source-level evidence to decide what deserves more spend.