# Native LinkedIn A/B test brief

Copy this brief. Fill **Plan** before native setup. Fill **Observed** only from actual Campaign Manager and CRM records. Write `Unknown` where evidence is missing.

## Plan

| Field | Your entry |
| --- | --- |
| Account, test owner, and decision owner |  |
| Business question and hypothesis |  |
| LinkedIn test variable: Ad, Audience, Placement, or Optimization |  |
| One difference between A and B |  |
| Variant A label and planned setting |  |
| Variant B label and planned setting |  |
| Shared objective, audience, geography, offer, landing page, conversion event, placement, bid/optimization, and creative settings that are not the chosen variable |  |
| Selected native test metric and why it answers the hypothesis |  |
| Budget currency; total test budget; A ad-set budget; B ad-set budget; daily or lifetime basis |  |
| Creation date; start date; end date; result review date; CRM lag/cutoff date; time zone |  |
| Business guardrail and decision rule, including acceptable cost per qualified outcome if used |  |
| CRM qualification rule, person/account dedup key, variant attribution rule and window, outcome window, and expected source/ID join |  |
| Prelaunch checks: tracking, consent/region effects, audience, overlapping Brand Lift test, and owner of native setup |  |

The native test must be created in Campaign Manager. LinkedIn's documented flow uses two ad sets and a single selected variable. Its help pages say the test starts no earlier than the day after creation, runs 14–90 days, and pauses both ad sets at the end. Check current account eligibility and the platform's displayed budget requirements before launch; this worksheet does not assert that a proposed plan was accepted. Avoid changing variants during the run.

## Observed: native result

| Field | Evidence from Campaign Manager |
| --- | --- |
| Actual account/test ID and A/B ad-set or variant IDs |  |
| Actual start/end, spend A/B with currency, and selected metric |  |
| Native result category and A/B metric values |  |
| Result evidence URL/export, capture time, and any edits, pause, or extension |  |
| Native interpretation: winner, directional/inconclusive, similar, pending, or unknown |  |

Record LinkedIn's result category as shown. A directional or inconclusive result is not a proven winner. A result can change after the test ends, so keep the capture time and review again if needed.

## Observed: CRM outcome

| Field | Evidence from CRM/export |
| --- | --- |
| Export/query ID, data snapshot, capture time, and outcome cutoff |  |
| A and B distinct attributed cohorts; matching rule and attribution window |  |
| Qualification definition, qualified A/B counts, and disqualified/unknown counts |  |
| Person/account dedup key; duplicate and unmatched join counts by variant |  |
| Pipeline/opportunities A/B, stage rule, and value/currency if claimed |  |
| Cost per qualified outcome A/B, only when spend and matched counts are comparable |  |
| Cohort comparability, lag completion, and attribution limitations |  |

Keep different snapshots or attribution windows labeled. An empty or broken join is **Unknown**, not zero qualified leads. Do not compare A and B as if they had the same cohort when IDs, dates, qualification rules, or attribution windows differ.

## Decision log

| Field | Decision |
| --- | --- |
| What LinkedIn supports |  |
| What CRM supports and what remains unknown |  |
| Guardrail pass/fail/unknown and reason |  |
| Next action, owner, and date |  |