Forge Academy: Performance Loop

The useful output of a weekly performance review is a defensible decision and a record of the evidence behind it. You might keep a message, clarify a buyer problem, replace weak proof or leave the inputs unchanged while gathering more information. A chart alone does not tell you which of those decisions is justified.

Forge retains marketing context associated with content and brings available supported-post metrics into Performance. This lesson shows how to use that context in a repeatable review. The routine includes a simple working record maintained by your team; it does not assume a built-in experiment ledger, native CRM attribution or automatic retraining of the model.

Set a question before opening the metrics

Choose a question connected to a decision you can make. “Did our content perform well?” is too broad. “Do catalogue managers respond more often to preparing product samples or defining the image views they need?” is narrow enough to guide a review, provided the published messages and available evidence actually address those audiences and problems.

Write the intended outcome beside the question. For an educational message, a relevant comment may reveal understanding. For a message offering a useful resource, link clicks may be a closer observable signal. For a commercial decision, verified enquiries or qualified conversations need their own business record. Do not silently replace one outcome with another because a different number looks better.

Use the same question across a reasonable set of comparable publications. A single unusually visible post can suggest an angle to investigate, but cannot establish a dependable preference. The message-testing guide develops stronger comparison designs when the decision requires more evidence.

Check whether the evidence is available

Open Performance and confirm which metrics are present for the posts you intend to review. According to the Forge User Guide, first metrics for supported posts are expected around 6 hours after publication, with timing affected by platform availability and permissions. Older posts refresh less frequently. Do not expect every newly published message to have a complete measurement record immediately.

Account support matters. The documented integration provides available analytics for supported LinkedIn Company Page, X, Facebook, Instagram and Threads posts, subject to permissions and format. LinkedIn personal-profile post analytics are not currently available inside Forge. If you examine a personal-profile result directly on LinkedIn, label that source in your working record instead of implying it came from Forge.

A dash can indicate unavailable or not-yet-arrived information. Do not enter it as zero. Check the account connection, the post's age and the latest information shown. The analytics and reporting guide explains metric definitions and missing-data handling in more detail.

Create a compact evidence record

For the selected posts, record enough context to reconstruct the comparison. A spreadsheet or team document can serve this purpose. Keep the original measurements and your interpretation in separate fields so a later reviewer can distinguish what happened from what you think it means.

• Identity: publication reference, date, account, surface and Node.
• Message: buyer, pain, proof, main angle and requested next action.
• Observation: metric source, observation date, post age and available counts.
• Conditions: relevant changes such as a different image, a changed offer or unusual external attention.
• Decision: what to keep, change or investigate, with the reason and next review date.

Do not collect unnecessary personal details to make the record feel comprehensive. A comment can be summarised as a relevant question about the offer without copying the author's profile into the knowledge base. Use authorised business evidence when updating a Node.

Make the comparison fair enough to be useful

Compare similar formats on the same account where possible, and observe posts at a similar age. A recent text post and an older image post have several differences before you consider their wording. If the sets are materially different, describe the difference and treat any conclusion as tentative.

Keep denominators explicit. Impressions count displays, while reach represents unique people or accounts as reported by the platform. A platform's engagement rate may use its own definition. If you calculate a separate rate, state the formula and use the same inputs across the compared posts. Do not relabel a custom calculation as the platform's own metric.

Also inspect the content itself. A click can lead to an unrelated page. A comment can be a complaint, a supplier pitch or a useful buying question. The count does not settle its commercial meaning. Review the nature of the response before changing the positioning.

Work through a hypothetical weekly comparison

Suppose a fictional catalogue product-photography service publishes 4 text posts on the same company account. There are 2 posts about preparing product samples and 2 about defining required image views. The team records each post after 7 days and aggregates the results by angle. These figures are illustrative and are not Forge customer performance.

Angle	Posts	Impressions	Link clicks	Clicks ÷ impressions
Preparing product samples	2	2400	36	1.5%
Defining image views	2	1800	36	2.0%

The image-view angle has the higher observed click-to-impression ratio, while both groups produced 36 clicks. Those are the observations. The team still needs to ask whether the destination links were equivalent, whether one post received unusual distribution and whether the visitors took a useful next step. With only 2 posts per angle, the table is a starting signal for further review.

A measured decision could be to keep both angles and clarify the image-view requirements in the Node using the language found in relevant audience questions. That is more defensible than removing the sample-preparation angle or announcing that image-view content converts better. The table measures clicks per impression, not verified conversions or causal impact.

Choose one input change and state your expectation

Return to the affected Node and identify the narrowest change supported by the review. If the audience repeatedly asks which views to request for a catalogue, the pain description may need to name that uncertainty. If people misunderstand the service's output, the description or proof may need to explain the deliverable more clearly.

Write the update and the reason before making it. For example: “Clarify that the photography brief identifies required front, side and detail views, because relevant questions repeatedly asked which images the catalogue needs. In the next comparable posts, inspect whether the same ambiguity remains.” This expectation is observable without promising a numerical uplift.

Keep other important choices stable where practical. Changing the buyer, offer, image, posting frequency and link simultaneously makes interpretation harder. If an urgent correction requires several changes, make the correction; simply record that the subsequent comparison will not isolate one variable. The Knowledge Base Setup lesson covers how to maintain the source information responsibly.

Close the loop at the next review

At the next weekly session, revisit the prior decision before selecting a new one. Ask whether the updated message was actually published, whether comparable data exists and whether the expected clarification appeared. The result may justify retaining the change, revising it or gathering more evidence.

“No change” is a valid outcome when the evidence is weak or the content already meets its purpose. Repeated editing without a reason can make a knowledge base less coherent. Keep a short record of inconclusive observations so they can be reconsidered when more relevant evidence arrives.

For a wider discussion involving several messages, commercial outcomes and next-period priorities, use the Campaign Review lesson. That broader retrospective can draw on the weekly records without confusing an operational signal with proof of incremental revenue.

Open Forge, choose a small comparable set and write one evidence-based decision. If you need help checking which measurements your connected accounts can support, request a demo focused on the performance review loop. The Forge platform overview explains the role of retained marketing context in the complete workflow.