For Managers
Creating observations
Write structured performance observations with context, behavior, impact, recommended action, drafts, attributes, and Henry assistance.
Observations are the building blocks of Performance Blocks. Each observation captures a single moment of behavior — a strength to reinforce or an opportunity to coach on — in a structured form that you can later share, summarize, or review with the employee.
This article covers the anatomy of a strong observation, every field on the form, draft and autosave behavior, attribute tagging, and Henry-assisted creation on the Agentic plan.
Why observations matter
Most performance reviews fail because they rely on memory at the end of a cycle. Observations flip that pattern: you capture short, specific notes throughout the period, then synthesize them into a summary when the time comes.
A well-written observation:
- Is specific enough that a reader who was not in the room can picture what happened.
- Separates the behavior from your interpretation of it.
- Names the impact so the employee understands why it matters.
- Suggests a clear next step so feedback becomes actionable.
The form structure in Performance Blocks is designed to nudge you toward all four of these.
Anatomy of a strong observation
Every observation follows the same four-part structure:
Context
A one-sentence setup that anchors the reader. Include the situation, the timeframe, and any relevant participants. Avoid loading this with judgment — save that for impact.
Behavior
The observable action. What did the person do or say? Stick to facts you witnessed or that were reported to you firsthand. Avoid hedged language like "I felt" or "it seemed" in this section.
Impact
Why the behavior mattered. Concrete, measurable impact is best ("the launch shipped two days early") but qualitative impact is valid too ("the new engineer felt unblocked and shipped their first PR that afternoon").
Recommended action
What you would like the employee to do with this feedback. For Strengths, this is usually about replication or scaling — "do this more, here is where else it would help." For Opportunities, it is a small, concrete next step the employee can act on this week or this month.
The two observation types
Every observation is either a Strength or an Opportunity.
Strengths
A strength is a behavior worth reinforcing. Strengths are not just compliments — they are signals about what the employee should do more of, in what contexts, and at what scope. A great strength observation makes the person better at the thing they are already good at.
Opportunities
An opportunity is a behavior worth changing. The word "opportunity" is intentional — it frames the feedback as forward-looking rather than punitive. Opportunities should be coachable; if the behavior is a serious performance issue, escalate through your people team rather than relying on observations alone.
A healthy observation history shows roughly 2–3 strengths for every opportunity. Persistent imbalance in either direction is worth a conversation with the employee or with your own manager.
Creating an observation
You can create an observation from several places in the app:
- Dashboard — click New observation in the recent observations panel.
- Employee profile — click New observation on the Observations tab.
- Observations page — click New observation in the page header.
- Keyboard shortcut — press
c ofrom anywhere with no input focused.
Selecting any of these opens the observation form in a side panel.
Step-by-step
- Select the employee. If you opened the form from a profile, this is pre-filled. Otherwise, type the employee's name and pick from the autocomplete list. You can only author observations about your direct reports unless your org admin has enabled cross-team observations.
- Choose the type. Toggle between Strength and Opportunity at the top of the form.
- Set the observation date. Defaults to today. Use the date this behavior actually occurred — accurate dates make summaries and trend reports meaningful.
- Write the observation text. This is where you describe the behavior and context. Aim for 2–4 sentences.
- Write the impact. A separate field; one or two sentences is enough.
- Write the recommended action. Be concrete and actionable.
- Tag attributes. Add 1–3 attributes that describe the underlying competency. See Attributes and skills for guidance.
- Save. The observation is saved immediately and added to the employee's history.
After saving you can choose to Share to conversation to drop the observation into an active 1:1 thread, or Close to return to whatever you were doing.
Field reference
| Field | Required | Notes |
|---|---|---|
| Employee | Yes | One observation = one employee. Use multiple observations for team moments. |
| Type | Yes | Strength or Opportunity. |
| Observation date | Yes | Defaults to today; backdating is permitted. |
| Observation text | Yes | The Context + Behavior portion. Soft cap at 2,000 characters. |
| Impact | Yes | The Impact portion. Soft cap at 1,000 characters. |
| Recommended action | Yes | The Recommended action portion. Soft cap at 1,000 characters. |
| Attributes | No | Up to five attributes per observation. |
| Visibility | Yes | Manager-only, or shared with the employee on save. Default is set by your org admin. |
Visibility
By default, observations you author are visible only to you and (depending on your org's policy) other managers in the employee's chain. When you share an observation into a conversation, the employee sees it then. Some organizations choose to make all observations visible to the employee immediately — your org admin sets this policy.
To check the policy in your org, look at the Visibility field at the bottom of the form. If you do not see a control, your org has a fixed policy.
Drafts and autosave
The observation form autosaves every few seconds while you type. Drafts are stored against your account, not the employee, so a draft will not appear on the employee's profile until you explicitly save.
You can have multiple drafts in progress at once. Open drafts appear in the Drafts section at the top of the Observations page, with the employee, type, and last-edited time. Click any draft to resume.
Drafts are kept for 30 days. After that, untouched drafts are deleted automatically. Drafts are not synced into summaries — only saved observations count as evidence.
Editing observations
You can edit any observation you authored. Open the observation from the Observations page or the employee profile and click Edit.
Edits are tracked in the observation's history. The history shows every change with a timestamp and author. Other managers in the employee's chain can see the history; the employee sees only the current version.
If you need to materially change a saved observation — for example, you remembered new context — edit it. If you want to add adjacent context, write a new observation rather than rewriting the old one. Multiple observations from the same period are valid and useful.
Archiving and restoring
You cannot delete a saved observation, but you can archive it. Archived observations:
- Are hidden from the default Observations view.
- Are excluded from quick stats and trend reports.
- Are excluded from new summary drafts.
- Are still visible in the employee's history with an Archived badge.
To archive: open the observation and click Archive in the overflow menu.
To restore: switch the Observations page filter to Include archived, open the observation, and click Restore.
Archive observations only when the underlying event turned out not to have happened, or when the observation is materially incorrect and a correction is not possible. Don't archive observations because the employee disagrees — disagreement is feedback to discuss in a conversation, not a reason to delete.
Tagging with attributes
Attributes are the competency tags your org admin maintains in the attribute library. Tagging observations with attributes is what powers attribute-based summaries, manager attribute ratings, and team-wide trend reports.
When you start typing in the Attributes field, the form autocompletes from your org's attribute library. You cannot create new attributes from the observation form — request additions from your org admin if a needed attribute is missing.
A few rules of thumb:
- Tag conservatively. 1–3 attributes per observation is ideal. More than five is too many.
- Pick the most specific attribute. "Communication" and "Written communication" should not both be tagged on the same observation.
- Don't tag what is not demonstrated. An observation showing strong code quality should be tagged with code-quality attributes, not with "Leadership" because the person is on a senior track.
For more, see Attributes and skills.
Examples
Strong example: Strength
Context: During the Q1 onboarding for the new product analytics tool, Sara ran the live walk-through for the 12-person engineering team.
Behavior: She prepared a written outline in advance, paused after each section to collect questions, and followed up the next day with a short FAQ that captured the questions she could not answer in the moment.
Impact: Of the 12 attendees, 10 had created their first dashboard within a week — the highest first-week activation we have measured. Two engineers cited the FAQ specifically in their team retro.
Recommended action: Sara, this is the right pattern for any internal training. Volunteer for the design-systems rollout next month and reuse the outline-plus-FAQ format. We should also share your outline as a template with other teams.
Attributes: Communication, Enablement.
Weak example: Strength
Sara is great at training. She did a really good job with the analytics rollout. Keep it up!
The weak version skips context, omits behavior detail, makes impact vague ("good job"), and offers no actionable next step. It is a compliment, not an observation.
Strong example: Opportunity
Context: In the design review for the billing redesign on April 14, Marcus presented three layout options and walked the room through trade-offs.
Behavior: When two reviewers pushed back on the chosen layout, Marcus restated his original argument three times rather than digging into the reviewers' concerns. The discussion ended without a decision and the project slipped a sprint.
Impact: The team lost a week of build time. The reviewers told me afterward they felt unheard, which makes them less likely to engage in future reviews.
Recommended action: For your next review, after presenting your recommendation, ask each dissenting reviewer to state their concern in their own words and reflect it back before responding. Aim to leave the room with either a decision or a concrete follow-up action, not a restated position.
Attributes: Collaboration, Decision-making.
Weak example: Opportunity
Marcus needs to get better at handling pushback in meetings.
The weak version is vague, attributes intent ("needs to"), provides no context for the employee to recognize the moment, and gives them nothing to do differently.
Henry assistance
Plan availability: Henry-assisted observation creation is available on the Agentic plan. Team plan customers can upgrade in Settings → Billing.
On the Agentic plan, you can use Henry to draft, refine, or extract observations from raw notes.
Draft from a moment
In the Henry panel, type a few sentences describing what happened — for example, "Sara ran the analytics rollout this morning and 10 of 12 engineers built dashboards by Friday." Then ask Henry to "turn this into an observation."
Henry runs the Create observation guided flow:
- It asks you to pick the employee from your direct reports.
- It asks you to confirm the type (Strength / Opportunity) — pre-suggested based on your input.
- It drafts the four sections (context, behavior, impact, recommended action).
- It suggests attributes from your org's library based on the content.
- You review, edit, and save.
You always see and approve the draft before anything is written to the employee's record. Nothing is saved without your explicit confirmation.
Refine an existing draft
If you have a draft observation that feels weak, open it and ask Henry to "make this more specific" or "add a recommended action." Henry will suggest a revised version inline; accept, edit, or discard.
Importing observations from email, Slack, Teams, and Chrome
Plan availability: Email, Slack, Teams, and Chrome capture are part of the Henry Agent suite on the Agentic plan.
On the Agentic plan, you can capture observations without leaving the surfaces where the work happens:
- Email — forward an email to your organization's Henry inbox to start an observation draft.
- Slack and Teams — message Henry with a moment, or use the message action on any thread to capture context into a draft.
- Chrome extension — highlight text in any browser tab and use the Capture observation action to send it to Henry.
In each case, Henry creates a draft observation seeded with the captured content and notifies you in the web app to review and save. Nothing is published without your review.
Keyboard shortcuts
| Shortcut | Action |
|---|---|
c o |
Open the new observation form |
Cmd/Ctrl + Enter |
Save the current draft |
Cmd/Ctrl + Shift + S |
Save and share into the employee's open conversation |
Esc |
Close the form (autosaves) |
Troubleshooting
I cannot find an employee in the picker
You can only author observations about your direct reports. If someone is missing, check the org chart and confirm the reporting relationship is current. Cross-team observations may be enabled by your org admin under organization settings.
My draft disappeared
Drafts are kept for 30 days. If yours is older than that, it has been deleted. If it is recent, check whether you are signed in as the correct account — drafts are scoped to the user that created them.
An attribute I want is not in the list
Attributes are managed by your org admin. Send them a request describing the competency you want to capture. See the admin docs for how the attribute library is curated.
What to read next
- Running 1:1s and conversations — share observations into the conversation flow.
- Writing performance summaries — observations are the evidence base for every summary.
- Attributes and skills — get the most out of attribute tagging.