> ## Documentation Index
> Fetch the complete documentation index at: https://docs.cortex.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Run OpEx reviews that end in decisions

> Practical lessons from teams running OpEx reviews with Cortex, from the first meeting to the habits that turn it into a lasting practice

PUBLIC BETA

An operational excellence (OpEx) review is the recurring meeting where engineering leaders look at how the organization is actually operating and decide what to do about it. [OpEx Review](https://docs.cortex.io/improve/opex-review) handles the prep: on the schedule you set, the agent pulls your Cortex data, organizes it by the DRIVE pillars (Delivery, Reliability, Initiatives, Vigilance, and Efficiency), and writes an executive briefing of what changed and what needs attention.

The report gets you to the meeting. This recipe covers what happens next, with lessons from teams running OpEx reviews in Cortex, from the first meeting to the habits that make the practice last.

## Before you start

**1. Get access to configure the report**

To set up OpEx Review, you need the `Configure OpEx` permission, which only Admins have by default.

**2. Give your leads access to the report**

Team leads need `View OpEx` to read reports, `Edit OpEx Comments & Follow-ups` to comment and manage follow-ups, and `Enable AI Chat` to use Cortex AI in the report. Managers and Admins have all three by default. See the [permissions reference](https://docs.cortex.io/configure/settings/managing-users/permissioning/permissions-reference.md).

**3. Check that your data is flowing into Cortex**

Starter metrics need deployment, version control, and incident data. For the supported tools and fallbacks, see [What do I need to get started?](https://docs.cortex.io/improve/opex-review/opex-faq.md#getting-started) in the OpEx Review FAQ.

## Step 1: Schedule your first review now

The hardest part is getting started, so book a recurring review before you think you're ready. It doesn't need to be perfect.

**Start with what you already have.** Your first few reviews will show you which metrics matter most, so start before you have a complete set. Teams often begin with one delivery metric, one reliability metric, and one or two Scorecards they already run. Delivery and Reliability are a good first pair because they counterbalance each other: shipping faster only counts if your services stay healthy. Expect to swap metrics in and out over time as your areas of friction and strategic priorities change. There's no perfect set.

**Use your Scorecards.** If you already score services on security, DORA metrics, or AI adoption, you have most of what you need for a first review. Adding those Scorecards to the Initiatives or Vigilance pillar also gets progress moving on the ones that have stalled.

**Line up the report with the meeting.** In the **Set schedule** section of the report configuration, the **Report period** sets both how much data each report covers and how often it generates. Match it to your meeting cadence, then set **Repeat on** to a day or two before the meeting so leads have time to read the briefing. For setup steps, see [Configuring OpEx Review](https://docs.cortex.io/improve/opex-review/configuring-opex.md).

The right meeting format depends on the size of your organization:

| | Startup or mid-market | Enterprise, local team | Enterprise, org-wide |
| - | - | - | - |
| **Cadence** | Weekly, 60 minutes | Weekly or biweekly, 30 minutes | Weekly or biweekly, 60 minutes |
| **Attendees** | VP of Engineering, EMs, tech leads, off-going on-call | Open invite within the team | VPs, directors, EMs, tech leads, off-going on-call across domains |
| **Report scope** | Team or domain | A single team, broken down by service | Team or domain |
| **Focus** | Full DRIVE report, critical and secondary metrics | Inward-facing metrics, uncurated SLOs, pager noise | Critical metrics only, major anomalies, and a deep dive on one team or service |

**Expect your first reports to surface data problems.** That's a good sign, because it means the review is working. Incidents left open for 90 days skew time to resolution, and incidents mapped to the wrong team hide where reliability problems are. Assign a follow-up to fix the process behind each one, and your data gets more accurate every cycle.

## Step 2: Send the executive briefing before the meeting

Long pre-reads don't get read. The executive briefing is designed to be short: a summary of what improved, what regressed, and where to focus, followed by highlights and concerns.

1. When the report generates, open it and click **Share** in the upper-right corner to copy a link to that exact report. Send it to the team leads a few days before the meeting. Anyone who opens the link needs the `View OpEx` permission.
2. Ask leads to review it and add context as [comments](https://docs.cortex.io/improve/opex-review/using-opex.md#commenting-on-a-report) on the summary, a highlight or concern, or a metric card. Comments stay with that reporting period, so the context is there during the meeting.
3. Encourage leads to use [Cortex AI](https://docs.cortex.io/improve/opex-review/using-opex-ai-chat.md) to dig into anything that stands out. If cycle time went up, for example, Cortex AI can trace the slowdown to reviews piling up on a few senior engineers.

When leads arrive already knowing the cause, the meeting can go straight to decisions.

## Step 3: Follow a consistent agenda

The same order every time keeps the meeting focused and makes sure follow-ups get checked:

1. **Review follow-ups from the last meeting.** Open the **Follow-ups** tab. The **Outstanding follow-ups** list shows every open follow-up plus anything completed since the previous report, along with the completion rate.
2. **Review anything that went red since the last meeting**, like SLO breaches and critical incidents, and assign follow-ups. If you track Datadog SLOs, click the SLO summary to see the SLOs closest to breaching at the top of the list.
3. **Pick one or two topics** from the briefing's highlights and concerns, like rising cycle time or a stalled Scorecard, and decide what needs a follow-up.

You won't get to every metric, and you don't need to. Note where you left off, and pick it back up at the next meeting.

A few things help the meeting run well:

* **Keep the collaboration in the report.** Discuss metrics in comment threads and `@`-mention the people who need to weigh in, instead of scattering the conversation across Slack and docs.
* **Answer questions in the room.** When someone asks which team is driving a spike, click **Ask Cortex AI** on the metric card instead of turning the question into a post-meeting task.
* **Have technical leaders drive.** Both executive and staff-plus technical leaders should be the early adopters and lead the discussion in the meeting.
* **Keep it blameless.** The review should target problems in the system, not individuals.

## Step 4: Capture every decision as a follow-up

Teams whose earlier OpEx processes failed tend to say the same thing: they didn't record decisions, and the meeting turned into reviewing metrics for the sake of it.

* **Give every decision an owner.** Click **Add follow-up** from any tab of the report, describe the action item, assign an owner, and add a link to the ticket or doc. You can also ask Cortex AI to draft one, for example *Create a follow-up for Platform Engineering to look into the cycle time increase.* For details, see [Tracking follow-ups](https://docs.cortex.io/improve/opex-review/using-opex.md#tracking-follow-ups).
* **Make follow-ups small enough to finish before the next meeting.** For example, update an SLO definition based on what the review showed, or have a team discuss how to spread code reviews more evenly after review times go up.
* **Turn bigger work into an Initiative.** If a follow-up is bigger than one cycle, break it into smaller steps or create an [Initiative](https://docs.cortex.io/improve/initiatives) so you can track progress across teams between reviews.

> [!NOTE]
>
> Assigning a follow-up doesn't notify the assignee yet, and it doesn't give them access to OpEx Review. Let owners know about their follow-ups, and make sure they have the `View OpEx` permission.

## Step 5: Make it stick

* **Keep OpEx top-of-mind between meetings.** Ask leads to check the report between reviews. Most of the culture change comes from that habit, because leads start thinking about these signals regularly, not just once a quarter. [Comment notifications](https://docs.cortex.io/configure/settings/notifications) for mentions and thread activity help keep the conversation going, and you can reply to them from Slack.
* **Incorporate what you learn into the next meeting.** If a metric never drives a decision, swap it out. If the same question keeps coming up, add the metric that answers it, or add an existing [custom dashboard](https://docs.cortex.io/improve/eng-intelligence/dashboards/custom) as a tab so the view is ready next time. Configuration changes appear in the summary starting with the next report.
* **Track your follow-through.** The follow-up completion rate is a check on whether the review is actually changing what teams work on. If it drops, make follow-ups smaller or move larger ones into Initiatives.

> [!TIP]
>
> If an OpEx review hasn't stuck for your team before, it was likely because the meeting turned into status updates. Let OpEx Review handle the prep and set the agenda, and spend the meeting on decisions.

## Related resources

* [OpEx Review overview](https://docs.cortex.io/improve/opex-review)
* [Configuring OpEx Review](https://docs.cortex.io/improve/opex-review/configuring-opex.md)
* [Using the OpEx Review report](https://docs.cortex.io/improve/opex-review/using-opex.md)
* [Using Cortex AI in OpEx Review](https://docs.cortex.io/improve/opex-review/using-opex-ai-chat.md)
* [The DRIVE framework](https://www.cortex.io/report/drive-framework)


## Related topics

- [OpEx Review](https://docs.cortex.io/guides/product-areas/opex-review.md)
- [Configuring OpEx Review](https://docs.cortex.io/improve/opex-review/configuring-opex.md)
- [Using the OpEx Review report](https://docs.cortex.io/improve/opex-review/using-opex.md)
