If you manage writers, you have probably sat down to write a content team performance review and realized you didn't have much to point to. You know who writes well and who struggles, but "this one is a strong writer" is hard to defend when the person across the table asks what exactly you mean. This guide is for managers of writers and content marketers who are already on the team, whether you run an agency delivery team or an in-house group. By the end you'll have a short rubric, a way to pick the work you review, a set of fair process and outcome signals, and a cadence your writers can predict.
A quick note on scope. This is about people who are already on your team. It doesn't cover hiring or how to structure roles. The steps below are a practical framework, not a validated instrument, so adapt them to your writers, your content types, and your goals.
Why reviewing writers feels so hard
Most managers assume the problem is that writing is subjective and traffic is objective, so the answer must be to lean on the numbers. The research doesn't really back that up. CIPD's evidence review of people performance measures reports validity and reliability concerns with both objective and subjective measures. A manager's rating can be shaped by the rater, the relationship, and the setting. A hard number can also be an incomplete picture of the work.
The same review points to a better approach, which is to combine several kinds of evidence and keep possible bias in mind while you do it. For a content role that means looking at the quality of the work, how reliably it gets delivered, what it did for readers and the business, and how the person has grown. You don't need to pick one of those. You need all of them, each used for what it can honestly tell you.
One more thing before the steps. Performance management is often thinner than people assume. CIPD's 2023 UK data found that 48% of employees had at least a basic level of performance management, meaning specific objectives set for their job, and just over a quarter had what it calls systematic performance management. If your team has been reviewed on a gut feeling so far, you're in common company, and building something simple is already a step forward.
Step 1: Set the review period and agree on expectations
Start at the beginning of the period, not the end. Sit down with each writer and agree on what success means for the work they actually own. Write down the review period, their responsibilities, the kinds of work they'll produce, the quality standard, the process expectations, any development goals, and the evidence you'll look at.
Keep the common criteria the same across writers and document what's different. At an agency that might be client briefs, voice requirements, and approval dependencies. For an in-house marketing lead it might be different products, audiences, or channels. If priorities change in the middle of the period, write the change down and agree on the new expectation together, so nobody gets scored against a target that moved without them knowing.
You're done when the writer can explain the criteria in their own words, knows which evidence will count, and has a record of the goals and any constraints you both agreed to.
Common mistake: goals like "write more," "be strategic," or "improve quality." Nobody can tell later whether those were met. Turn them into something you can observe, such as hitting agreed delivery dates, producing work that meets the shared editorial standard, taking on a set level of assignment complexity, or improving one named outcome.
Step 2: Build a short rubric that makes quality observable
This is the step that does the most to answer the "good writing is hard to score" problem, and it's where you start when you evaluate writer performance in a way people can follow. Before you score anyone, write a shared rubric with a few dimensions you can use consistently. Here is a set that works for most content teams:
- Reader usefulness and goal completion. Is the audience clear? Does the piece answer the reader's question or help them finish the task? Would they leave satisfied?
- Accuracy and substantiation. Are factual claims backed by appropriate sources? Are product, customer, and company claims correct? Are limits and uncertainty stated honestly?
- Originality and added value. Does the piece add real insight, experience, or synthesis, or does it just reword what already exists?
- Brief, audience, and brand fit. Does it meet the assignment's purpose, format, voice, and agreed claims or exclusions?
- Structure and usability. Does the title describe the piece accurately? Can a reader find what they need? Are spelling, style, and formatting errors under control?
Google's own self-assessment questions for helpful content are a useful backdrop here, since they ask about the intended audience, expertise, original value, and whether a reader leaves satisfied. They also say plainly that there is no preferred word count, which is worth remembering before you put length in a rubric. The five dimensions above are our adaptation, not a Google scoring scheme. For more on how quality raters think about good content, see Google's quality rater guidelines in plain terms.
If you need numbers, a simple four-level scale is enough:
- 1, Below the agreed standard. Recurring material gaps, and the work needs substantial correction or close support.
- 2, Inconsistent. Meets some expectations but needs repeated, targeted improvement.
- 3, Meets expectations. Reliably delivers the agreed standard for the assignment and level of support.
- 4, Exceeds expectations. Consistently stronger work or a broader contribution, shown by specific examples, without giving up accuracy or the agreed requirements.
These anchors are examples, not a research-backed scale. They only become useful when you attach real examples from your own team's work to each level before the cycle starts. Avoid labels like "great writer," "not strategic," or "not a culture fit," because they can't be checked against anything. If you want a deeper look at scoring drafts themselves, our rubric for scoring draft quality goes through that in more detail.
You're done when each rating level has an example from your team's real work and every writer is judged on the same core dimensions.
Pro tip: Have the people who'll do the reviews score two or three sample pieces together before the first formal review. When you disagree, point to the rubric and the work, then fix whatever wording was unclear. It's much easier to sort out a vague criterion on a sample piece than in the middle of someone's review.
Step 3: Review a fair sample of work, not a memorable outlier
Memory is the enemy of a fair review. The article that went great last week, or the one the client complained about, will crowd out the other eleven months if you let it. So collect a sample that covers the whole period and the formats or clients the writer worked on. Include strong work and work that needed improvement, and pull in the brief, the source materials, the feedback history, and the revisions along with the published piece.
Then record concrete examples against the rubric. For each one, you should be able to say which criterion it speaks to and why.
Take care not to compare things that aren't alike. A short social post and a researched long-form article are different assignments, and treating them the same punishes the person who took on the harder work. Think about complexity, how clear the brief was, how much research it needed, whether the client or stakeholders changed direction, how long approvals took, and whether the writer had access to a subject matter expert. A good review separates the decisions the writer made from the constraints they were handed. Good briefs help here too, and a repeatable briefing system makes it much easier to tell whether a miss came from the writing or from the assignment.
You're done when the review has examples from across the whole period and you can explain how each one maps to an agreed criterion.
Common mistake: using one article, one complaint, or the last few weeks as a stand-in for the whole cycle. The sample exists to make your reasoning visible and to reduce your reliance on memory. It's not a portfolio contest.
Step 4: Score delivery and process separately from prose quality
A writer can be reliable without being a strong stylist, and the other way around. If you fold both into one number, you lose the ability to tell the person what to work on. So keep delivery and process as their own part of the review, and agree up front on which behaviors matter for the person's assignments. Reasonable ones include:
- Delivering against deadlines you both agreed to, with dependencies and scope changes recorded.
- Following the brief, format, sourcing, review, and approval requirements.
- Flagging blockers, unclear claims, missing inputs, or deadline risks early.
- Communicating clearly at handoffs and responding well to editorial feedback.
- Reducing preventable repeat errors and avoidable rework over time.
Be careful with revision counts. A lot of revisions can mean an unclear brief, a client who changed their mind, new facts, or normal editorial back and forth. Note why a piece needed heavy rework before you attach it to the writer. And don't reward raw volume unless the work also meets the standard and the assignments are comparable, because output on its own can push people toward speed at the expense of the things the rubric is meant to protect.
If your team works with a stored brand voice and product facts, some of the repeat-error problem goes away on its own. DeepSmith's Deep IQ keeps your positioning, personas, product facts, and brand voice as shared context, so drafts start from the same facts and voice no matter who is writing. That makes it easier to see what's genuinely a writer's judgment call and what was just a missing input. It does not score people, and it isn't meant to. It's one way to take avoidable noise out of the process evidence you're collecting. If you manage several client accounts, you may also want to see how agencies govern content quality across clients.
You're done when each process criterion is something the person can reasonably control, deadlines and exceptions have context, and quality and output are not collapsed into a single number.
Step 5: Connect content outcomes to the purpose of each assignment
Now you can bring in the numbers, which matter even more in a content marketer performance review, since marketers are often closer to the channels and conversions. There's one rule: pick a metric because it fits the purpose of the piece, not because it's easy to count. The Content Marketing Institute's measurement framework recommends matching metrics to the purpose and the audience journey, and it describes indicators across three broad areas:
- Awareness and reach. Sessions or users, with a suggestion to watch engaged sessions rather than rely on raw traffic alone. In that framework an engaged session lasts at least 10 seconds, includes a conversion event, or has two or more page views.
- Engagement. Social interactions relative to reach, video play and watch time, scroll depth, downloads, shares, and increases in branded search.
- Conversion and later impact. Lead-form completions, content-assisted conversions, customer advocacy, and retention or NPS among audiences who engage with content.
A how-to article meant to educate should be judged differently from a landing page meant to drive demo requests. Where you can, compare like with like, meaning similar formats, similar audiences, similar distribution support, and similar measurement windows. If you want more help choosing the right signals, we've written about which blog KPIs actually matter and about measuring content performance beyond traffic.
The part that matters most for fairness is separating one writer's contribution from a shared business result. A writer is accountable for the quality of the work, whether it fits its agreed objective, whether they apply what the data teaches them, and the improvements they make within their remit. They aren't solely accountable for lead volume, revenue, or retention, because those depend on distribution, product, sales, budget, and other people. Privacy rules, cookie changes, and platform shifts also make attribution incomplete, and multi-touch attribution doesn't prove one person caused a conversion. If your team publishes only a few posts a month, the metrics that matter at four posts a month are a better fit than a big dashboard. This is a real operating problem, too. In the Content Marketing Institute's 2025 B2B research, measuring the results of content efforts was the second most cited challenge, at 47%.
If AI search citations are part of what your team is trying to earn, page-level citation data can serve as one outcome signal. DeepSmith's AEO area shows which of your pages AI engines actually cite and which prompts drive those citations, so you can look at the result of a piece with the same context you'd use for any other outcome. As with traffic, use it as evidence to discuss with the writer, not as a verdict on them. If traffic is the number your leadership keeps asking about, this piece on organic traffic as a metric can help you frame the conversation.
You're done when every outcome indicator is tied to an agreed purpose, has a stated measurement window, and is read with its dependencies and data limits in mind.
Common mistake: ranking writers by pageviews or leads without accounting for topic demand, assignment type, distribution, how old the post is, or how the conversion path is attributed. Raw traffic can be affected by plenty of things the writer never touched.
Step 6: Assess development and growing scope against agreed expectations
Growth is part of any content marketer performance review, but it's not a requirement that everyone follows the same career path. Agree on what growth means for each person in the work they already do. Examples are handling more complex briefs on their own, doing original research or stronger source synthesis, making sound editorial decisions with less intervention, using audience or performance insight in later work, or taking on an agreed mentoring or review contribution where it makes sense.
Compare a writer to their own earlier expectations and development goals, not to a colleague who had easier assignments, better inputs, or more visible clients. Write down what support, practice, or feedback you'll provide in return. It's also worth saying out loud that a bigger scope shouldn't quietly become a permanent expectation without a conversation, because writers notice when that happens and it erodes trust in the whole process.
Keeping the editing standard consistent across contributors is one place where growth is easy to see. If you're also working on voice consistency, our guide to keeping one brand voice across many contributors is a useful companion.
You're done when the review names a specific skill or responsibility that changed, gives examples of it, and records a realistic next development action.
Step 7: Hold regular check-ins and a formal review conversation
A steady content team feedback cadence means nobody is surprised at the formal review. CIPD materials describe regular check-ins, continuous feedback, and coaching alongside end-of-year assessments, and another CIPD source gives weekly or monthly check-ins as examples. What the evidence doesn't do is show one correct frequency. So treat the cadence below as an adaptable starting point, not a study finding.
- During each assignment. Give prompt, specific editorial feedback while the work and context are still fresh.
- Monthly. Hold a short one-to-one about workload, blockers, feedback themes, and progress on goals.
- Quarterly. Look at a sample of work and the agreed process and outcome indicators, and adjust objectives if the work or business priorities have materially changed.
- At the end of the formal cycle. Summarize the period against the original criteria, hear the writer's view, agree on development actions, and record any decisions through your organization's normal process.
If your publishing rhythm is different, pick another regular interval. What matters is that feedback is timely and people know when the formal assessment happens. If you run several accounts at once, setting cadence and capacity across a client roster can help you keep those check-ins from getting squeezed out.
In the review meeting itself, ask the writer to bring a self-assessment and a few examples. Compare those with your own evidence, talk through any differences using the same rubric, and agree on concrete next steps. Keep feedback specific by describing the work or behavior, what effect it had, and what should happen next. Record strengths as well as areas to improve. If the review affects pay or promotion, follow your organization's policies and the laws that apply to you, since this guide isn't legal advice.
You're done when the writer understands the assessment and can see how the evidence, the criteria, the feedback, and the next steps connect.
Fairness checks to run all the way through
A few habits protect every content team performance review, and they're worth keeping in view even after the steps are in place.
- Agree the criteria first. They shouldn't be invented after you've already decided whether someone is a strong or weak performer.
- Use the same core rubric. Tailor the goals to the assignment, but evaluate everyone with the same method.
- Ask for examples. Replace "overall impression" with prompts about which expectation was met, what evidence supports that, and what should happen next.
- Revisit the original goals and note any changes in scope or circumstances.
- Use more than one kind of evidence. Neither a single manager rating nor a single KPI tells the whole story.
- Look at patterns across the team. Before you finalize ratings, compare how you applied the criteria to different people and look for unexplained differences. Calibration is a safeguard, not a guarantee.
- Don't let the formal review be the first time someone hears about a recurring issue. Use check-ins for that.
SHRM's practical account of fixing biased evaluations recommends the same basic moves, which are defining a rubric in advance, applying it consistently, asking managers to use specific evidence, and reviewing the original goals. It reports that consistent criteria reduced gender gaps in ratings in one organizational case. That's a useful reported case and not proof that a rubric removes bias everywhere. CIPD's evidence is similarly nuanced. It reports that rater personality traits accounted cumulatively for 6% to 22% of the variance in performance ratings in the reviewed studies, and it found little evidence of overall gender bias in real work settings, while noting a substantial pro-male bias in studies with only male raters. None of that means every rating is biased, and none of it means a form fixes the problem. It does mean that a clear shared rubric, specific examples, and a check on your own patterns are worth the effort.
What to do next
You don't have to roll all of this out at once. Pick one writer and one review period, write down the criteria with them, and put a monthly one-to-one on the calendar. After the first cycle, look at which parts of the rubric were hard to use and fix the wording before you extend it to the rest of the team. Over a couple of cycles you'll have examples for each rating level, and the reviews get easier to write and easier for writers to trust.
If your team produces a lot of content and you'd like to spend less of your own time on the mechanics of drafting, linking, and formatting, you can start a DeepSmith free trial and see how much of that work a platform can take on, so the review conversations can stay about judgment and growth.
