A lot of people treat review goals as something you pick from a list. You find a page of ready-made sentences, choose three that sound like your job and change the number. But a copied goal has no starting point and no record behind it. So when the review comes round, nobody can say for sure whether you met it.
The research on goals shows why that matters. In Building a Practically Useful Theory of Goal Setting and Task Motivation, Edwin Locke and Gary Latham looked back over 35 years of studies. They found that specific, difficult goals consistently led to higher performance than simply urging people to do their best. That’s a strong finding. But a specific goal only helps you at review time if you and your manager can later agree whether you hit it, and that agreement is what this post is about.
You’ll see how to work out where you stand now and decide what evidence the work will leave behind. You’ll also see how to agree the measure with the person who will judge it. Eight worked goals then show the method for different kinds of work. The research behind it comes from studies of specific tasks rather than review goals, and the section on where goals fail explains what that means for you.
This post is about goals you set for yourself for the next cycle. If you’re setting targets for a company or a department, the guide to performance management goals for organizations covers that. Managers setting goals for their own role will find more in performance goals for managers. If you need to rate the year that just ended, that’s a self-evaluation and it comes up again near the end.
What makes a review goal one you can prove later?
You can test any goal with three plain questions. The first is where you’re starting from. If you don’t know your current number, there’s nothing to compare the end of the cycle against.
The second question is what will exist afterwards to show the work happened. That could be a report, a finished piece of work or a comment from someone who used it. If you can’t name it now, you’ll be trying to reconstruct it from memory in a year.
The third question is who agrees that this is the standard, which is usually your manager. The time to get their agreement is when the goal is written, because at the review it’s too late to argue about what should have counted.
These questions sit on top of the usual advice. A SMART goal is meant to be measurable, but SMART doesn’t tell you where the measuring starts or what record you’ll keep. The baseline and the evidence trail fill that gap. They’re also what goal setting looks like as a skill you use all year, rather than a form you fill in once.
It also helps to keep the list short. Locke and Latham describe an older study by Locke and Bryan (1969) in which people got feedback on several parts of their driving. They improved on the parts they had goals for but not on the others. A goal pulls your attention toward what it names and does little for the rest. So ten goals won’t give you ten improvements. Three or four goals that cover the work that matters most will serve you better, because you can actually keep track of them.
The same paper says goals work best when people are committed to them. When later studies failed to repeat the effect, the authors trace part of the reason to missing feedback or missing commitment. That’s the practical case for the three questions. A baseline and an evidence trail give you feedback during the cycle, so you know whether you’re on track before the review tells you. Agreeing the goal with your manager makes the commitment real for both of you.
You might expect goals you write yourself to work better than goals you’re handed, but the paper is more mixed on that. Locke and Latham describe a study by Latham and Yukl (1976) in which word-processing operators with specific high goals improved whether the goal was assigned to them or set with them. They also report that people who helped set their goals tended to set higher ones. But they say the main gain from taking part is that people share what they know. So the reason to draft your own goals is to make sure you and your manager understand them the same way. It gives you a say in what will count as evidence, and that’s worth having even if it doesn’t make you work any harder.
How do you find your baseline when nobody has measured it?
A baseline is where you stand before the goal starts. Many people skip it because nobody has ever measured their work. But there’s usually more to go on than you’d think, and each source has its own trap.
| Where the starting point comes from | What it looks like | The trap |
|---|---|---|
| A number your tools already hold | Tickets, deals, defects or turnaround times in a system you use | One busy or quiet season can make it misleading |
| A sample of your own work, scored | Ten recent pieces of work scored against a short rubric you write down | A small sample tells you little, and changing the rubric later breaks the comparison |
| A one-question check with the people who use your work | The same short question sent to two or three people | People are polite, so a vague question gets a vague answer |
| A count of one behavior over a few weeks | How often you interrupt, reply late or hand work back | You only count what you remember to note |
| A plain before-state | ”I can’t do this without help yet” | It’s honest but it isn’t a number |
The first row is the easiest place to start, because the number already exists. You can get past the seasonal trap by using several months of data or by comparing with the same period a year earlier. The scored sample works for work that has no count attached, like reports or designs, but you have to write the rubric down before you score anything and keep it the same at the end.
The one-question check gets better answers when the question is specific. “How am I doing?” gets a kind reply. “Did my last handover tell you everything you needed?” gets a useful one. The behavior count works best when you keep the tally somewhere you’ll see it every day. The before-state suits a learning goal better than an outcome goal, which the next section explains.
One more check applies to every row. Ask how much of the number is yours. A turnaround time that depends on another team’s approvals will move for reasons you can’t touch, which doesn’t rule it out but means you’ll need to pair it with something you do control, as the sales example below shows.
Sometimes no baseline exists at all. Maybe the role is new, or the work has never been tracked, and in that case a fair goal is to set one up. You agree to measure the work by a date, such as the end of the first month, and you set the target at that first checkpoint. By then you know the real starting point.
Tell your manager plainly that this is the plan. It can feel weaker than a goal with a number in it, but a target picked before anyone knows the start is a guess, and you’d be judged against that guess for a year.
When should the goal be to learn, not to hit a number?
Most advice on review goals assumes the goal should be a number. Locke and Latham’s paper says that’s sometimes wrong. On tasks that are complex for the person doing them, the authors report that a hard performance goal sometimes does worse than simply asking people to do their best. It can make people so anxious to succeed that they scramble for ways to do the task without testing them in any orderly way. The remedy the authors describe is a learning goal that is still specific and challenging. Their example is to discover a set number of different strategies for mastering the task, which draws on work by Winters and Latham (1996).
That gives you a simple rule. If you’ve never done this kind of work at the level your role now needs, set the goal on what you’ll learn or try by a certain date. Once you’ve done enough to have a baseline, add the outcome goal. That’s usually at the first checkpoint, a couple of months in.
A learning goal isn’t “take a course” or “get better at Excel”. Neither of those can be assessed, because finishing a course says nothing about what you can now do. A learning goal you can be assessed on ends in something real. It might be something you’ve built or run. It could also be work someone else will review, or a skill you can show in front of other people. If you need to plan several learning goals against a timeline, an individual development plan is the better tool.
Goals about people skills need the same care. “Improve communication by 20%” sounds measurable, but there’s no honest way to measure it. Nobody can say what 20% more communication would look like. The measure for a people-skill goal is a change in behavior that other people can see.
That might be a colleague’s one-line answer to the same question at the start and end of the cycle. Or it might be a count of how often the behavior happens. Neither is as tidy as a sales figure, but both can be shown at a review, and both come from the people who actually see the change.
What do good review goals look like for different kinds of work?
The eight goals below cover three kinds of goal across different jobs. Each one shows the situation and where the baseline came from. It also shows how the goal was worded, what evidence the work leaves behind and what the person might say at the review. The people here are illustrative composites made up for this guide, and so are the numbers. They’re there to show the method. Your own numbers have to come from your own baseline. So read the example closest to your work first, then skim the others for moves you can borrow.
Outcome goals
Sophie works on a customer support team at a software company. Her team’s tool records every ticket, including the ones customers reopen because the first answer didn’t fix the problem. Nobody had asked her to watch that number, but she pulled it for the last twelve weeks and found that 160 of her 1,140 closed tickets had been reopened. That’s a reopen rate of about 14%.
She used twelve weeks rather than one month because a single month can be thrown off by a product release. Once she had the number, she looked at what was behind the reopens, and most came from tickets she’d closed after one reply without checking the fix had worked. So she built her goal around that habit. Over the next two quarters she’d bring her reopen rate from about 14% to 10% or lower, measured the same way in the same tool. The target came from her own number. Cutting the rate by roughly a third looked hard but possible, because one habit caused most of the problem.
The evidence is the tool’s own report, which she exports every four weeks and saves in one folder. She doesn’t need to write anything up, because the tool keeps the record for her. At the review she’d lead with the change and the period: “Over the last two quarters my reopen rate went from 14% to 9%, measured the same way in the ticket tool. The main change was checking the fix with the customer before closing.” Leading with the period and the source tells her manager the comparison is fair before they’ve had to ask.
Her manager might still push back: “Couldn’t that be because the product got more stable?” It’s a fair question, so Sophie should have the answer ready. She can pull the team’s reopen rate for the same two periods from the same report. If the team’s rate barely moved while hers dropped, the change is hers. If both fell, she can say so honestly and point to the part that was different for her.
Oliver is an account manager who looks after existing customers in one region. His goal sheet asks for growth in expansion, which means existing customers buying more. But his numbers swing with the seasons, because most of his customers set budgets at the start of the year. His region also lost two large accounts in a reorganization that had nothing to do with him, so comparing this quarter with last quarter would mislead everyone.
The fairer baseline is the same quarter a year earlier. Oliver pulled it from the CRM and found that 9 of his accounts had expanded in that quarter, and he also looked at something he controls directly. After each renewal call, how often did he send a follow-up with a proposal within two working days? Going through his CRM notes, he found it happened about half the time.
So his goal has two parts. The outcome part compares expansion this year with the same quarter last year, with the two lost accounts left out of both years. The input part is a follow-up within two working days after at least 9 in 10 renewal calls. He controls the second part completely and the first only partly. Pairing them means he won’t be judged only on what his customers’ budgets do. Both numbers come from the CRM his manager already uses, because it records the date of every follow-up.
At the review he might say: “Twelve accounts expanded this quarter against nine in the same quarter last year, leaving out the two we lost. I followed up within two days after 92% of renewal calls, up from about half.” His manager might answer that follow-ups are activity, not results. Oliver’s reply is that the input goal was never meant to stand alone. It’s there because the outcome depends partly on budgets he can’t see. If expansion had stayed flat while his follow-ups doubled, they’d both know the problem lay somewhere else.
Fiona processes supplier invoices in a finance team. Her system already records how long each invoice takes from arrival to payment, but nobody reads it. Her first instinct was to write “pay invoices faster” and pick a target of three days. Instead she did it the other way round. She pulled the number first, and the median turned out to be six working days. Only then did she set a target of four days, because about two of those days went on waiting for her own follow-ups about missing purchase order numbers.
Two months in, the company added an extra approval step for larger invoices. Turnaround went up for everyone, and none of it was Fiona’s doing. She flagged it at her first checkpoint and showed her manager the system report with the date the new step began, and they agreed to change the measure. From then on her goal covered only the part she controls, which is the time from an invoice arriving to it being ready for approval. So her review won’t show a clean win on the original goal. It will show the goal changing and the date she raised it. It will also show how she did on the part that was hers.
Learning goals
Owen is a data analyst who has built every weekly report in spreadsheets. His team is moving its reporting to Python, and he has never written more than a few lines of it. So the honest baseline isn’t a number but the plain fact that he can’t yet build a working report in Python without help.
A performance goal here would be something like halving the time it takes to produce the weekly report. But he has no idea how long a Python report will take him, so that target would be a guess. It would also push him to grab the first method that seems to work, which is the scramble Locke and Latham describe. So his goal for the first two months is about learning. He’ll rebuild three existing weekly reports in Python by the end of month two. Cleaning the raw data is the trickiest step, so he’ll try at least two different approaches there and note which one held up.
The evidence is the reports themselves and a code review on each one from a senior analyst. Each review sits in the team’s code repository with a date on it. The outcome goal waits until the first checkpoint, and by then Owen will know how long a report takes him, and he and his manager can set a time target from that real starting point. At the review he can describe both stages. He’ll show what he learned first, and then what he delivered once he knew enough to commit to a number.
Hannah joined a payroll team eight months ago, and payroll runs under strict tax and employment rules. She’s careful, but a senior colleague still checks every line of her year-end work. She started from the end, which helps when knowledge is hard to count. She wrote down what she’d want to be able to say at her next review. It was that she now prepares the year-end checks on her own, and a named senior colleague has signed them off without corrections.
Working backwards from that sentence gave her both the goal and the evidence. One piece of evidence is the set of year-end checks with a sign-off from Marcus, the colleague who reviews her work now. The second is a twenty-minute walk-through she’ll give the team on one rule she had to learn this year. If she can explain it to people who know the work well, that shows the knowledge more clearly than any rating could. The goal is about what she’ll know by the end of the next cycle. It doesn’t try to judge how much she knew last year. That’s a rating of the past cycle, and the performance review phrases for job knowledge cover that side.
People-skill goals
Tom is a project lead who runs a weekly status meeting for five people from other teams. He suspected the meeting wasn’t working, but a suspicion isn’t a baseline. So he started with the evidence he’d want at the end. Before changing anything, he sent each of the five the same question: “In one line, what did you get from last week’s meeting?” Two said they’d left with a decision they needed, while three said nothing in particular. He also counted from his calendar how many of the last ten meetings had run over, and it came to seven.
Only then did he write the goal. By the end of the quarter the meeting would finish on time in most weeks. At least four of the five people would also be able to name something they got from it. He’ll ask the same question of the same people at the midpoint and again before the review. Their replies are the evidence once he saves them with the dates. They won’t add up to a score, but they show the change in the words of the people who sit in the meeting, and that’s harder to argue with than his own opinion.
Claire is a product designer. In design reviews she tends to stay quiet about problems she sees in an engineer’s plan. She raises them later in private messages instead, often after the build has started. Her first draft goal was “communicate more openly with engineering.” She read it back and couldn’t say what she’d show at the review to prove it. Nobody else could have scored it either.
So she rewrote it around one behavior. First she kept a short private log for three weeks. Each time she noticed a concern in a review, she noted whether she raised it in the room or afterwards. The log showed she’d raised one concern out of six in the room. Her new goal was about that ratio. She’d give constructive feedback on the plan during the review itself, so most concerns would come up before the work started. The log is the evidence, and she’ll keep it running through the cycle. At the checkpoint she can also ask the engineering lead whether he’s had fewer late surprises from her. That’s softer than a count, and she should present it as softer.
Ben was a senior engineer until six months ago, when he became lead of a team of four. He still writes much of the hardest code himself, and his manager has told him he’s become a bottleneck. Delegation was the obvious goal, but “delegate more” can’t be measured. So he looked at how work moves through the team. Over four weeks he listed every task he took on that someone else could have done, and there were eleven. His goal was to hand off tasks like those with a clear outcome and a deadline. He’d also track how many came back to him unfinished.
At the midpoint the evidence was mixed. He’d handed off nine tasks and five came back done. But four came back for him to finish, and he’d taken two of those back himself before anyone asked for help. Ben told his manager exactly that at the checkpoint, without dressing it up. Handing off more was working, but taking work back too early wasn’t. They kept the goal as it was and added one check. Before he takes a task back, he’ll ask the person what they need first. Whether that’s enough to call the goal met will be his manager’s call at the review.
If your job isn’t covered here, look for the example with the same shape. A project role usually has a delivery date. That works like Fiona’s turnaround figure, including the risk that someone else moves it. Design and writing roles can use a quality score from peer review, as long as the same reviewers score the work at the start and at the checkpoint. That’s the same idea as Tom asking the same people each time. Customer success roles often have a retention or renewal number. That needs Oliver’s same-period comparison and an input goal beside it. A recruiter might track how many candidates reach a final interview. That depends partly on hiring managers, so it needs the same split between what you control and what you don’t.
What do you collect along the way, and what do you say at the review?
Evidence is easiest to collect when you decide what it is at the start. That way the record builds up as you go, so you aren’t trying to remember a whole year when the review form opens.
When you agree the goals, write down the baseline for each one and where it came from. Next to it, add the measure and where the evidence will live. That might be a folder, a saved report or a thread with your manager. It takes one sitting of about half an hour.
Pull the number and save it at the first checkpoint, which is usually four to six weeks in. If you have a learning goal, this is also when you set its outcome target. Add any comment from someone who used your work while it’s still fresh. After that, ten minutes every couple of weeks is usually enough to keep the record current.
At the midpoint, check whether each goal still fits the work. If something has changed, raise it now. A mid-year review is the natural time, but you don’t need to wait for one. Send your manager a short note saying what changed and when it happened. Add what you’d suggest instead. Keep that note, because it’s your evidence that you flagged the problem.
Put it all together in the weeks before the review. The guide to writing your self-evaluation covers the write-up and how to keep a running log, so there’s no need to repeat that here. The difference now is that your evidence already matches your goals, because you chose it when you wrote them.
At the review itself, state the result without inflating it. Give the baseline period, the measure and what you did. Also separate what you did from what the market or your team did. “Renewals went up 20%” invites the question of how much of that was you. “My renewals went up 20% while the team’s went up 12% over the same period” answers it before anyone asks.
A goal that changed, was missed or got overtaken needs the same care. Say what changed and when you flagged it. After that, say what you did instead and point to the note that shows you raised it in time. A missed goal you raised in month three reads very differently from one that first comes up at the review.
How do you get your manager to agree the goals?
The point of this conversation is to agree the measure before the work starts. If you skip it, the review can turn into an argument about what counted. You’re not asking your manager to approve your wording. You’re asking them to agree what they’ll look at when they judge each goal.
Bring your draft goals with their baselines and the evidence you plan to keep for each one. If your company doesn’t have its own form, a free SMART goals template can hold all of it in one place. It also helps to show how your goals connect to your team’s. The guide on how to align goals covers that part.
Some managers will hand you different goals. That’s normal, and it’s often because they can see priorities you can’t. So don’t argue about the wording. Ask what they’ll look at to decide whether each goal was met. After the meeting, write the measures down and send them back so you both have the same version.
Other managers don’t really engage and just nod at your draft before moving on. If that happens, send a short written summary of what you understood and ask them to correct anything that’s wrong. Silence isn’t agreement, but a dated note that nobody corrected is still a record.
Sometimes goals are set above your manager, and neither of you can change them. In that case the useful question is which part of each goal is yours. Agree that in writing, so the review looks at your share and not just the whole number.
This conversation can matter beyond the review. A 2015 Gallup article reported a 2014 survey of employees in Germany. Among those who strongly agreed their manager helps them set work priorities, 38% were engaged. Only 4% of those who disagreed were engaged. That finding is about priorities rather than goals. It also comes from one country and is now several years old. Still, it suggests that half an hour spent agreeing what matters most is time well used.
Where do review goals fail?
The first problem is that a measured goal pulls effort toward the measure. The driving study from earlier cuts both ways here, because the parts without goals didn’t improve. So a support agent with a reopen goal may give more care to the tickets that get counted and less to the work nobody tracks, like helping a new colleague. Nobody has to cheat for this to happen. The goal makes some work visible and leaves the rest unseen, and over a year that can shift where your time goes. If part of your job matters a lot but has no number, say so when you agree the goals.
Outcomes you don’t control are the second problem. A market can drop or a team can be reorganized, and another team can miss the deadline you depend on. Any of these can make a well-written goal look like a miss. A paired input goal helps, and so does the note you sent when things changed. But neither one erases the number at the top of the page. At the review, say plainly which part was yours. Your manager may still weigh the headline figure more than you’d like.
Priorities also change, so a goal that was right in January can be wrong by April. Renegotiating it then isn’t a failure, and keeping it unchanged just to look consistent would be worse.
There’s also a limit to the evidence behind this post. The studies cited here were about specific tasks like driving and word processing, and none of them tested annual review goals. The sections above rest on what those studies found about specific goals, feedback and commitment. They don’t prove that these eight goals will work in your company. Goals at work also run into office politics and shifting budgets, which a lab task doesn’t have. So treat the method as a sensible way to work rather than a tested recipe.
A manager who won’t agree a measure leaves you with a record but not an agreement. Your notes show what you did and when you asked. But the review may still come down to their judgment, and a record can’t make them read it your way.
Some work doesn’t fit a goal sheet at all. Support during a crisis is one example, and a first year in a new role is another. Squeezing that work into measured goals can cost more time than it saves. One learning goal and an honest note about the situation may be all that’s useful.
People-skill goals have their own weakness, because the evidence is softer. A colleague’s one-line reply or your own log is real, but it isn’t a hard number. Don’t present it as one. A reviewer who spots a soft measure dressed up as a hard one will trust the rest of your evidence less.
Even a well-written goal with clean evidence doesn’t decide your rating. Ratings also depend on your manager’s judgment and on how your company compares people across teams. Your evidence is the best case you can make, but it isn’t a guarantee.
Test your goals on a skeptical manager first
The hardest moment in agreeing goals often comes when your manager asks how they’ll know a goal was met. If you’ve never said your answer out loud, it tends to come out vague. Merlin can play your manager in that conversation. You can present your baseline and draft goals. Merlin will then ask the questions a manager might, including that one. If you’d like to see where your goal setting stands first, you can take the goal-setting assessment.
Merlin runs natively in Slack and Microsoft Teams, so you can practice where you already work. Coaching you start yourself is private. If your company formally assigns coaching to you, it sees a summary and engagement data. It never sees the details of what you said.
Practice the goal conversation with Merlin before you book the meeting with your manager.