Skills Gap Analysis: A Template and Process for Small Teams
How to run a skills gap analysis at a small company: define the skills that matter, rate them honestly, build the gap matrix, and act on it.
What follows is the skills check I wish I had run before we hired another person into delivery because projects were shipping late. Six months and one salary later, they were still shipping late. The constraint was never headcount. One person could turn a client request into a written scope, and everything queued behind her whether we had four people or five.
I had solved a capability problem by buying headcount, which is the most expensive way to be wrong about your own team. Nobody had ever written down what the role actually required, so nobody could see that the queue formed at a single skill held by a single person.
A skills gap analysis sounds like something that needs a competency framework and a consultant. In a small company it needs a grid, a four-point scale, an afternoon, and the willingness to write an honest number next to somebody you like. This is how I run it, including the parts that go wrong.
I build the people and records tooling for businesses without an HR department at FirstHR. General guidance, not legal advice.
What a Skills Gap Analysis Is and What It Is For
A skills gap analysis compares the skill level a role requires against the level the people in it currently have, and turns each difference into a decision. The output is a grid with a number in every cell, not a report, and the numbers only matter because of the actions attached to them.
It answers three questions founders otherwise answer by instinct. Where is work slower or worse than it should be because of capability rather than effort? Which capabilities sit inside exactly one head, so a resignation or a two-week vacation stops something the business needs? And what will you need to do in twelve months that nobody can do today?
The third question is the one small companies skip. A grid with a second required column for where the business is going becomes a piece of workforce planning that tells you what to start teaching now.
It is worth separating the analysis from two adjacent exercises. Individual measurement methods, including work samples and tests, belong to skills assessment. Deciding what training to deliver belongs to a training needs assessment. The gap analysis sits between them and stays open about whether training is the answer at all.
Defining the Skills That Actually Matter for a Role
Start from the work the role produces, not from a list of skills you found somewhere. A skill belongs on your grid only if you can name a specific piece of work that proves it and you would notice within a month if it disappeared.
Generic lists fail for a specific reason: you cannot rate them. Communication, attention to detail, and problem solving are real qualities and completely unratable, because two managers rating the same person produce different numbers and neither can say why. Rewrite each as a task and the rating becomes almost mechanical.
| Generic skill you cannot rate | The version you can rate | What makes the second one work |
|---|---|---|
| Communication | Handles a client complaint through to resolution without escalating it to you | There is a countable event and a visible outcome |
| Attention to detail | Runs the pre-delivery data check and catches errors before the client does | You can look at the last ten deliveries and count |
| Excel skills | Builds the monthly margin report from raw exports, unaided | Names the actual artifact the role produces |
| Leadership | Runs the Monday planning meeting so the team leaves with owners and dates | Observable in a recurring event on the calendar |
| Product knowledge | Quotes a non-standard configuration correctly without checking with anyone | Failure is visible: the quote is wrong or it is not |
| Being organized | Closes the month within three working days with no open items | Has a deadline and a definition of finished |
Cap the list at six to ten skills per role. My first attempt had twenty-two skills for one role, and by the fourth person I was rating from fatigue rather than evidence. Six skills that describe the work honestly beat twenty that describe the job description.
Three sources fill the list quickly. Name the capabilities behind the work the role actually delivered last quarter. Look at what gets escalated to you, because every escalation is a capability gap announcing itself. And check what the company turns down, since work you cannot take is a capability you do not have.
For a formal version, the US Office of Personnel Management publishes competency definitions and proficiency guidance used for federal roles. It was built for a very different scale, but the discipline is the useful part: a competency is defined by observable behavior at each level, not by an adjective.
Use your job descriptions as raw material rather than as the answer. They usually describe duties rather than capabilities, and a duty list never tells you what a person must be able to do to carry them out.
Setting the Required Level Before You Rate Anyone
Decide what the role requires before you look at a single person. If you rate people first and set the bar afterwards, you will describe your current team and call it a standard, which guarantees the grid finds almost nothing.
The required level is a business decision, not a wish. Ask what happens if this skill sits one level lower: does a customer notice, does something get redone, does it land on your desk? If not, the required level is lower than you first wrote, and you just saved yourself a training plan.
Not every skill needs the top level of the zero to three scale. Set most tasks at level 2, meaning the person does it alone to standard without review. Level 3 belongs to a few skills per role, usually the ones where somebody has to teach the next person or notice when the method itself is wrong.
Then set a second number most templates leave out: how many people need to reach that level. This is coverage, and it is where small companies find their real exposure.
A skill where exactly one person meets the bar is a risk however well that person performs, which is the logic behind cross-training and any honest succession planning. My rule is two people at the bar for anything that stops work or money within a week.
Assessing Current Level Without Fooling Yourself
Use three inputs in a fixed order of weight: demonstrated work first, manager assessment second, self-assessment last and never alone. The rating is a claim about what somebody has actually done, so it should be built from what somebody has actually done.
Demonstrated work means recent, real output: the last five scopes the person wrote, the reports that shipped without correction, the escalations that were handled and the ones that came to you. This is the strongest input and it is sitting in your systems already, which is why the assessment is far easier at a small company than the enterprise literature suggests.
Manager assessment fills in what output cannot show, mainly consistency and independence. The question is not whether the person can do it, but whether they do it without a second person answering questions. That distinction is exactly the line between level 1 and level 2.
Self-assessment is the input people reach for first because it scales, and it is the one that should carry the least weight. Research on self-awareness reported by Harvard Business Review in October 2018 found that while 95 percent of people think they are self-aware, only 10 to 15 percent actually are. Skill ratings inherit that gap directly.
The error runs in both directions, which is why no blanket adjustment fixes it. People early in a skill rate themselves high, because they cannot yet see what good looks like. Careful, highly skilled people rate themselves low, because they compare against an expert rather than the requirement. A grid built only on self-ratings ends up inverted in places.
What self-assessment is genuinely good for is different information: what somebody wants to learn and which parts of their job they quietly dread. Use it when you write the plan, not when you set the number.
Every rating gets one line of evidence. Not a paragraph. “Quoted eleven non-standard configurations last quarter, two needed correcting” is a rating. “Seems solid on quoting” is an impression, and impressions are what make this exercise argue with itself six months later.
Nick Anisimov, FirstHR founderKeep the ratings separate from your performance review cycle. The moment people believe a skill rating feeds their raise, the honest answers stop, and a grid built from defensive answers is worse than no grid. Use a regular one-on-one to check ratings, and say out loud that this measures the role, not the person’s standing.
Scoring: Four Points, Not Ten
Use a zero to three scale where every level maps to something observable. Zero is “cannot do it,” one is “can do it with help,” two is “can do it alone,” and three is “can teach it and improve the method.” Subtract current from required to get the gap, and put only positive gaps on the plan.
Ten-point scales fail in a specific way. Everybody lands between six and eight, the differences are undefendable, and the first time somebody asks why they got a seven instead of an eight you discover that you cannot answer. Four levels force a real judgment: either the person does it alone or somebody helps.
Level 3 is the level people under-record, and it is the most commercially useful cell in the grid. Every person at level 3 is a teacher already on the payroll, which converts a training budget question into a scheduling question. That is the mechanism behind the training a small business can realistically run: structured on-the-job training from somebody who already does the task well.
The Gap Matrix Template
Copy the structure below into a spreadsheet. Skills go down the side written as tasks, the required level sits in the second column, one column holds each person, and the last two columns hold coverage and the chosen response with its owner. That is the whole template.
| Skill (written as a task) | Required | Person 1 | Person 2 | Person 3 | Meets required | Response and owner |
|---|---|---|---|---|---|---|
| What goes in this column | 0 to 3, set before rating | 0 to 3, plus evidence | 0 to 3, plus evidence | 0 to 3, plus evidence | Count at or above required | Train, hire, contract, redesign |
| Scope and price a new project alone | 3 | 3 | 1 | 2 | 1 of 3 | Train Person 3, owner Person 1 |
| Run month-end invoicing end to end | 2 | 2 | 0 | 0 | 1 of 3 | Train Person 2, owner Person 1 |
| Build the monthly margin report from raw data | 3 | 1 | 3 | 1 | 1 of 3 | Redesign, move report to Person 2 |
| File quarterly sales tax returns | 3 | 0 | 0 | 0 | 0 of 3 | Contract, owner you |
| Handle a client escalation to resolution | 2 | 3 | 2 | 2 | 3 of 3 | None, covered |
| Configure a new client in the system | 2 | 2 | 1 | 1 | 1 of 3 | Write the procedure, owner Person 1 |
Read it two ways and the findings appear without analysis. Down a person’s column is their development plan for the year, in priority order. Across a skill row is the organizational picture: a row of zeros with one three on it is concentration risk, and a row where everybody sits one level below the requirement is almost never five individual skill problems.
That last pattern deserves a rule of its own. When every person fails the same skill in the same way, the cause is usually a missing standard rather than missing ability, and the fix is one written procedure rather than five training plans. Writing the thing down as a standard operating procedure closes a surprising share of what presents as a skills gap.
Keep the evidence somewhere. A note on the cell or a second tab is enough. If you drop it, the grid degrades into opinion within one cycle, and the second run becomes an argument instead of an update.
This is that second tab. One line per rating, written while you are rating rather than afterwards, naming the piece of work and where the claim came from. The last column is the one that turns the log into next quarter's plan, because writing down what would move a rating up forces you to say what level 2 actually looks like on that task.
| A | B | C | D | E | F | G | H | I | |
|---|---|---|---|---|---|---|---|---|---|
| 1 | Person | Skill, written as a task | Required level | Rating given | Evidence: the actual piece of work | Source | Date rated | Re-rate on | What would move this up one level |
| 2 | Replace this row | Quote a non-standard configuration without checking with anyone | 2 | 2 | Quoted eleven non-standard configurations last quarter, two needed correcting | Demonstrated work | One quarter of quotes with no corrections needed | ||
| 3 | |||||||||
| 4 | |||||||||
| 5 | |||||||||
| 6 | |||||||||
| 7 | |||||||||
| 8 | |||||||||
| 9 | |||||||||
| 10 | |||||||||
| 11 | |||||||||
| 12 | |||||||||
| 13 |
From Role Grids to a Workforce Skills Gap Analysis
A workforce skills gap analysis is the same grid run across every role and then read as one sheet. You get there by repeating the role-level pass rather than by starting bigger: do two or three roles, act on them, and add the next roles at the following cycle until the company is covered.
Three things have to hold for the sheets to add up. Skills written the same way, so one task does not appear as three differently worded rows. One scale, applied identically. And a single owner of the vocabulary, usually you, because as soon as two managers invent parallel skill names the roll-up stops comparing anything.
Then, if your role grids do not carry it yet, add the forward-looking column: the level each skill will require twelve months out. Comparing that against the required level you set for today is where this becomes workforce planning, because it names the capabilities to start teaching before the work that needs them arrives.
Read the combined sheet by skill rather than by person. Coverage counted across the whole company, not one team, is what tells you whether a capability is genuinely two people deep or whether both of them sit on the same team and are out in the same week.
A Worked Example for a Small Team
Here is the grid filled in for the four-person delivery team at a 26-person services firm that sells to other businesses. Required levels were set first, in a separate sitting, before anybody was rated. Scores use the zero to three scale above.
| Skill | Required | Priya | Marcus | Dana | Luis | Meets required |
|---|---|---|---|---|---|---|
| Scope and price a new project alone | 3 | 3 | 1 | 2 | 1 | 1 of 4 |
| Run a client status call without escalating | 2 | 3 | 2 | 2 | 2 | 4 of 4 |
| Build the reporting in the analytics tool | 3 | 1 | 3 | 1 | 2 | 1 of 4 |
| Run the pre-delivery data quality check | 3 | 2 | 3 | 2 | 2 | 1 of 4 |
| Set up a new client account end to end | 2 | 2 | 2 | 3 | 1 | 3 of 4 |
| Handle contract and renewal paperwork | 2 | 3 | 0 | 0 | 0 | 1 of 4 |
The bottom row is the finding that pays for the whole exercise. Only Priya can process a renewal, and renewals are how this business gets paid. That makes this row a concentration risk on incoming revenue rather than a training need in the ordinary sense, and it is cheap to close because the expert is sitting right there.
The first row is the constraint on growth. One person meets the bar on scoping, so the firm can only sell as fast as Priya can write scopes, which is exactly the trap that made me hire a fifth person instead of teaching the fourth. Dana at level 2 is one level away, so the response is Priya walking Dana through the next four scopes rather than a job posting.
Row four is the pattern that looks like a people problem and is not. Three of four people sit at level 2 against a required 3 on the data quality check. When the same gap appears across nearly the whole team, the cause is usually that nobody wrote down what the check is. One page fixes it, and no training budget is involved.
Row three is the mirror image of row six: Marcus is the only person at the required level on the technical build. That makes two single points of failure on the same four-person team, and neither was visible from performance conversations, because both people are performing well.
Row two is worth noticing because nothing is wrong with it. Coverage is complete, and the value of the row is that it stops you spending attention there.
Common Examples of Skill Gaps in the Workplace
The gaps I run into most often at small companies are variations on a short list: a task only one person can do, a manager promoted for output rather than for managing people, a tool the team uses at a fraction of its capability, and work you sell but have never written down. Here is what they look like once they are on a grid.
| Example of a skill gap | How it shows up in the work | The response that usually fits |
|---|---|---|
| Only one person can price non-standard work | Quotes wait on one calendar, and the discount depends on who wrote them | Train, with the expert walking two people through live quotes |
| A first-time manager cannot run a difficult conversation | Performance problems arrive on your desk instead of being raised with the person | Train, paired with a written script for the first few conversations |
| Nobody can build the report the business runs on | The number lands late, assembled by hand, and does not match the previous one | Redesign, by moving the report to whoever is at teaching level |
| Quarterly tax filings belong to nobody | The deadline is met by panic, or missed | Contract, because the need is intermittent and the rules move |
| The team uses the system but not the parts that save time | Data gets entered twice and half the features sit unused | Train, usually an hour of setup rather than a course |
| The pre-delivery check lives in one head | Errors reach the client and each person fixes them differently | Write the procedure, because the whole team fails the same way |
Notice what is absent from those examples: adjectives. Each one names a task and a visible failure, which is what makes it something you can rate rather than something you complain about. If you cannot say when the gap last cost you money or time, park it and work on the ones you can.
Train, Hire, Contract, or Redesign the Role
Every gap that matters gets exactly one of four responses, chosen deliberately rather than by default. In my experience, most small business gaps are one level wide with a teacher already on the payroll, which makes training the right answer far more often than founders assume.
Hiring is the option that feels fastest and rarely is. It carries a search, a notice period, a ramp, and a permanent cost against a gap that may be temporary, and the market is not obliging. Before you post the role, check whether the same money spent on two months of a contractor plus internal teaching closes the gap permanently.
Contracting is underused at small scale. Quarterly tax filings, a one-off technical build, a specialist review: these are gaps you can buy your way out of this month without adding a permanent role, and a contractor can also cover the months while somebody trains into the work. Classification rules matter, so read the boundary between an employee and an independent contractor first.
Redesigning the role is the option nobody lists and the one I have used most. If the same gap shows up on every person you put in a seat, the seat is the problem: you combined three unrelated skills into one job that no realistic candidate holds. Splitting it or moving one task elsewhere is cheaper than a third failed hire.
When the gap is real and the hire is right, hire for the capability rather than the resume. That is the practical argument for skills-based hiring, and a gap analysis gives you exactly the input it needs: a specific level on a specific task, which is far easier to test for than a job title.
Turning the Matrix Into a Plan With Owners and Dates
Pick three gaps, not nine. A grid becomes a plan only when each action has a named owner, a date, and a defined piece of evidence that closes it. Everything else is a description of the world.
Two things make the plan hold: the actions are mostly internal, which removes budget approval as a reason for delay, and people can see their own column.
That visibility has a payoff. Gallup research published in October 2025, from a survey of 15,968 US employees, found 57 percent had participated in on-the-job training and 45 percent had received training to develop new skills. Yet one in four reported lacking opportunities for career advancement (Gallup). A plan of three dated actions is a cheap way to be on the right side of that number.
Each person’s column, read top to bottom, is most of an employee development plan already written. Add what they told you they wanted to learn during the self-assessment, and you have something worth putting in front of them.
Where This Goes Wrong at Small Scale
The failure I see most often is scope: running the analysis on every role at once, producing a grid nobody finishes and a list of gaps nobody acts on. Two or three constrained roles produce a usable answer in an afternoon.
The second failure is rating without evidence. It feels faster and it costs you the second run, because a grid built from impressions cannot be defended, updated, or discussed with the person it describes. One line per cell is the whole discipline.
The third is treating every gap as a training need. Some gaps are missing procedures, some are missing tools, and some are people doing work the role should never have contained. Running training on a broken process produces a team that executes it more confidently.
The fourth is confusing a low score with a bad employee. A level 1 on a skill the person has never been taught is a finding about your onboarding, not about them. This is exactly why a new hire’s column looks alarming and means almost nothing: in month two, gaps across the board add up to an upskilling schedule, not a performance signal.
The fifth is the quiet one: running the analysis, finding real gaps, and doing nothing. That is not neutral, because you have told your team you know exactly where they are weak and then chosen not to help.
The sixth is building the grid around the people you have rather than the work you sell. Required levels reverse-engineered from your current team come back clean every time, right up until the work changes and nobody can do it.
My advice for a first run: choose two or three roles where work queues today, rate each one in a single sitting, and leave with three dated actions and their owners.
Frequently Asked Questions
What is a skills gap analysis?
A skills gap analysis is a side-by-side comparison of the level of skill each role calls for and the level the people in that role have today. You end up with a scored grid rather than a written report, and each gap worth acting on gets a decision beside it. The exercise settles three things. It shows where capability, not effort, is making work slower or sloppier than it should be. It exposes skills that only one person on the team holds. And it flags work the business will need a year from now that no one can yet handle. Limited to two or three roles instead of the whole company, it takes roughly four hours and needs no software beyond a spreadsheet.
How do you conduct a skills gap analysis step by step?
Seven steps. First, start narrow: pick the two or three roles in which work piles up or the same mistakes keep coming back, not the whole company. Second, write between six and ten skills for each of those roles, phrasing every one as a task you could point to in real work. Third, decide the level each skill requires before you rate anybody, so the people you happen to have do not quietly become the benchmark. Fourth, score each person, trusting proof from real work over a manager’s view, and a manager’s view over the person’s own rating. Fifth, work out each person’s gap and, for every skill, how many people already meet the bar. Sixth, pick train, hire, contract, or redesign for each gap worth closing. Seventh, commit to three actions, each with someone responsible, a deadline, and the specific proof that will show it is done.
What should a skills gap analysis template include?
Seven columns when you rate three people, and no more. The first names the skill as a task rather than a topic, and the second holds the required level, scored from zero to three. The next three hold each person’s current level, one column per person, so every extra person you rate adds a column. A coverage column then counts how many people already meet the requirement, and the final column holds the response, one of four words (train, hire, contract, or redesign), with the name of whoever owns it. A longer template tends to be abandoned halfway, and a short grid that gets finished beats a long one that does not. Beyond those columns, keep the evidence behind every rating, because it is what turns an impression into an assessment. If the grid gets crowded, move that evidence into a cell comment or onto its own tab. Due dates go on the three-action plan you build from the grid.
What is the difference between a skills gap analysis and a training needs assessment?
A skills gap analysis measures how far current capability sits from what a role requires, without assuming why the gap exists or how to fix it. A training needs assessment begins from the assumption that training is the fix, then works out which training and who should get it. That makes the gap analysis the broader tool, because it can end in any of four answers: teach the skill, hire someone who has it, buy it from a contractor, or reshape the role so the seat no longer needs it. In practice the analysis usually feeds a training plan, since in my experience the typical small-company gap is a single level deep and someone who can teach it already works for you. Even so, assuming every gap calls for training leads to courses nobody needed.
Is self-assessment reliable for rating employee skills?
Not on its own. In a research program that Harvard Business Review described in October 2018, 95 percent of people thought they were self-aware, but only 10 to 15 percent really were, and that blind spot carries straight into how people score their own skills. It skews both ways. Someone new to a skill often overrates it, since they cannot yet recognize good work when they see it, while a highly capable person often underrates it by measuring against an expert instead of against what the role needs. Either error distorts the grid. Treat the self-rating as the lightest of your three inputs, behind proof from real work and the manager’s view, and use it mainly to learn what the person hopes to develop and where they feel unsupported.
What scoring scale should you use for a skills gap analysis?
Use four points, numbered zero to three. A zero means the person cannot do the task, a one means they manage it only with help, a two means they handle it alone to standard, and a three means they can teach it to others and make the method better. Four levels work because each one describes something you can watch happen, not a feeling about the person. A ten-point scale invites fake precision: ratings bunch up between six and eight, and nobody can explain why someone earned a seven rather than an eight. To get the gap, subtract the current score from the required one; only a result above zero belongs on the plan. A side benefit is that every three marks a person who can train others, which is where most of your teaching capacity already sits.
Should you train or hire to close a skills gap?
Train when the gap is a single level, the person has already shown the underlying ability, and a colleague already at teaching level can coach them. Hire when no one in the company has the skill at all, the work is steady and full time, and nobody on the team is within a single level of what the role requires. Expect hiring to take longer than your plan allows: in SHRM research fielded in February 2026, 68 percent of HR professionals reported trouble filling full-time positions, while 41 percent said they train current employees for roles that are hard to fill. Two further options get overlooked. A contractor covers an occasional or specialist need without adding a permanent cost, and redesigning the role is the honest fix when every person you put in the seat ends up with the same gap.
How often should a small business run a skills gap analysis?
A full pass once a year, and a partial re-rating whenever something structural changes. Once every role has its own grid, the full pass covers all of them and takes most of a day at a small company. The partial pass covers only the affected cells and is triggered by a specific event: somebody resigns, you adopt a new system, you win a type of work you have not done before, or you promote someone into a role they have not held. Between passes, do not re-rate the whole grid. Re-rate the cells you built a plan against, using the evidence you defined when you wrote the action. Rebuild the entire grid every quarter and the exercise becomes administration, which is how it quietly stops happening.
Make HR the easy part of running your business
Onboarding, e-signatures, training and core HR in one place. Up and running in 5 minutes.
- 7-day free trial
- No credit card required
- Cancel anytime