We Analyzed 830 Keywords From 48 Job Descriptions. Here's What Resumes Actually Get Wrong.

Most resume keyword advice starts from the wrong end. It tells you which words to add to your resume without first asking a more basic question: what are employers actually asking for?

We wanted to answer that question with data instead of assumptions. So we extracted every keyword from 48 real job descriptions that came through Ajusta's production pipeline, then matched those keywords against 22 real resumes to see where the gaps actually are. The dataset produced 830 unique keywords and 1,482 total keyword mentions across job titles ranging from Data Scientist to CNC Operator to Investment Banking Analyst.

What we found challenged some things we expected. The biggest keyword gap in resumes is not technical skills or programming languages. The terms most consistently missing from resumes are words like "collaboration," "teamwork," and "agile development." And 71% of the keywords employers use appear in only a single job posting, which means generic keyword lists are even less useful than most people realize.

The long tail problem: 71% of keywords are unique to one job

Career advice sites love publishing lists of "top ATS keywords" or "power words that get you hired." The implicit promise is that there is a universal set of keywords that works across jobs. Our data suggests otherwise.

Of the 830 unique keywords we extracted from 48 job descriptions, 590 of them (71.1%) appeared in only a single posting. Another 178 appeared in just 2-3 postings. Only 40 keywords (4.8%) showed up in 5 or more job descriptions. And only 12 (1.4%) appeared in 10 or more.

The practical implication is straightforward. If you are working from a generic keyword list, roughly 7 out of every 10 keywords in any given job description will not be on that list. They are specific to that particular role, at that particular company, in that particular industry. A list of "top 50 resume keywords" might cover 5-10% of what a specific posting actually asks for.

This does not mean cross-cutting keywords are worthless. But it does mean that the single most effective keyword strategy is also the most obvious one: read the actual job description you are applying to and match its specific language. There is no shortcut that replaces this.

What employers actually ask for (it is not what you think)

Despite the long tail, some keywords do appear across many different job descriptions. These cross-cutting keywords tell us something about what employers broadly value, regardless of industry or role.

Here is what surprised us: the most common keyword across 48 job descriptions was not Python. It was not machine learning or data science. It was "communication." Over half of all job postings explicitly listed it.

Of the top 15 most common keywords, 7 are soft skills. Only 4 are technical. The remaining 4 are domain terms like "data science" and "engineering" that describe fields rather than specific tools.

When we looked at the full dataset by category, the pattern held: domain-specific terms accounted for 78.3% of all keyword mentions, technical skills 12.9%, and soft skills 8.8%. But domain keywords are overwhelmingly unique to individual postings (they are the long tail). The keywords that actually cross industry and role boundaries are disproportionately soft skills.

Each job description contained an average of 30.9 keywords (range: 11 to 59, median: 30). More specialized roles tended to have more keywords. A generic entry-level posting might list 15 requirements. A senior data scientist role might list 50+.

The gap nobody talks about

Here is the finding that reframed how we think about resume optimization. When we scored 22 real resumes against their target job descriptions and categorized every keyword as either "found" or "missing," the results split cleanly along category lines.

Technical keywords had a 100% match rate. Every single technical term that a job description asked for was already present in the resume. Machine learning (found in 7 baselines), large language models (5), data analysis (5), generative AI (4), prompt engineering (3). People who work in technical fields already write their resumes using technical language. This is not where the problem is.

Soft skills had a 19% match rate. Out of 21 soft skill keyword instances across all baselines, only 4 were found. The single most commonly missing keyword in our entire dataset was "collaboration," which appeared in 46% of job descriptions but was missing from every single baseline resume we analyzed. Not most of them. All of them.

Domain-specific keywords fell in between at 35%. These are terms like "clinical research," "blueprint reading," or "payment" that are highly specific to a particular job or industry. The low match rate here is expected. When someone applies to a role outside their exact previous experience, domain terminology is naturally the first thing to diverge.

The pattern this reveals is worth stating plainly: resume writers overindex on the category where they already perform well (technical skills) and underindex on the categories where they actually lose points (soft skills and domain-specific terms). The conventional advice to "add more technical keywords" is solving the wrong problem for most people.

Why match rate matters more than keyword count

A common piece of advice is to "include as many keywords as possible." But our data shows that the percentage of job description keywords you match matters more than the raw number you include. The reason is that ATS scoring is relative to the job, not absolute.

Across 22 baseline resumes, the average keyword match rate was 44% (median: 44%). That means the typical resume, before any optimization, matched fewer than half of the keywords the job description listed. Match rates ranged from 12% to 70%.

The correlation between keyword match rate and overall ATS score in our data was 0.970. That is extremely strong. In our previous analysis of resume statistics, we found that keywords were the primary deficit driver in 100% of low-scoring resumes and accounted for 52.7% of total lost points. This dataset confirms that finding from a different angle: the percentage of job-specific keywords you cover is nearly a direct predictor of your overall score.

After optimization, the average keyword match rate jumped from 44% to 89%. The overall score rose from a mean of 46 to 74.

The improvement was not evenly distributed across score components. Keywords improved by an average of 61 raw points (from 29 to 90 on the 0-100 subscale). Skills improved by 20 points. Experience, education, and contextual fit barely moved, because they were already high. This matches what we described in our breakdown of how ATS scores are calculated: the keyword component carries 40% of the total weight, and it is almost always the component with the most room for improvement.

Keywords and skills: a related gap

Keywords and skills are scored as separate components in our system (keywords at 40% weight, skills at 25%), but the two are related. When we looked at the skills component specifically, the same soft skill gap appeared.

Across 26 optimization plans, the system flagged 87 missing hard skill instances and 86 missing soft skill instances. Almost exactly equal. But the nature of what was missing differed. Missing hard skills were highly fragmented (80 unique terms, most appearing only once). Missing soft skills were concentrated around a small set of terms that appeared repeatedly.

Collaboration, communication, and entrepreneurial spirit were each missing 7 times across 26 plans. Compare that with the matched skills: the most commonly matched skills were technical terms like "data analysis of structured and unstructured data" (4x), "Python" (4x), and "collaboration with product and engineering teams" (4x).

The pattern reinforces what the keyword data showed. People are already strong on technical language. The gap is in the softer, more process-oriented terms that job descriptions consistently ask for but that resume writers consistently leave out. Not because they lack these skills, but because they do not think to write them down.

What this means for your resume

We did not start this analysis expecting to write an article about soft skills. But the data led here, and we think the practical implications are worth spelling out.