# How to hire the best candidate: a practical guide for small businesses > A complete, evidence-based hiring process a small team can run without an HR department, from writing requirements to making the offer, with links to a detailed guide for each step. Sara at SortTheStack, October 9, 2026. Original: https://sortthestack.com/blog/how-to-hire-the-best-candidate Most hiring finds the person who is easiest to evaluate: the most years of experience, the recognisable employer, the most confident interview. Those are quick to judge and weak at predicting who will do the job well. **Hiring the best candidate means deliberately looking for evidence of the work instead, and comparing everyone against the same standard.** That sounds like a large-company process. It isn't. Here is the whole thing, sized for a small team with no HR department, with a link to a detailed guide for each step. ## The process at a glance 1. Define the job as evidence 2. Write the requirements, and decide the knockouts 3. Post the job and collect applications in one place 4. Screen every application against the requirements 5. Run a short phone screen 6. Ask for a short sample of the work 7. Run a structured interview 8. Score, combine and decide 9. Check references for facts, not opinions 10. Make the offer, and review your process ## 1. Define the job as evidence Before writing a posting, write down what the person will actually do in their first six months, and what someone who can do it would already have done. "Run our month-end close" leads to "has run a month-end close independently". That second sentence is something you can look for. "Detail-oriented self-starter" is not. ## 2. Write the requirements, and decide the knockouts Turn that list into three to five **must-haves** and a few **nice-to-haves**, each written as evidence you could find in an application. Describe what the person must already have, not the duties they'll perform. [Must-have vs nice-to-have](https://sortthestack.com/blog/must-have-vs-nice-to-have) has a four-question test for each requirement and a worked before-and-after. Be sceptical of year counts. Across decades of research, years of job experience correlated just .07 with later job performance (see [what actually predicts job performance](https://sortthestack.com/blog/what-predicts-job-performance)). Ask for what the years were supposed to produce. A **knockout** is a requirement that ends an application on its own: a licence the law requires, the right to work, a shift that can't move. Ask those as direct questions on the application form. ## 3. Post the job and collect applications in one place Wherever you advertise, get every application into one place: one inbox, one folder, one spreadsheet or one tool. Applications scattered across email and job board messages are how good candidates get missed. Posting rules vary by place. Ontario, for example, requires employers with 25 or more employees to include expected pay and to say whether AI is used in screening in publicly advertised postings ([Ontario Ministry of Labour](https://www.ontario.ca/document/your-guide-employment-standards-act-0/requirements-related-publicly-advertised-job)). Check what applies where you hire; our [guide to job posting disclosure rules](https://sortthestack.com/blog/job-posting-disclosure-rules) summarises the rules in several places. ## 4. Screen every application against the requirements Read every application against the same short list, mark each requirement as shown, unclear or not met, write a few words of evidence, and sort by a rule you decided in advance. Then re-read a random sample of the applications you turned down, to check the screen is working. **Detailed guide:** [how to screen a large pile of resumes without missing good candidates](https://sortthestack.com/blog/how-to-screen-a-pile-of-resumes). If the pile is too big to read carefully, see [can AI help?](https://sortthestack.com/blog/can-ai-identify-the-best-applicants) and [do you need an ATS?](https://sortthestack.com/blog/do-you-need-an-applicant-tracking-system) ## 5. Run a short phone screen A 15-minute call with each candidate on your shortlist settles what the application couldn't: anything you marked unclear, practical details like availability and pay expectations, and whether they're still interested. Ask every candidate the same few questions and write down the answers. It is a small structured interview, and the same principles apply. ## 6. Ask for a short sample of the work A work sample asks candidates to do a realistic piece of the job: reconcile an account, draft a customer reply, fix a short piece of code. In the 2022 re-analysis of selection research, work samples had an estimated validity of .33, well above an unstructured interview (.19). The [U.S. Office of Personnel Management](https://www.opm.gov/policy-data-oversight/assessment-and-selection/other-assessment-methods/work-samples-and-simulations/) notes that applicants often see them as very fair, and that they suit roles where candidates should already have the skills on day one. [Skills-based hiring](https://sortthestack.com/blog/skills-based-hiring) shows how to design one, with a free planner. In our view, keep it short and close to the real work, give everyone the same task and the same time, and decide how you'll score it before anyone does it. A strong candidate with other options won't spend a weekend on your exercise. ## 7. Run a structured interview Ask every finalist the same job-related questions, in the same order, and rate each answer against examples of strong, acceptable and weak answers written in advance. In the same research, structured interviews were the strongest single predictor (.42), more than twice an unstructured interview. **Detailed guide:** [structured interviews: what the research says and how to run one](https://sortthestack.com/blog/structured-interviews), with worked questions and rating anchors. ## 8. Score, combine and decide Put every finalist on a scorecard: one row per requirement, a rating and the evidence behind it. Weight the rows equally unless you wrote down a reason beforehand. Each interviewer rates alone, then you combine the ratings by your rule, and only then discuss. Combining by a fixed rule has been found to predict performance substantially better than overall judgement. **Detailed guide and free template:** [candidate scorecard](https://sortthestack.com/blog/candidate-scorecard-template). ## 9. Check references for facts, not opinions The evidence on reference checks is thin. The 2022 re-analysis left them out of its rankings because there wasn't enough information to re-estimate their validity, so there is no current figure to compare with the methods above. Use references to confirm facts, such as dates, titles and the work the candidate described, rather than to compare candidates. ## 10. Make the offer, and review your process Move quickly once you've decided. Then, a few months after the hire, look back: did the scorecard predict what you've seen? Which requirement mattered more, or less, than you thought? That review is how the next hire gets better. [Hiring metrics](https://sortthestack.com/blog/hiring-metrics) sets out a simple way to do it, with a free calculator. Keep your hiring records, too. In the United States, [EEOC regulations](https://www.eeoc.gov/employers/recordkeeping-requirements) require covered employers to keep application and hiring records for at least a year. Other places have their own rules. ## The one-page version - Requirements written as evidence, before reading anything - Three to five must-haves; knockouts only for true yes/no requirements - Every application read against the same list; unclear kept separate from no - A random sample of rejections re-read - Same phone screen questions for everyone - A short, realistic work sample, scored the same way for everyone - A structured interview with rating anchors - A scorecard, rated alone, combined by rule, then discussed - References for facts - A look back after a few months None of this needs software. It needs the requirements written down first, and the discipline to hold every candidate to them. To track it, [download the hiring process checklist (PDF)](https://sortthestack.com/downloads/hiring-process-checklist.pdf), free, no sign-up. ## Where SortTheStack fits SortTheStack does step 4 for you: it reads every application against your requirements, quotes the evidence, marks what it couldn't find as unclear, and leaves every decision to you. It doesn't run interviews, tests or reference checks. [See how the screening works](https://sortthestack.com/product). ## Sources - Sackett, P. R., Zhang, C., Berry, C. M., & Lievens, F. (2022). Revisiting meta-analytic estimates of validity in personnel selection. _Journal of Applied Psychology, 107_(11), 2040–2068. [doi.org/10.1037/apl0000994](https://doi.org/10.1037/apl0000994) - Kuncel, N. R., Klieger, D. M., Connelly, B. S., & Ones, D. S. (2013). Mechanical versus clinical data combination in selection and admissions decisions: A meta-analysis. _Journal of Applied Psychology, 98_(6), 1060–1072. [doi.org/10.1037/a0034156](https://doi.org/10.1037/a0034156) - U.S. Office of Personnel Management, [Work Samples and Simulations](https://www.opm.gov/policy-data-oversight/assessment-and-selection/other-assessment-methods/work-samples-and-simulations/), accessed 9 October 2026. - Ontario Ministry of Labour, [Requirements related to publicly advertised job postings](https://www.ontario.ca/document/your-guide-employment-standards-act-0/requirements-related-publicly-advertised-job), accessed 9 October 2026. - U.S. Equal Employment Opportunity Commission, [Recordkeeping requirements](https://www.eeoc.gov/employers/recordkeeping-requirements), accessed 9 October 2026.