Candidates read employer reviews before they apply, and many read them before they accept an offer. A falling rating, a cluster of reviews about one manager, or an interview process described as chaotic will cost you applicants long before it shows up in any internal survey.
Employer brand teams therefore want the same thing product teams want from customer reviews: to know what is being said, where, and whether it is getting better or worse, for their own company and for the competitors they hire against. This tutorial covers how to do that across review sites, and where the lines are.
What to monitor
- Overall rating and how it moves over time
- Sub-ratings such as culture, leadership, compensation and work-life balance
- Review volume, since a rating built on a handful of reviews moves a lot
- Themes in review text: management, pay, workload, career growth, layoffs
- Interview reviews, which describe the candidate experience directly
- Employer responses, and how quickly they come
Monitor them per site, per country and per company. The same employer can have a very different reputation on a regional site than on the largest global ones.
Start with the data you are entitled to
Before collecting anything, use what the platforms give employers directly. The major employer review sites, Glassdoor among them, offer employer accounts or tools that let a company see and respond to reviews of itself, often with analytics. For your own company, that is the cleanest source: complete, permitted and maintained by the platform.
Public collection is mainly useful for two things: benchmarking against competitors you hire against, and tracking sites where you have no employer tools.
Know what is accessible, and what is not
Employer review sites protect their content more tightly than most. Many limit how much can be viewed without an account, and their terms often prohibit automated collection.
That shapes what a responsible programme looks like:
- Collect only what is publicly visible without logging in. Do not create accounts or use credentials to reach gated reviews.
- Respect each site’s terms. Where a site prohibits automated collection, use its employer tools, its official partner programmes, or licensed data instead.
- Keep volumes proportionate. Review pages change slowly; there is no reason to request them often.
If a question can only be answered by crossing those lines, the answer is to change the question, or to buy the data from someone licensed to provide it.
Collecting what is public
For sites and pages where public collection is appropriate, a few practices make it reliable and light.
Collect per locale. Review sites often run separate country editions with their own reviews and ratings, and what you see depends on where the request appears to come from. Collect each market from that market.
Walk newest-first and stop early. Sort by date where the site allows it, and stop when you reach a review you have already stored. Incremental collection is faster, cheaper and far gentler on the site than re-reading every page.
Hold a session across pagination so that page two follows from page one.
With the Shifter gateway, the market and session go in the credentials against p.shifter.io:443:
customer-USERNAME-country-gb-sid-reviews-acme-03-ttl-600:PASSWORDThe trade-off between holding and rotating exits is covered in sticky vs rotating residential proxies, and request pacing in rate limiting and request throttling.
Store reviews, not reviewers
Reviews are written by people, often anonymously and often by current employees who could face consequences if identified. The data model should reflect that.
| Field | Keep | Notes |
|---|---|---|
| Review ID and site | Yes | For deduplication and incremental collection |
| Company and locale | Yes | The unit of analysis |
| Ratings and sub-ratings | Yes | The core metrics |
| Review date | Yes | For trends |
| Role and location as displayed | Coarse only | Useful for themes, never for identification |
| Review text | Yes | For theme analysis |
| Employer response and date | Yes | Response rate and speed |
| Reviewer username or profile | No | Not needed, and a re-identification risk |
Never attempt to work out who wrote a review, and never let monitoring data feed any action against a suspected author. A monitoring programme that can be used to identify critics will destroy the trust it was meant to protect, and is likely unlawful in many places. Reviews contain personal data, so the same obligations apply as to any other; see residential proxies and GDPR compliance.
Turning reviews into something useful
Weight ratings by volume. A company with twelve reviews moving from 3.9 to 3.6 may be two bad weeks. Show volume next to every rating, and use rolling windows rather than point readings.
Normalise across sites. Rating scales, sub-rating categories and review prompts differ. Compare trends within a site before comparing levels across sites.
Extract themes. Topic and sentiment analysis over review text shows what is driving a rating, which is what an employer brand team can actually act on. The collection side of sentiment work, and its pitfalls, is covered in scraping social platforms for sentiment analysis.
Watch for bursts. A sudden cluster of very positive or very negative reviews in a few days often reflects an event or an organised push rather than a change in the workplace. Flag bursts rather than letting them drive trends.
Line reviews up with events. Rating dips often follow layoffs, reorganisations or leadership changes. Tracking those events alongside reviews explains the movement; see how to track layoffs and workforce changes in real time.
Benchmarking against competitors
For the companies you compete with for talent, compare trends in overall and sub-ratings, the themes that dominate their reviews, interview experience, and how actively they respond. The most useful output is usually not a league table but a small set of themes where a competitor is clearly stronger or weaker than you, which recruiters can speak to in conversations with candidates.
The general practice of monitoring what is said about a brand across the web is covered in how to track brand mentions across social, forums and review sites.
Respond through the proper channels
Monitoring is only half the job. Respond to reviews through the platform’s employer tools, consistently and without defensiveness, and route recurring themes to the teams who can fix them. A visible, thoughtful response to a critical review often does more for candidate perception than the rating itself.
FAQ
Is it legal to scrape Glassdoor reviews?
It depends on the jurisdiction and on the site’s terms, and many employer review sites prohibit automated collection and limit logged-out access. For your own company, use the employer tools the platform provides. For competitors, stay within what is publicly visible and permitted, or use licensed data, and take legal advice before building anything larger.
How often should reviews be collected?
Reviews arrive slowly. Daily is plenty for most companies, and incremental collection means each run fetches only what is new.
Can we find out who wrote a negative review?
No, and you should not try. Anonymity is the reason employees are candid, and attempting to identify reviewers creates serious legal and reputational risk.
Which sites matter most?
The largest global employer review sites, the review sections of major job boards, and regional sites in the countries where you hire. Weight them by where your candidates actually look.
The bottom line
Employer reviews are a candidate-facing signal that moves before internal metrics do. Use the employer tools platforms provide for your own company, collect only public pages within each site’s terms for competitors, walk newest-first and stop early, and store reviews rather than reviewers.
Weight ratings by volume, compare trends within sites, extract themes, and never let monitoring become a way to identify critics. The product view is on the residential proxies for recruitment page, with rates on the pricing page.