Glassdoor reviews to Power BI
Glassdoor reviews are written by the people you employ and read by the people you want to employ. That makes employer brand a People number, and People numbers belong in the model next to headcount, not in a browser tab somebody remembers to open.
Set up in under five minutes • No demo required
Topics in Cons
67 reviews- Senior leadership, mentioned in the cons of 41% of reviews in the last 90 days.
- Job security, mentioned in the cons of 33% of reviews in the last 90 days.
- Workload, mentioned in the cons of 22% of reviews in the last 90 days.
- Pay and benefits, mentioned in the cons of 14% of reviews in the last 90 days.
Written by the people you employ. Read by the people you want to hire.
Every other People number already has a home
Headcount, attrition, time to hire, open requisitions. Each of them lands in the model, gets a measure and is read on the same page every month. Employer brand is the exception, and it is the exception for a boring reason rather than a strategic one: nobody could get it into the model.
Glassdoor no longer supports API partnerships, so the honest options have been a CSV somebody exports by hand out of the Employer Center, or a scraper somebody maintains. Both of them stop the week that person is on leave, and neither of them ends up in the same report as anything else.
Reviewflowz publishes your Glassdoor reviews as an ordinary table. Relate it to the date table your People model already has and the rating sits on the same row of the same report as the month's leavers and the month's open roles.
Then the question stops being what the rating is. It starts being whether it moved with anything else you did, which is a question you can only ask once the two numbers are in the same place.
Reviews
Glassdoor reviews, from Reviewflowz
105 rows Same model, same months, so attrition and employer brand read on one page.
Anonymous by design, which is exactly why it belongs in aggregate
A Glassdoor review has no name on it. There is nobody to email, no account to look up and no case to close, which is why routing employer brand through a support queue never works. That tooling assumes each review is a thing one person actions, and this one is not a thing you can action.
It is a reading on a population instead. A single two star review in August is noise. Thirty one reviews in July against eight in May is a finding, and the only way to tell those apart is to count them, which is what a dashboard is for and what a browser tab is not.
The text arrives already split, and that is the part that makes it chartable. Glassdoor asks for pros and cons in separate boxes and takes a line of advice to management on top, so each of those lands as its own column. You are not parsing a paragraph to work out which half was the complaint.
Reviewflowz runs topic extraction over that text as reviews arrive, so a review about a manager and a review about a director land under the same topic. The share of reviews whose cons mention it becomes a measure rather than a reading exercise you do once a quarter and then stop doing.
The engineers I work with are the best I have had anywhere, and my own manager is honest about what she actually knows.
Nobody above director level has explained the plan. People find out their team is going from a calendar invite.
Say the plan out loud, including the parts you think we will not like. We are already assuming worse.
Nobody to reply to. Sixty seven of these since June.
The spike lands in the week nobody is looking
Glassdoor's own economists measured this. Review volume runs 44% higher in the week immediately after a layoff, and the effect is not a blip: their September 2025 study found layoffs cost a company 0.13 stars, that the employees who stayed dropped 0.16, and that ratings were still depressed a year later and had not fully recovered inside two.
That week is the week your People team is running the layoff. Nobody is refreshing a review site on the Thursday of a redundancy consultation, and the reviews that arrive that week are the ones a candidate reads in November.
A dashboard does not forget. Count reviews by month over the review date and the spike is on the page the following Monday whether anyone went looking or not, sitting next to the headcount change that caused it.
You cannot know in advance which month will be the hard one. That is the whole reason this belongs on a refresh schedule rather than on somebody's calendar reminder.
| Month | Reviews |
|---|---|
| January 2026 | 8 |
| February 2026 | 6 |
| March 2026 | 9 |
| April 2026 | 7 |
| May 2026 | 8 |
| June 2026 | 14 |
| July 2026 | 31 |
| August 2026 | 22 |
Glassdoor Economic Research: volume runs 44% higher the week after a layoff.
The reviews you never read are the ones not in your language
Reviewflowz pulls your Glassdoor profile across seven languages: English, French, German, Dutch, Portuguese, Spanish and Italian. Every review arrives tagged with its language code, so the German reviews are in the table whether or not anyone on the People team reads German.
This matters most at the moment it is hardest to notice. A profile page shows you a rating and the reviews in front of you. It does not tell you that the last fourteen French reviews all say the same thing, because you were never going to scroll that far.
Split any measure by language and you get a read on a population you had been averaging away. Average rating by language, review count by language, and the topic mix in the cons for each of them, side by side in one visual.
Topics are worked out from what a review means rather than from a keyword you had to guess, so a French review about la direction counts under the same topic as an English one about leadership. You are not maintaining a French word list beside an English one.
| Language | Reviews | Avg | Top topic in cons |
|---|---|---|---|
| en English | 41 | 3.6 | Job security |
| fr French | 14 | 2.9 | Senior leadership |
| de German | 8 | 3.5 | Workload |
| nl Dutch | 4 | 3.8 | Pay and benefits |
| Total | 67 | 3.4 | No single top topic across all languages |
3.4 is the number on the profile page. 2.9 is the reason.
An OData feed, not a connector somebody has to build
Reviewflowz publishes an OData v4 feed, and Power BI has read OData for years. Get data, OData feed, paste the URL, then authenticate with your token as the user name and a blank password. There is no custom connector to install, nothing to find in AppSource and no Power Query M to write.
The navigator lists the tables and you tick the ones you want: your reviews, your review profiles, your ratings history, and a change stream of inserts, updates and deletes for anyone building this properly.
After that Power BI refreshes on its own schedule. The review date is a datetimeoffset the feed filters on, which is exactly what Power Query folds a RangeStart and RangeEnd window into, so incremental refresh works instead of reloading five years of reviews every morning.
The same feed opens in Excel, Tableau or Qlik. The analyst who will not leave Excel is reading the same rows as the report, rather than a copy that drifted three weeks ago.
Navigator
OData v4
Larkfield, Glassdoor reviews
Basic rfz_bi_placeholder_0000 Power BI pulls the feed. Reviewflowz never pushes.
The trend line exists on the day you connect
The first run loads your Glassdoor profile's review history, all of it, across all seven languages. The chart is full from the first refresh instead of filling up over the next two quarters while you wait to find out whether anything is wrong.
That is the difference between a number and a judgement. A 3.4 on its own is a number. A 3.4 under four years of quarters that sat between 4.0 and 4.2 is a judgement, and you can make it in your first week rather than next spring.
It also gives the bad month something to be compared against. Whatever happened in June has the June before it and the June before that on the same axis, which is usually enough to settle the argument about whether this is a pattern.
Somebody senior will ask whether this is unusual, and they will ask before a dashboard that started empty could answer. Loading the history is how you have the answer the first time the question comes up.
Larkfield's average of the Rating column on Glassdoor reviews, by quarter, from 2022 Q1 to 2026 Q3. Larkfield connected Reviewflowz on 2 September 2026 and every quarter listed below was loaded on that first run. Only the five days from 2 September to 7 September 2026 were collected afterwards.
| Quarter | Average of Rating |
|---|---|
| 2022 Q1 | 4.25 |
| 2022 Q2 | 4.22 |
| 2022 Q3 | 4.19 |
| 2022 Q4 | 4.21 |
| 2023 Q1 | 4.15 |
| 2023 Q2 | 4.12 |
| 2023 Q3 | 4.10 |
| 2023 Q4 | 4.08 |
| 2024 Q1 | 4.11 |
| 2024 Q2 | 4.09 |
| 2024 Q3 | 4.06 |
| 2024 Q4 | 4.05 |
| 2025 Q1 | 4.02 |
| 2025 Q2 | 4.00 |
| 2025 Q3 | 3.98 |
| 2025 Q4 | 3.97 |
| 2026 Q1 | 3.90 |
| 2026 Q2 | 3.80 |
| 2026 Q3 | 3.40 |
A dashboard that starts today cannot tell you what normal was.
Put employer brand on the dashboard People already opens
Set up in under five minutes. No demo required.
Which Glassdoor fields land in Power BI?
One row per review, carrying the rating, the review date, the title, the pros, the cons, Glassdoor's advice to management, the language code, the link to the review and the topics Reviewflowz extracted from the text, alongside tables for your ratings history, your review profiles and a change stream. Glassdoor's six workplace factor ratings, culture and values, work life balance, senior management, compensation and benefits, career opportunities and diversity and inclusion, are not in the feed, so no measure here is built on them. Glassdoor calculates those separately from the overall rating on its own site in any case.
How do Glassdoor reviews actually get into Power BI?
Through an OData v4 feed that Reviewflowz publishes. You connect your Glassdoor profile, create a connection in Reviewflowz and copy its feed URL, then in Power BI choose Get data, OData feed and paste it. Authenticate with the token as the user name and a blank password. The navigator lists the tables and you load the ones you want. There is no custom connector to install and no Power Query M to write.
Can Reviewflowz post our replies to Glassdoor?
No, and no tool can. Glassdoor has no API for posting an employer response, so a response is always something a person publishes in the Glassdoor Employer Center, which needs a free Employer Account and shows publicly with the date and the job title of whoever posted it. Reviewflowz drafts the reply and links you straight there. It is worth doing: in Glassdoor's own survey, 80% of job seekers who read reviews said their view of a company improved after seeing an employer respond. One thing to know before you build a report around it, though: Glassdoor employer responses are not in the feed, so response rate is not a measure you can put on this dashboard.
Does the first load include our Glassdoor history?
Yes. Connecting your Glassdoor profile loads its review history across all seven languages Reviewflowz pulls, so the trend line is full from the first refresh rather than filling up over the next quarter. After that the feed carries what changed.
Does the report refresh on its own?
Yes. Power BI's scheduled refresh pulls from the feed on whatever cadence you set, and incremental refresh works too, because the review date is a datetimeoffset the feed can filter on. Nobody re-runs an export and nobody remembers to.
Which plan includes the BI feed?
The BI feed is a Premium feature and it is not part of the free trial, so budget for that rather than discovering it on day three. Everything else works the way the trial suggests: you connect your Glassdoor profile yourself and see your reviews in minutes.
Do I need a demo to get started?
No. You connect your Glassdoor profile and create the BI connection yourself, in under five minutes. If you get stuck, support is the people who build it.