Rules
Part of Provider comparisons: a small-business evaluation method
Provider comparisons reviews: what they can't tell you
Provider comparisons reviews can't tell you about your own connection. Learn what a review really shows and the trial that reveals your real service quality.
A review of a restaurant describes the restaurant. A review of a connection describes one building, on one segment, at one time, under one set of expectations. The reviewer's experience and yours can both be accurate and have nothing to do with each other.
This page is about what to do with that. Provider comparisons reviews are where most buyers start, so it is worth knowing what they can and cannot show. The shortlist itself gets built in provider comparisons.
What to take away
- Service quality is a property of your address and your segment. A star rating aggregates addresses, which is the one thing that must not be aggregated.
- Reviews are useful for the parts of a provider that are the same everywhere: billing, contract behavior, and how support responds.
- The replacement for a review is a trial with a written log. Nobody else's experience substitutes for two weeks of yours.
Why the ratings are structurally unreliable
Four effects, all of them working in the same direction:
- Who writes.People write after an outage. Satisfied customers rarely post about a line that simply worked for three years, so the sample is weighted toward failures by construction.
- What is being rated.A single rating merges the physical path, the contract, the billing and the support desk. Those are run by different people and can be independently good or bad.
- Where the reviewer is.A congested segment in one town and an empty one in the next are the same product and a different service.
- When.A provider that was overloaded two years ago may have rebuilt the segment since. The review does not age out of the average. The FCC's household broadband guide is a better starting point than any rating, because it describes requirements rather than experiences.
None of this makes reviews worthless. It makes them evidence about the company and not about the connection.
Where the reviews you're reading come from
Three different things get called a review, and they answer different questions. Availability aggregators such as BroadbandNow and HighSpeedInternet.com show which providers list service at a given address; that is a coverage claim, not a quality measurement. Community forums such as DSLReports collect long threads of real measurements and hardware, billing and troubleshooting problems. General rating platforms such as the Better Business Bureau and Trustpilot collect company-level scores, which is the category this article says is worth reading.
None of the three can see your segment, and none will tell you how your line behaves at five in the afternoon on a Tuesday. The rule is the same for all of them: believe what they say about the company, not about the line.
What a review can actually tell you
What a review can tell you
Because it is the same for everyone
- Billing behavior, price changes, renewal practice
- The contract is standard
- Whether support answers, and who answers
- The support desk is central
- How a cancellation goes
- Process, not infrastructure
- Whether commitments are honored
- Policy, not geography
Ignore reviews for
- Billing behavior, price changes, renewal practice
- Speed at your address
- Whether support answers, and who answers
- Whether a fault gets fixed in your street
- How a cancellation goes
- Reliability in your building
- Whether commitments are honored
- The quality of the local segment
Reviews of a cellular service carry an extra problem: the reviewer's coverage
What Reviews Can and Cannot Tell
Read reviews for
- Billing behavior
- Contract is standard
- Support desk
- Central process
- Cancellation
- Process, not infrastructure
- Commitments honored
- Policy, not geography
- Speed at address
- Not relevant
- Fault fix in street
- Not relevant
- Reliability in building
- Not relevant
Ignore reviews for
- Billing behavior
- Not relevant
- Support desk
- Not relevant
- Cancellation
- Not relevant
- Commitments honored
- Not relevant
- Speed at address
- Local segment
- Fault fix in street
- Local infrastructure
- Reliability in building
- Building-specific
Read the one-star reviews for the process complaints and the five-star ones for nothing at all. A complaint that says "three weeks to send an engineer" is about the company. A complaint that says "never gets the speed" is about a street you do not live on.
The trial that replaces the review
You cannot borrow somebody else's measurement. You can make your own cheaply.
Steps to Test Your Connection
- Measure at busy hour repeatedly
- Measure upload direction separately
- Log interruptions, not averages
- Do it wired
- Keep the log
- Measure at the busy hour, repeatedly.A single test at ten in the morning tells you almost nothing. The FCC's own measurement program, Measuring Broadband America, exists because a one-off test is not a measurement, and its method is a reasonable model for your own: same equipment, same place, over time.
- Measure the upload direction separately.Calls, backups and video meetings live there, and it is the direction that is usually narrower.
- Log the interruptions, not the averages.Four short drops a day matter more to a phone system than a lower average does. What delay and variation do to a call is in voip systems.
- Do it wired.A wireless measurement measures your wireless network.
- Keep the log.It is the evidence you need if you later have to argue that a commitment was missed.
If the log shows a commitment being missed, the FCC's consumer complaint center and your state public utility commission are the two places that take that complaint. Bring the log: dates, test results and ticket numbers are what a complaint is built from.
Asking for references instead
A reference from a same-size local business on the same product is worth more than a thousand ratings. Ask the provider for one.
Then ask the reference three questions: what broke, how long the fix took, and who they called at midnight. The answers are specific enough to be checkable, and a provider who cannot produce a local reference has told you about their local presence.
Common questions
Is there any rating I should trust?
FCC-published complaint and performance data is better than crowd ratings, because the sample is defined and the categories are consistent. It still describes a national average, not your street.
A provider has terrible reviews and is the only option. Now what?
Buy the contract terms rather than the service. Short term, clear repair commitment, a documented exit, and a second path that does not depend on them. Read the term and the notice period on the contract page itself, and get the cancellation procedure in writing, because a commitment that is not written down is not a commitment.
How long should a trial run?
Two weeks minimum, including a month end if your work has one. A month is better and rarely costs much more. What the resulting figures mean is in business internet.
Should I post my own review afterward?
If you do, write the parts that generalize: how billing behaved, how support responded, how long a repair took. Your speed figure is about your building and helps nobody else.
