Nine credit unions, seventy branches, and one question: when a member needs their own branch, can they actually reach it?
A member of twenty years has just been told the bank branch she has used since 2004 is closing. She picks up her phone and types three words.
Credit union near me. What happens in the next forty seconds decides where a household's deposits, its mortgage and its car loan spend the next decade. Not your rate sheet. Not your community programme. Not the fact that you are member-owned. Three results on a map, each with a name, a rating, a distance and a set of opening hours.
That is the entire competitive arena for an institution whose central claim is that it is closer to people than a bank. And it is the part of the member journey that almost no credit union has audited.
Nine institutions are in this study. They are anonymised, and the letters are deliberately not in rank order — so if you work in this industry, there is a reasonable chance one of these scorecards is yours and you will not know which until you reach the end.
Read this as a list of things you can fix. Most of the defects here cost nothing but attention, and in a category where members choose on trust, a branch that cannot be contacted is a trust problem before it is a marketing one.
Read this as a list of things to ask about. One of the findings below is a member-security matter, not a marketing one, and it is visible to anyone with a browser right now.
Amplispot sells digital presence management, including a review and listings product called ReviewSpot. We have an obvious commercial interest in a report concluding that local presence is mismanaged, and you should read it with that in mind.
What we have done to make it checkable anyway: the scoring formula and the full deduction schedule are published, so you can recalculate any score yourself. Every figure came from public Google listings, so any claim can be independently verified. No institution was contacted before publication, paid for inclusion, or given advance sight of its score — and we excluded one institution from the ranking despite its high rating, because it has no branch estate to measure. ReviewSpot is not mentioned again in this document.
Every institution with a physical location has two shopfronts. One is the building. The other is a box on a map that most boards have never deliberately designed, and many have never read.
For credit unions the second shopfront does more work than the first. Membership is a considered decision made in a moment of friction — a branch closing, a fee that finally became intolerable, a move to a new city, a first car loan. People search, compare three results, and choose. The search result is not an advertisement for the relationship. It is the relationship's beginning.
In those forty seconds a person forms a complete impression from four things: whether your name looks like a real, specific place; whether your rating is above the two institutions beside you; whether enough other people have been there to make that rating mean anything; and whether you are open right now. Nothing on your website is consulted. No rate comparison. No annual report.
Each of those four can be silently wrong. A branch can publish hours that were reduced years ago and never restored. Two profiles can exist for one building, splitting the reviews so neither looks established. A corporate office can be listed as a public destination and quietly absorb complaints meant for the institution.
Because of what is being decided and what members write about. A restaurant listing carries opinions about dinner. A credit union listing carries a member describing a frozen account, a disputed charge, a hold on a deposit, or a suspected fraud on their own money — permanently, publicly, at the top of the search result the next prospective member reads.
And the decision itself is unusually durable. Somebody choosing a restaurant badly loses an evening. Somebody choosing a financial institution badly stays for years, or leaves and tells people why.
Search for one of these institutions in its home city and you get six to eight results carrying the same name, frequently the same phone number, often the same hours, distinguished only by a street address. A member cannot tell which branch they visited, which has the service they need, or which is closest to open.
Google is in the same position. Location-qualified listing names are one of the few direct levers left on local placement, so an estate of undifferentiated listings competes against itself for the same searches rather than against the bank down the road. It also means calls cannot be attributed to a branch — so nobody can see which branches generate demand.
None of this is visible from inside the building. Nobody on your team searches for their own branch, and nobody calls the number published on it. That is the entire reason this report contains anything you did not already know.
If that front door matters this much, the obvious question is who else is standing in the same doorway. In this category the answer is four thousand institutions — and shrinking.
Next, Part TwoFour thousand institutions, one shared habitAmerican credit unions are numerous, member-owned, and shrinking in number while growing in size — the classic conditions for central functions to be under-resourced at branch level.
As at the first quarter of 2026, down from 4,411 a year earlier, per NCUA system performance data. Consolidation is long-running and continuing.
Membership at federally insured institutions reached 145.8 million in Q1 2026, having added 2.5 million over the year. Roughly one in three Americans.
Credit union membership covers about a third of the country, which makes branch findability a mass-market question rather than a niche one.
A merging industry inherits estates. Branches arrive from acquired institutions with their own listings, their own names and their own phone numbers, and get folded into a central contact structure because that is operationally simpler. Marketing owns the brand, operations owns the hours, the contact centre owns the phone number — and the branch listing, which sits across all three, ends up owned by nobody.
The institutions in this study range from single-metro operators to national names. The pattern below does not correlate with size.
Nine of them are the subject of this study. Before the numbers, here is precisely how we selected them and what our method genuinely cannot see.
Next, Part ThreeHow we lookedNo institution gave us data. None was contacted before publication. Every finding was equally available to the institution itself.
We searched each institution by name across its principal market — the metro where it is headquartered or most concentrated — rather than sampling nationally, so that branches are compared against genuine local peers.
For every branch we captured the listing name, street address, telephone number, published opening hours, star rating, total review count, and the review text Google displays.
Most defects only appear across a whole estate. Shared phone numbers show up as one number repeated on every listing. Duplicates as two profiles at near-identical addresses. Naming failures by comparing names across the estate. Hours defects by finding a branch that contradicts its own network. This is why internal teams miss them.
Components were calculated from the captured figures using the formulas and deduction schedule in Part Five. No judgement was applied afterwards and no institution was moved for editorial reasons.
A tenth institution met every criterion except the branch threshold. It operates a digital-first model with effectively no retail estate — one surfaced location, carrying more reviews on its own than several entire branch networks in this study. Scored as-is it would have placed near the top purely because a single high-volume location produces a per-branch average no real network can match.
That is an artifact of our method rather than a finding about the business, so it is excluded. It is worth noting for a different reason: an institution with no branches has no branch-listing problem at all, which is one legitimate answer to everything in this report.
We expected the interesting variable to be service quality. It was not. It was something far more mechanical, and it appeared in eight of nine estates.
Next, Part FourThe finding we did not expectIt is the most consistent observation in the study, and it is also the one where an experienced operator will disagree with us — so we are putting both sides here rather than burying the counter-argument.
In eight of the nine estates, every branch listing publishes the same national contact number. We found no branch-level direct line anywhere in those eight. Members write about it constantly: that the national line cannot help with a matter specific to a branch, that there is no after-hours route, and in one case that a branch could not supply the phone number of another branch of the same institution.
Centralised contact routing is a deliberate operating model, not an oversight. A contact centre gives you longer hours, call recording for compliance, workforce management and quality assurance that a branch answering its own phone cannot match. And branch staff pulled off the counter to take calls serve the member standing in front of them worse.
So we removed it from the score. It is reported as an observation and carries no deduction. When we tested the ranking with and without it, the order did not change at all — which is the clearest possible evidence that a finding applying to eight of nine institutions does not discriminate between them. It was never doing analytical work; it was only doing rhetorical work, and that is not what an index is for.
Two things, and they are about the listing rather than the call routing.
Attribution is lost either way. Whatever your routing strategy, if every listing publishes an identical contact route you cannot attribute a call to a branch. A centralised contact centre can still capture which location a member is calling about — several do — and publish tracking numbers that route to the same place. The operating model and the listing data are separable decisions, and most institutions here appear to have made only the first one.
The member complaint is real regardless of who is right. Members are saying publicly that they cannot get help with a branch-specific matter. That may be an argument for branch lines, or it may be an argument that the contact centre needs branch context — the diagnosis is yours. But an unanswered complaint on a public listing is a fact whichever way the operational question resolves.
On one institution's branch listing, a member has posted a public warning telling others not to call the number shown. They describe dialling it and being routed through an interception layer that captured their details before forwarding the call to the genuine support line.
We are not in a position to verify that account, and we are deliberately not naming the institution or reproducing the technique. That is the point: the warning is sitting in public on a branch listing, unanswered, where members read it. Whether it is accurate or mistaken, an institution would want to know it is there — and if it is accurate, it is a member-security matter rather than a marketing one. This is the one finding in the study that we scored heavily, and it has nothing to do with call routing.
That is one component of four. Here is the full measure, published so you can recalculate your own score.
Next, Part FiveA measuring stickAn index you cannot audit is a marketing device, not a measurement.
Reviews per branch. Count is a ranking input in its own right, not just a confidence signal for humans. Log-scaled, because the gap between 30 and 90 matters more than between 900 and 1,000.
Whether the estate is correctly represented and contactable. Starts at full marks and loses points for each defect found.
Volume-weighted average across the estate, scaled 2.5 to 4.9.
The gap between the best and worst branch. A wide spread usually means nobody is watching at branch level.
Integrity begins at 35 points. Each defect costs the following:
Applied to seventy branches, that measure produces a table that looks very little like a ranking by asset size.
Next, Part SixThe ranking| # | Institution | Branches | Reviews | Per branch | Rating | Defects | Score |
|---|---|---|---|---|---|---|---|
| 1 | Institution F | 8 | 8,141 | 1,018 | 4.79 | 2 | 85 |
| 2 | Institution C | 8 | 2,426 | 303 | 4.33 | 2 | 74 |
| 3 | Institution I | 8 | 674 | 84 | 4.54 | 1 | 73 |
| 4 | Institution A | 7 | 2,704 | 386 | 4.41 | 3 | 70 |
| 5 | Institution H | 7 | 255 | 36 | 3.78 | 2 | 57 |
| 6 | Institution D | 8 | 762 | 95 | 3.43 | 2 | 56 |
| 7 | Institution B | 8 | 389 | 49 | 3.80 | 3 | 48 |
| 8 | Institution G | 8 | 515 | 64 | 3.22 | 3 | 43 |
| 9 | Institution E | 8 | 826 | 103 | 3.73 | 3 | 42 |
The largest institution in this study by assets and membership finishes last. The institution in first place operates in a single metropolitan market. Scale gives you a bigger contact centre; it does not give you a findable branch.
Two variables drive nearly all of the separation, and the size of both is larger than we expected.
Next, Part SevenWhat the ranking revealsFrom 36 reviews per branch to 1,018.
Two institutions here are collecting member feedback at genuine scale. Several are collecting almost none. Because volume feeds local ranking independently of score, an institution at the bottom of this chart can be outranked in search by a lower-rated competitor with more reviews on file. It also means a handful of complaints can move a branch rating by half a star, which is why the bottom of this chart and the bottom of the next one overlap.
Take a location sitting at 4.6 stars. In one difficult month it receives three one-star reviews — a staffing gap, a bad week, one genuinely poor experience shared by a family or a group. Nothing unusual.
What that costs depends entirely on how many reviews were already on file.
| Reviews already on file | Rating after | Drop | Five-star reviews to recover |
|---|---|---|---|
| 40 | 4.35 | −0.25 | 27 |
| 100 | 4.50 | −0.10 | 27 |
| 250 | 4.56 | −0.04 | 27 |
| 500 | 4.58 | −0.02 | 27 |
| 1,000 | 4.59 | −0.01 | 27 |
Two things in that table matter more than anything else in this report.
The same three reviews do twenty-five times more damage at a thin location than a rich one. And the recovery cost — 27 five-star reviews — is identical at every volume. It does not get cheaper because you are large. It is fixed by arithmetic.
So a thin estate takes visible hits and pays the same repair bill. That is why volume behaves like insurance rather than marketing: you cannot prevent bad reviews, you can only dilute them, and dilution has to be running before the bad month, not after it. Institutions here range from 36 to 1,018 reviews per branch. At the bottom of that range, three unhappy members visibly move a branch's public rating.
From 3.22 to 4.79. This category separates on sentiment in a way most do not.
In several industries we have audited, ratings cluster so tightly that they tell a marketing team nothing. Credit unions are the opposite: a 1.57-star spread across nine institutions, with individual branches ranging from 1.9 to 4.9. Members clearly distinguish between these institutions, and they distinguish between branches of the same institution — two estates here carry a 1.6-star internal spread.
Setting aside the shared-number observation discussed in Part Four, we counted nine scored defects: four naming failures, four sets of inconsistent or restrictive opening hours, one duplicate listing, one headquarters listed as a member destination, one estate trading only under an acronym, and one reported contact-security problem. Every institution carried at least one — the only study we have run where that is true.
Below is every institution, with its component scores and each defect we found, described generically.
Next, Part EightThe scorecardsWhat you take from this depends entirely on which side of the table you sit on.
Next, Part NineWhat to do about itIf any of the nine scorecards looked uncomfortably familiar, there is a straightforward way to find out.
Next, Part TenWhich institution are youInstitutions in this study are anonymised. We hold a private key mapping each letter to its institution, along with a verification code that appears nowhere in this document.
Email us from a company domain and we will confirm your letter, quote your verification code back to you, and send the full underlying detail for your estate — every branch, every defect, named and specific. No charge, no meeting required, and we will not add you to anything.
If you are the institution referenced in Part Four's security finding, write to us and we will send you the specific listing and the member's post immediately, ahead of any other correspondence. We will not confirm any other institution's identity to you, and we will not confirm yours to anyone else.