Review sources

How to aggregate reviews from multiple sources in one widget

10 September 2026 · JustReview

Pulling Google, Facebook and marketplace reviews into a single widget is a solved problem technically, and almost nobody writes about the part that actually bites: the sources do not agree on what a star is. You can merge the feeds in a few minutes. Deciding what the combined average should say, which reviews sit next to which, and what happens to a visitor who reads three languages in a row takes longer, and it is the part that decides whether the widget helps.

This post uses numbers from our own production database, measured on 10 September 2026, not from a vendor comparison page.

Roughly one in four shops runs more than one source

Across the accounts holding at least one live review, 3,728 in total, 883 draw from two or more sources. That is 23.7%. The tail is thinner than the marketing around “all your reviews in one place” suggests, and the shape of it matters if you are deciding how much effort to spend here.

Counting configured connections rather than reviews actually collected raises the multi-source share slightly, to 1,205 of 4,701 accounts, or 25.6%, because some connections never return anything. The busiest account in the database holds reviews from 13 different sources; the busiest by configuration has 14 connectors attached.

So the realistic planning assumption is two sources, not ten. Build for two done properly and you have covered three quarters of the multi-source population.

The pairing that dominates is also the one with a scale mismatch

Sorted by how often each combination actually occurs, one pairing is not merely first, it is most of the category.

Google with Facebook covers 687 of the 883 multi-source accounts, 77.8%. Everything else is a long tail. That concentration is convenient, because it means the single most valuable thing to get right is the behaviour of exactly those two feeds together, and that is where the trap is.

Facebook has two star values. Google has five.

We checked every distinct star value each source has ever delivered into our database. The sources do not share a scale, and two of the biggest do not even come close.

Facebook replaced star ratings with a yes-or-no recommendation years ago, and that is exactly what reaches us: 47,347 reviews scored 5 and 1,768 scored 1, so 96.4% of every Facebook review in our database is a five. There is no such thing as a Facebook three. eBay is nearly as blunt, because its feedback is positive, neutral or negative and lands as 5, 3 or 1: 196,747 fives against 1,405 threes and 1,389 ones, 98.6% positive. Allegro is the opposite problem, arriving pre-averaged in thirds so that a “4.33” is not a rating anyone ever gave.

Averaging those together and printing one number is arithmetically easy and editorially misleading. A yes/no vote is not a four-star opinion, and stacking a source that literally cannot produce a three next to one that can will bend the result upward.

What the merge actually does to your average

The honest answer is: usually very little, occasionally a lot. Across the 687 accounts holding both Google and Facebook reviews, we compared the Google-only average with the combined average.

The mean moves from 4.817 to 4.840, a shift of +0.023 stars. For most shops the merge is cosmetically invisible. But the average of the shift hides the shops it happens to: 105 of those accounts, 15.3%, see their displayed number move by at least a tenth of a star, and for 79 of them, 11.5%, it moves down, not up. A source that is 96.4% five-star can still drag your number down if your Google reviews were better than that.

That is the practical rule. Merging is safe for roughly six shops in seven and materially changes the headline number for the seventh, in either direction. Check it before you publish rather than after, and if the combined figure is worse than your best single source, show the sources separately instead. Our own comparison of which review sources still work in 2026 is a reasonable starting point for deciding which feeds deserve the space at all.

Deduplication: what it can and cannot do

Every review we import is keyed on the platform’s own review ID, scoped to your account and that one source. Before an import runs we read the set of review IDs already stored for that account and source, and skip anything already present. That is why re-running an import, or running it nightly for years, never duplicates a review: the identity comes from the platform, not from matching text.

It also defines the limit precisely. Google IDs and Facebook IDs live in different namespaces and have nothing in common, so there is no key on which to match a person across two platforms. If the same customer praised you on Google in March and on Facebook in April, those are two real reviews from two real accounts, and both are shown.

We measured how often that overlap is even visible. Scanning all 687 Google-and-Facebook accounts by author name, 226 account-and-name pairs appear on both sources, spread over 41 accounts, so 6.0% of multi-source Google-plus-Facebook accounts have at least one person on both sides. It is not nothing, and it is concentrated: most accounts have none, a few have many. If yours is one of them, hiding individual reviews is a per-review decision you make in the panel, not something an import should guess at.

Be sceptical of any widget that advertises cross-source deduplication as a feature. Matching on a display name is guesswork, two different people are often called Anna K., and the failure mode is silently deleting a genuine review.

Language is the bigger practical problem

Multi-source accounts are usually multi-language accounts, and this catches people out far more often than duplicates do. Of the 883 multi-source accounts, only 383 hold reviews in a single language. 500 of them, 56.6%, hold two or more, 295 hold three or more, and the widest holds 69.

A visitor landing on your English page and reading a Dutch review, then a Polish one, then a German one, learns nothing and leaves. Filtering the widget by page language is the fix, and it is a display setting rather than an import setting, so nothing is thrown away: the reviews stay in your account and simply surface on the pages where they can be read. The filter by page language option exists for exactly this and should be the first thing you switch on after connecting a second source.

Ordering is the smaller sibling of the same question. Newest-first is the sane default, because it is the only order a visitor can verify and it makes the widget look maintained. Sorting by rating so the fives float to the top is the fastest way to make a real review wall look bought.

Doing it in JustReview

The reason to run this here is that the merge is a display decision rather than a migration. You connect a second source, and the same widget you already embedded starts drawing from both; the snippet on your page does not change, so there is no redeploy and nothing to re-test on your site. Sources can be turned on and off, filtered by language and by rating, and the star average recalculates from whatever is currently visible, which is what lets you check the combined number before you commit to showing it. Our own form is one of those sources, so first-party reviews collected on your site sit in the same widget as the Google ones, keyed and deduplicated the same way. Full setup for the most common starting point is on the Google reviews widget page, and the supported sources list shows what else can be added next to it. You can create an account and connect two sources without touching your site’s code.

Two caveats worth knowing before you start. Marketplace sources behave differently from local listing sources and are worth reading up on separately in our note on marketplace reviews on your website. And the number of sources and locations you can attach depends on the plan, so if you are heading toward four or more feeds, check pricing first rather than after you have connected three.

If you are running two sources today, the useful next hour is not spent on integration. It is spent looking at your combined average against your best single source, and switching on the language filter. Those two decisions are what separate a review widget that convinces someone from one that just sits there.