Ask a restaurant owner whether their first few reviews mattered and most will tell you it was a slow start, that they found their feet, that it worked itself out. We wanted to know whether the data agreed.
So we took every Saudi business in our corpus with at least 100 written Google reviews - 1,799 of them, across Jeddah, Riyadh, Dammam and Mecca - and compared each one's earliest ten reviews against everything that came afterwards. Not against other businesses: against its own later self. The question was simple. Does a venue's opening predict where it ends up?
It does. The correlation between a business's first ten reviews and all its later ones is 0.625 - strong enough to be unmistakable, loose enough to leave room for change.
The interesting part is the shape of that room.
The trap we nearly fell into
Our first pass produced a cheerful headline: of the businesses whose first ten reviews averaged below 4.0 stars, 65% now sit above 4.0. Two in three recover.
That figure is true and it is nearly meaningless, because 4.0 is the wrong line. The average rating across these 1,799 businesses is 4.243. Clearing 4.0 does not mean rejoining the pack - it means reaching the edge of it. Most of those "recoveries" land in the dead zone between 4.0 and 4.24: above the round number, still below the market.
Measured against the real baseline, only 20% of businesses that opened below 4.0 have climbed above the average. Not two in three. One in five.
The residue
Businesses with a rough opening do genuinely improve. Their first ten reviews averaged 3.285; today they average 4.012. That is a climb of nearly three quarters of a star, and it is real.
But they stop short. They sit 0.23 stars below the market average, and they stay there - the gap is as statistically solid as anything in this analysis.
If early reviews were simply noise - a small, unlucky sample that says nothing about a business - then every group would drift back to 4.243 given a hundred reviews to work with. They don't. Here is where each opening band actually lands:
| First ten reviews averaged | Businesses | Rating today | vs. the 4.24 average |
|---|---|---|---|
| Under 3.0 | 132 | 3.82 | −0.43 |
| 3.0 – 3.5 | 194 | 4.00 | −0.25 |
| 3.5 – 4.0 | 283 | 4.12 | −0.13 |
| 4.0 – 4.5 | 444 | 4.24 | −0.01 |
| 4.5 – 5.0 | 746 | 4.44 | +0.19 |
Every step down in opening quality is still visible a hundred reviews later. The worst openers - the 132 businesses whose first ten reviews averaged below 3.0 - remain 0.43 stars behind the market today. Nothing in that column flattens out.
The mirror image is softer than we expected. A third of businesses that opened above 4.0 have since drifted below the average. A good start is an advantage, not a guarantee.
Who it binds hardest
The grip of a first impression is not the same everywhere.
| Category | Businesses | Correlation |
|---|---|---|
| Hotels | 479 | 0.70 |
| Restaurants | 827 | 0.63 |
| Cafés | 289 | 0.55 |
| Gyms | 204 | 0.48 |
Hotels are the most path-dependent business in the dataset: where a hotel starts is where a hotel largely stays. Gyms are the loosest - a gym's opening reviews tell you less about its future than any other category here. If you run a gym, your early reviews are a weaker verdict than you might fear. If you run a hotel, they are a stronger one than you might hope.
City makes almost no difference. The correlation sits between 0.60 and 0.66 in all four cities. This is not a Jeddah phenomenon or a Riyadh one.
What we are not claiming
We can measure that a weak opening and a weaker present travel together. We cannot show that one causes the other, and we want to be plain about that, because the two explanations lead somewhere different.
It may be that early reviews do damage: a 3.2 rating suppresses footfall, thin footfall means thin margins, thin margins mean the understaffed Friday night that produces the next bad review. Reputation as a trap that closes.
Or it may be that the first ten reviews are simply an early and accurate reading of a business that was always going to struggle - weak location, weak concept, weak operator - and the rating is reporting that, not causing it.
Our data cannot separate those. Both are consistent with everything above. The practical advice happens to be the same either way, which is why we are comfortable publishing it: the opening months are worth more attention than their share of your trading life suggests.
What to do with this
If you are opening. Your first ten reviews will be read as a verdict for longer than seems fair. Delay the soft launch rather than open under-staffed. The cost of a quiet extra fortnight is smaller than the cost of 0.43 stars you will spend years paying down.
If you opened badly. It is recoverable - these businesses climbed three quarters of a star. Just do not expect the slate to come clean. Plan for a long grind back toward average rather than a turnaround, and judge progress against your own trajectory rather than against your neighbours.
If you opened well. A third of businesses that started above 4.0 are now below average. The advantage is real and it is not self-sustaining.
Method
This analysis covers 373,874 written Google reviews across 2,219 businesses in Jeddah, Riyadh, Dammam and Mecca, pulled at 100% coverage using keyset pagination and verified against an exact count rather than a capped query.
The study set is the 1,799 businesses with at least 100 written reviews. For each, we took the ten earliest reviews by date and compared them against that business's remaining reviews and its current stored Google rating. The design is within-business throughout: every business is compared against its own later self, never against a group of other businesses, so the finding cannot be an artifact of which businesses happened to be measured when.
Two honest limits. First, ten reviews is a small and noisy sample, so part of the movement we see in both directions is regression toward the mean - the residual gap that remains after a hundred reviews is the part that is not regression, and it is what we have reported. Second, as set out above, this is a correlation and not a causal finding.
We make no claim about how many reviews businesses receive over time. Our corpus grew substantially during 2026 as businesses were added, so review counts by period reflect when a business entered our data as much as customer behaviour.