Anatomy of a Review

Written By Chad McGuire (Sparrow Intel)

Overview

A review looks simple — a star rating and a paragraph of text — but every review in Sparrow Intel has structure underneath. Knowing what's there helps you read faster, reply faster, and know what your rules can act on. This lesson is a vocabulary lesson. Open any review alongside it and match what you read to what you see.

Review vs. response

A review is what the guest wrote about their stay. One stay = one review per channel (you can't get two Airbnb reviews from the same reservation). A response is the reply you (or Chirp AI) post publicly under the review. Responses are governed by the OTA — see Replying to a Review for what each channel allows. When this curriculum (or the platform) says "review," we mean the guest's post. When we say "response" or "reply," we mean what's sent back.

Review-level fields

Every review has metadata you can see on the detail page (click Analyze from the inbox row) and use as filters or rule conditions:

Field

What it is

Guest

First name (last name is rarely shared by OTAs), language

Reservation

Confirmation code, dates, OTA, status, property

Property

The listing or unit being reviewed

OTA / Channel

Where the review lives (Airbnb, Booking.com, Google, etc.)

Posted date

When the review went live on the OTA

Star rating

Overall rating, normalized to your scale setting (5-star or 10-point)

Category ratings

Per-OTA sub-ratings — Cleanliness, Communication, Check-in, Location, Accuracy, Value (varies by channel)

Body

The actual review text

Original language

Auto-detected language of the review

English translation

Stored even if the review is already English — useful for non-English reviews

Reply status

Awaiting reply / Replied / No reply needed

Workflow flags

Starred (team-visible), Archived (out of queue), Snoozed (delayed), Marked for removal

Review-level fields describe the whole review. Most don't change after the review posts — the exceptions are the workflow flags and reply status, which evolve as your team works the inbox.

AI-extracted fields

Chirp AI reads every review the moment it arrives and adds a structured analysis layer:

Field

What it is

Document sentiment

Overall positive / neutral / negative read on the whole review, with confidence

Sentence-level sentiment

Per-sentence breakdown — useful when a review is mostly positive with one stinger

Opinions

Specific things the guest praised or complained about ("loved the hot tub," "WiFi was unreliable")

Cleaning issues

Cleaning problems mentioned ("bathroom wasn't cleaned," "hair on the pillow")

Maintenance issues

Maintenance problems mentioned ("dishwasher didn't drain," "AC clicked all night")

Detected language

What language the review was written in

Most of this is generated the moment a review arrives. You don't have to tag anything by hand. We cover how to read this analysis in Review Analysis: Sentiment, Opinions & Issues.

What "sentiment" actually means

Two different sentiment values show up in the UI, and they are not the same thing:

  • Document sentiment — Chirp's read on the whole review. This is what most rules and filters use.

  • Sentence sentiment — per-sentence reads. Useful when a 4-star review hides a sharp complaint in one sentence, or when a 2-star review still says something nice you should acknowledge in the reply.

A review can be overall positive but contain a negative opinion ("everything was great except the parking situation"). Treat opinions, not just the overall sentiment, as the thing your reply should address.

A few terms you'll see in the UI

  • Starred review — a team-visible flag for "come back to this"; anyone in your account sees the same starred state

  • Archived review — out of the active queue, still searchable, won't show in default views

  • Snoozed review — temporarily hidden until a date you pick; re-appears in the queue

  • Marked for removal — flagged as a policy-violating review you're pursuing removal of (covered in Tracking Review Removals)

  • Generated reply — an AI-drafted response shown in the composer, awaiting your send

  • Host-to-guest review — your team's review of the guest, posted back to the OTA (covered in Host-to-Guest Reviews)

You'll meet all of these properly in upcoming lessons.

Up next

You know how to find reviews and what's in them. Time to Reply to a Review.