TrustScore Monitoring → Reputation Response

Detect TrustScore drops, analyze root causes with Mave Agent, and generate response templates

Scenario

Your TrustScore dropped from 4.5 to 4.2 this month. Why? You pull recent negative reviews, send them to Mave for root cause analysis, then generate response strategies and specific reply drafts. This creates a rapid-response loop: detect score drops → understand causes → respond systematically.

Flow: Trustpilot GET /business-units/{id} (score) → Compare to baseline → If dropped: GET /reviews (recent 1-3 star) → Mavera POST /mave/chat (root cause) → POST /generations (response drafts)

Architecture

Code

Python

import os, requests, time

TP_KEY = os.environ["TRUSTPILOT_API_KEY"]
MV = os.environ["MAVERA_API_KEY"]
BU_ID = os.environ["TRUSTPILOT_BU_ID"]
TP_BASE = "https://api.trustpilot.com/v1"
MV_H = {"Authorization": f"Bearer {MV}", "Content-Type": "application/json"}

# 1. Current TrustScore
bu = requests.get(f"{TP_BASE}/business-units/{BU_ID}",
    params={"apikey": TP_KEY}).json()
current_score = bu.get("score", {}).get("trustScore", 0)
review_count = bu.get("numberOfReviews", {}).get("total", 0)
star_dist = bu.get("score", {}).get("stars", {})

BASELINE_SCORE = 4.5

print(f"Current TrustScore: {current_score} (baseline: {BASELINE_SCORE})")
print(f"Total reviews: {review_count}")
print(f"Star distribution: {star_dist}")

# 2. If score dropped, pull recent negative reviews
if current_score < BASELINE_SCORE:
    drop = BASELINE_SCORE - current_score
    print(f"\n⚠ Score dropped by {drop:.1f} points")

    negative_reviews = []
    for stars in [1, 2, 3]:
        r = requests.get(f"{TP_BASE}/business-units/{BU_ID}/reviews",
            params={"apikey": TP_KEY, "stars": stars, "perPage": 20,
                     "orderBy": "createdat.desc"})
        r.raise_for_status()
        negative_reviews.extend(r.json().get("reviews", []))
        time.sleep(0.2)

    # 3. Build review block for analysis
    review_block = "\n\n".join(
        f"[{r.get('stars',0)}★] {r.get('title','No title')}\n{r.get('text','')[:300]}\nDate: {r.get('createdAt','')[:10]}"
        for r in negative_reviews[:20]
    )

    # 4. Mave root cause analysis
    analysis = requests.post("https://app.mavera.io/api/v1/mave/chat",
        headers=MV_H,
        json={"message": f"""Our Trustpilot TrustScore dropped from {BASELINE_SCORE} to {current_score}.
Total reviews: {review_count}. Star distribution: {star_dist}.

Recent negative reviews:
{review_block}

Analyze:
1. Root causes — what themes drive negative reviews?
2. Frequency — which issues appear most?
3. Severity — which issues cause 1-star vs 3-star?
4. Trend — are issues getting worse or improving?
5. Actionable fixes — what operational changes would address each root cause?
6. Response strategy — how to reply to each theme category"""}).json()

    print("\n=== Root Cause Analysis ===")
    print(analysis.get("content", "")[:1500])

    # 5. Generate response templates
    gen = requests.post("https://app.mavera.io/api/v1/generations",
        headers=MV_H,
        json={"prompt": f"""Based on this root cause analysis, generate 4 review response templates:

{analysis.get('content','')[:1000]}

For each template:
- Category name (e.g., "Shipping Delay", "Product Quality")
- Tone: empathetic, solution-oriented, professional
- Acknowledge the specific issue
- Offer a concrete next step
- Keep under 100 words each
- Never use corporate jargon"""}).json()

    print("\n=== Response Templates ===")
    print(gen.get("output", gen.get("content", gen.get("text", "")))[:1200])

else:
    print(f"✓ Score stable at {current_score}")

Example Output

TrustScore: 4.2 (baseline: 4.5)
⚠ Score dropped by 0.3 points

=== Root Cause Analysis ===

## Theme Analysis (20 recent negative reviews)

| Theme | Mentions | Avg Stars | Trend |
|-------|----------|-----------|-------|
| Shipping delays | 8 | 1.8 | Worsening (5 in last 2 weeks) |
| Product quality | 5 | 2.4 | Stable |
| Customer support wait | 4 | 2.0 | New issue (all from last month) |
| Billing errors | 3 | 1.3 | Stable |

## Root Cause: Shipping
Carrier switch in March caused 3-5 day delays. Reviews mention "used to arrive in 2 days."

## Actionable Fixes
1. Shipping: Revert to previous carrier or add expedited option
2. Support: Hire 2 additional agents — wait times exceeded 45 min
3. Billing: Audit subscription renewal flow — 3 cases of double-charge

=== Response Templates ===

### Shipping Delay
"Hi [Name], I'm sorry your order took longer than expected. We recently
changed carriers and are actively fixing the delays. I've escalated your
order — you'll receive tracking within 24 hours. If it doesn't arrive by
[date], reply here and I'll personally ensure a full refund."

### Customer Support Wait
"Thank you for your patience, [Name]. Wait times have been longer than our
standard, and we take that seriously. We've added staff this week. I'd love
to resolve your issue now — can you DM us your case number?"

Error Handling

TrustScore precision

TrustScore is rounded to 1 decimal. A 0.1 change can represent dozens of reviews. Track review volume alongside score for context.

Baseline storage

Store your baseline TrustScore in a database or config file. The code uses a hardcoded baseline — replace with dynamic comparison in production.

Response posting requires OAuth

Reading reviews uses the API key. Posting replies requires OAuth 2.0 (POST /reviews/{id}/reply). Set up OAuth before automating responses.