Service Specification¶

Executive Summary¶

Customer Match is a feature that allows you to upload reviews from your customers and match them to the existing customer database. When supplied with review information, this service will attempt to identify which customer is most likely to have left this review. In the event that a high-confidence answer cannot be provided, the service will provide a list of possible customers sorted by likelihood instead.

Objective: Enrich our data to power opportunities for our clients to accelerate customer driven growth
Correct customer data is critical to the success of our clients. We want to be able to provide our clients with the most accurate customer data possible. This will allow them to make better decisions and improve their customer experience.
Objective: Increase new customer acquisition for our clients
Key Result: Referrals contribute to a 20% total customer vocality rate by the end of 2022 from 15% baseline.
- "Get Reviews" campaign performs better by identifying when customers have left a review.
- Customers that have left a review are identified as doing so, which means they are correctly entered into the Referrals campaign (entry rule is that a known customer has left a positive review). More entries into the Referral campaign mean more referrals.

Value to the Growth Platform¶

Customer Match is a feature that allows us to identify customers who have left reviews. This allows us to provide our clients with the most accurate customer data possible, and helps us capitalize on opportunities to increase new customer acquisition for our clients.

Service Level Agreements (SLAs)¶

Throughput¶

Median requests per second: 0.1 95th percentile requests per second: 0.6 Variability in Daily Request Distribution: Medium Variability in Weekly Request Distribution: Medium Times of Year with Normal Expected Traffic: None Request 200 status: 98%

Latency¶

Median: 3.3 seconds 95th Percentile: 6.6 seconds 99th Percentile: 7.1 seconds Worst Case Latency: 59 seconds

Schema¶

-Instance
    - ID: gid
      Description: Global Unique Identifier for the incoming Review
      Data Type: str
      Runtime Restrictions: None
    - ID: namespace
      Description: Company Namespace of the incoming Review
      Data Type: str
      Runtime Restrictions: None
    - ID: author_first_name
      Description: First Name of the Review Author. If null, pass empty string.
      Data Type: str
      Runtime Restrictions: None
    - ID: author_last_name
      Description: Last Name of the Review Author. If null, pass empty string.
      Data Type: str
      Runtime Restrictions: None
    - ID: author_email
      Description: Email Address of the Review Author. If null, pass empty string.
      Data Type: str
      Runtime Restrictions: None
    - ID: created_at
      Description: Timestamp of when the Review was created (in UTC)
      Data Type: str
      Runtime Restrictions: None

-Parameters
    - ID: max_candidates
      Description: Maximum number of candidates to return (default 3)
      Data Type: int
      Runtime Restrictions: None

-Prediction
    - ID: candidates
      Description: List of candidate matches sorted by likelihood
      Data Type: list of str
      Runtime Restrictions:
    - ID: perfect_match
      Description: Whether or not the prediction yields a perfect match
      Data Type: boolean
      Runtime Restrictions: None

Feedback Mechanisms¶

In cases when the model fails to perfectly match a review, we provide a list of candidates that are supplied to the user. These candidates are sorted by likelihood of being the correct match. The user can then select the correct match from this list. This feedback can be used to improve the model over time by creating more robust, difficult data points for re-training.