Understanding the Quality AI at SquadStack

Quality assurance is a critical element for any business. As the world is rapidly moving towards automation and cost-efficiency, human interaction has almost become a liability. We want to automate replies, use bots, and make the best use of technology to empower them and keep the human in the loop limited.

But for businesses hinged on customer interaction, humans play a key role in establishing that trust and building a relationship. Telecalling is a cost-effective and foolproof method to connect with your customers in a targeted and personalized manner. And quality assurance is what makes this entire process reliable.?

How SquadStack is Leading the Way

SquadStack is a comprehensive solution for businesses that seek fully managed telecalling services. It covers everything from hiring, training, and managing telecallers so that the partner businesses can focus on their growth.?

Simply put, it is a full-stack solution that augments human expertise with the power of technology to achieve 100% productivity with quality.

Top Indian companies in industries like E-commerce, Financial Services, and Edtech collaborate with SquadStack's fully managed telecalling support to improve their business results, enhance customer experience, and safeguard their brand.

SquadStack has been a game-changer for several Indian companies seeking exponential growth and business success. Upstox, a top investment platform, achieved a 40% increase in their active user base by leveraging SquadStack's sales experts and AI-driven outreach. Choice India was able to slash their customer acquisition cost by 200% with the help of SquadStack's sales experts and robust tech stack. WOW Skin Science achieved an impressive 8x ROI with SquadStack's abandoned cart recovery calling expertise. MoneyView also saw significant improvement in their internal team processes and rapid business growth by leveraging SquadStack's telecalling expertise.?

The Innovation that is Driving SquadStack’s High-Quality Calls?

SquadStack's fully managed telecalling stack relies on an AI-powered continuous quality monitoring process, which is a crucial component of its infrastructure.

The Automated Quality Check (AQC) engine helps measure the quality of calls at scale, making it easier and faster to ensure high-quality calls. Traditional quality monitoring and measurement methods are slow and costly. SquadStack’s AQC removes the need for human intervention in the initial stage, which provides fast and accurate measurements while allowing for easy scalability.?

AQC is estimated to be 3-5X cheaper than traditional methods. This makes it an affordable and reliable method for businesses actively looking to improve their call quality. It can provide real-time audits, identify areas of improvement, and partially detect fraud or anomalies.?

With its deterministic measurement process (a process where the outcome of the measurement can be determined with 100% accuracy) and the ability to answer various questions for different use cases, AQC provides a human-in-the-loop framework that ensures high-quality results. It is a smart solution that is deployable almost instantly, saving time and effort on management and hiring.

Choosing the right customer service BPO provider

A Closer Look at Quality AI

If one were to bucket the whole process of achieving Quality AI - it requires a combination of technical expertise, ethical considerations, and rigorous testing and validation processes throughout the entire AI development lifecycle.

If an AI system has the ability to consistently perform its intended functions under different conditions and produce correct error-free results, that model is reliable and accurate.?

The Process at SquadStack

The distinctive feature of SquadStack’s call quality monitoring, as compared to other telecalling services, is its utilization of an AI-based system to monitor the quality of conversations between callers and customers.

After the calls are completed for the day, a sampling generator is used to select a set of connected calls that consider all critical factors, including business outcomes and a minimum of two calls per telecaller. These selected calls are then passed through the Audit UI, where the machine learning system assesses 23 parameters in a yes/no format (each parameter has a score assigned based on the binary assessment), and the output is stored in a database.

All 23 parameters are directly linked to the C5 attributes the team experts have established as critical to a high-quality call experience.?

The parameters are related to four aspects of the calls: connecting, conveying, convincing, and closing. Additionally, there are compliance-related parameters for the telecaller throughout the call to ensure that the telecaller is consistent and compliant.?

The system calculates the quality score by multiplying the parameter score with the parameter weight at the parameter level, as shown in the following formula:

Formula: ?Σ[(Parameter Score)*(Parameter Weight)]Σ(parameter weight)

The quality score is an amalgamation of the aforementioned C5 facets that encapsulate the skills that each telecaller needs to focus on. The score is computed at the VPM level, and the actioning team uses this feedback to improve the telecaller's performance.

After the score is computed, it is then passed through Auditing the Auditor to ensure the reliability of the system’s quality score. For instance, out of 100 connect calls, if 10% are selected through the sampling generator, 50% of the selected calls go through this auditing process.?

Significance of Auditing the Auditor?

To put it simply, auditing the auditor helps increase confidence in the AI system and the quality score that it generates and reduces the margin of error.? Over time, machine models can deteriorate because data is dynamic and constantly changing (this is often known as data drift which refers to the deterioration of a machine learning model's performance in production over time, often caused by changes in input data distribution). Auditing the auditor ensures that if the model starts performing badly, the required checks are already in place - a form of prevention as opposed to cure.?

How is this measured?

Confidence% is a measure of the degree of certainty that is in place for the reported quality score. For example, if we state that the quality score is 80% with a 90% confidence level, it means that we are 90% confident that the true quality score falls within a range of values that includes 80%. The remaining 10% represents the margin of error or uncertainty in the score.

Sample Set Distribution?

As a thumb rule in statistics, uniform distribution of samples from a set is used to evaluate how well any system or process is performing. However, this method does not always guarantee that all the callers will be checked and monitored at all times, and there might be some outliers.

To address this issue, SquadStack has set a target of 100% campaign checking for all telecallers. So at least two connected calls for each caller from each campaign are audited, and all outcomes are covered. And the system ensures that more than just the mere quantity, all types of calls that may matter to the business concerned and to SquadStack and its processes are checked.

Ensuring Caller Adherence to the Standards

To ensure adherence to quality standards, the system aggregates the score for a minimum set of calls to check the caller's performance. For instance, if at least five calls have been audited in the last 15 days for a particular caller, the score is based on the average for those calls, and the caller is then categorized as red, amber, or green. The red category indicates poor performance and a marker of the fact that the specific caller might need replacement, while the amber category suggests that the caller needs more training - all of which is fully taken care of and managed by the SquadStack team.?

Sample Audit Parameters

Given the varied experience across multiple industries and customers, SquadStack has managed to fully understand and become industry experts on the different quality monitoring requirements when it comes to multiple use cases. This allows the customers to completely rely on the set of parameters based on which callers for each campaign are monitored.?

Here’s a sample of three key parameters -

Creating a sense of urgency

A common tactic often used during sales calls, this allows the callers to push potential customers to make decisions quickly. For example, if a caller is trying to sell a product or service, emphasizing the benefits of a time-sensitive offer, highlighting the potential for a future price increase, or changes to terms and conditions can do the trick.?

Handling objections and rebuttals

When faced with objections or rebuttals from a potential customer, it is important to listen carefully and address their concerns directly.?

Avoid Misselling

Convincing customers that your product is superior to that of your competitors by highlighting the unique features and benefits can be effective. However, it is important to avoid misrepresenting or exaggerating these features to inflate your conversion numbers.?

Automation and Manual Efforts

The integration of automation and manual efforts is a critical consideration in achieving optimal productivity and cost-efficiency when it comes to quality monitoring.?

Today (as of March 2023),SquadStack’s team of data scientists manually reviews 10% of the first sample, with a 50% pass rate. The decisions made are based on this subset, with 9 out of 23 parameters already surpassing the target. The vision is to achieve 80-85% confidence in sales use cases and 90% in operations and pre-sales use cases through manual efforts.?

Why SquadStack’s Quality Monitoring Surpasses Industry Standards

Smart Sampling

SquadStack utilizes a selective approach to call sampling based on previous behavior or results rather than random selection. By targeting specific campaigns and outcomes, false qualifications are minimized while achieving a sampling rate of 10%, which surpasses any other BPO in the market (industry standard rate is at 1-2%).

Reducing Linear Cost Increase

SquadStack employs machine learning to reduce the linear cost increase and instead applies a logarithmic increase in cost. Because of automation, this results in significant savings as the company scales.

Faster Feedback Loop

The feedback loop today has an average duration of 24 hours, whereas a manual process would take a minimum of 36 hours and up to 48 hours. SquadStack's feedback loop is faster and can be accelerated further based on the increase in the utilization of technology.

Deterministic Approach

SquadStack’s approach to call assessment is free of human bias with is different from regular BPOs . Whether a call is checked today or in 10 days, the results will remain the same. This approach provides clear and objective answers to questions and allows the company to take action in an objective manner (except in cases where guideline changes occur).

Flexible Parameters

SquadStack offers the flexibility to add or remove parameters or adjust frequency based on business needs. This feature is not commonly found among competitors.

Expertise

Based on past experiences, SquadStack understands the requirements of various use cases and has identified the best method to check and ensure quality, making it an expert in standard parameters. This eliminates the need for each customer to create their own parameters, thus streamlining a successful and foolproof process.