E-commerce
September 2, 2026
Wondering how to offer a size recommendation that reassures without misleading? The key lies in transparency: the chatbot must explain its sources, whether they are the customer's measurements, the size guide, or purchase history, while clearly signaling its limitations.
A justified recommendation builds trust and reduces returns, but it should never hide uncertainties related to varied cuts or limited stock.
So how do you structure this logic to maximize conversion without misleading the visitor? On the agenda:
What essential criteria need to be explained to make a size credible?
How do you integrate purchase history while respecting privacy?
What limitations should be communicated to avoid misunderstandings?
What transfer-to-human strategy should be adopted in case of doubt?
How do you measure the actual effectiveness of these recommendations using relevant KPIs?
Here we go.
Summary
Why is it crucial to explain the reasoning behind a recommendation?
A suggested size without justification risks seeming arbitrary or magical to the customer. If a visitor hesitates between several options, they have an imperative need to understand the logic behind the advice, whether it is a size S, M, or 42.
Transparency is the primary tool for building trust: the chatbot must make its recommendation understandable by listing the data used. This includes the precise measurements provided by the customer, the specific cut of the product, the expressed comfort preference, and any inherent margin of uncertainty.
Make the logic explicit
Cite the customer's measurements (chest circumference, hip, etc.).
Analyze the cut of the product (fitted, loose, deconstructed).
Mention the material (stretch or rigid) that influences the choice.
Indicate whether purchase history was used to refine the advice.
A recommendation will inspire trust provided it explains its reasoning in simple language. This transforms a technical suggestion into personalized and reliable guidance, preventing the customer from feeling guided by an opaque black box.
Convert over 2,000 customers on average per month with Qstomy.
The world’s 1st Shopify AI dedicated to customer conversion
Empowering 200+ e-commerce merchants
What specific data should feed the chatbot's logic?
To function properly, the bot must draw from a diverse ecosystem of data. It can use the official size guide, the anatomical measurements provided by the user, and their saved usual size.
Exploiting feedback from other customers is also valuable for validating the accuracy of the recommended size. Finally, purchase history helps to understand the style or fit that the customer already prefers, provided it is not over-generalized.
The diversity of sources
The website's size guide for standard measurements.
The purchase history to know sizes already worn successfully.
Customer reviews and returns to adjust according to specific models.
Product details such as fit (slim, regular) and material.
The bot must limit the use of history to elements that are truly useful. It is imperative to explain clearly if personal data influences the recommendation to maintain customer trust and respect their privacy.
How to manage purchase history without scaring the customer?
Using purchase history can greatly improve the accuracy of advice if a customer has already bought a similar size. However, the chatbot must approach this point with sensitivity and clarity to avoid any feeling of intrusive surveillance.
If the system detects a previous purchase in the same category, it can simply mention: "I am also basing this on your previous purchase in this category" to refine its suggestion. This personalized approach reassures the customer by showing that they are recognized.
The importance of consent
Explicitly offer the use of past data.
Explain that this helps avoid making a wrong choice again.
Allow the customer to refuse the use of their history.
Offer an alternative recommendation based on the standard guide in case of refusal.
If the customer does not want their history to be used, the bot must immediately switch to a standardized recommendation based on the measurements provided or the general guide. This flexibility shows complete respect for privacy preferences while maintaining the quality of service.
How should the response be structured to cover the nuances of cutting?
The presentation of the recommendation should not be limited to a one-size-fits-all approach. The chatbot must suggest a primary size tailored to the desired fit, while offering an alternative based on the customer's comfort preference.
For example, it can advise: "Choose M for a fitted look, or L if you prefer a looser fit". This approach recognizes that the choice often depends on personal preference rather than raw measurements. It avoids locking the customer into a single option.
Handling Hesitations
Identify if the customer is "between sizes".
Suggest a smaller size for a shaping or tight-fitting effect.
Suggest a larger size for more comfort or layering.
Avoid presenting the size as certain if hesitation persists.
The chatbot should remain cautious when it detects that the customer is between two measurements. In this case, it is more useful to provide guidance on the desired fit rather than guessing a perfect size that might turn out to be unsuitable.
Which limits must absolutely be flagged by the AI?
Sizes vary considerably depending on the brand, fit, and materials used. The chatbot is responsible for reiterating these conditions to prevent the recommendation from being interpreted as an absolute guarantee.
In particular, it must emphasize exchange policies, especially if the product is on final sale, customized, or particularly difficult to return. The recommendation should assist in making a choice without hiding the potential risk associated with a non-exchangeable product.
Transparency on risks
Specify fit differences between collections.
Mention the stiffness or elasticity of the material.
Remind about exchange conditions for sale or customized items.
Highlight that the choice remains an estimate based on statistical data.
An honest recommendation helps the customer make an informed decision rather than selling them a dream that would clash with the reality of the garment. This preserves trust and reduces subsequent frustration.
What is the ideal workflow for processing a complex request?
An efficient flow must recommend transparently throughout the interaction. The first step is to identify the product, the category, the associated guide, the fit, and the customer's hesitations.
The chatbot must then collect precise measurements, the comfort preference, and, if authorized, the useful purchase history. It then proceeds to a rigorous comparison with the available data before formulating its response.
Process Steps
Identify the product and the specific fit concerned.
Collect the customer's measurements and style preferences.
Verify the compatibility of the data with existing guides.
Provide an explained recommendation with an alternative if necessary.
Transfer inconsistent or complex cases to a human.
The flow ends with a transfer to a human agent for inconsistent measurements, non-exchangeable products, highly personalized advice requests, or high-volume professional orders. This ensures that each case receives appropriate attention.
What key messages should be used to guide the customer?
The formulation of the chatbot's responses plays a crucial role in the perception of the recommendation. For a standard suggestion, it is necessary to be direct and fact-based: "With your measurements and the cut of this model, size M seems the most suitable."
If the history is used, the tone should invite collaboration: "I can use your previous purchase to refine the recommendation, if you wish." For borderline or hesitant cases, the message should reflect this uncertainty.
Examples of formulations
"As you are between two sizes, the choice depends mainly on the desired look."
"Size S is recommended for a fitted look, but M offers more room."
"Due to the unstructured cut, you can opt for a smaller size."
"We advise you to check the product dimensions before confirming."
These messages help guide the user without rushing them, leaving them with the final responsibility for the choice while ensuring they have all the information to act.
When is it necessary to transfer the conversation to a human?
Although AI is powerful, certain scenarios exceed its capabilities or present business risks that require human intervention. Therefore, handing over is essential in several critical situations.
The chatbot must hand over if the measurements provided are inconsistent with each other, if the product can neither be easily exchanged nor replaced, or if the customer's request requires highly personalized expert advice.
Cases for Handover
Contradictory or imprecise data provided by the customer.
Non-exchangeable products (final sales, customized items).
Highly technical or model-specific advice requests.
Bulk professional orders requiring manual verification.
During the handover, the chatbot must provide a comprehensive summary including the product concerned, the measurements used, the history consulted or not, the proposed recommendation, the alternative, and the remaining doubt. This allows the agent to take over immediately without asking the customer for the information again.
Which performance indicators should be tracked to optimize reliability?
To ensure that the chatbot provides useful and accurate recommendations, it is imperative to track a series of key performance indicators (KPIs) specific to sizing. This allows for the evaluation of the actual impact on conversion and customer satisfaction.
The data to monitor includes the recommendation acceptance rate, returns triggered by incorrect sizing, exchanges avoided thanks to advice, and opt-outs of history usage. This provides a clear vision of the system's effectiveness.
Essential KPIs
The rate of recommendations accepted by customers.
The number of returns due to incorrect sizing.
The volume of exchanges avoided thanks to AI advice.
The rate of history opt-outs by users.
Customer satisfaction after receiving the product (reviews).
This data allows for continuous improvement of sizing guides and verifies that the recommendation does not create false confidence. If the return rate remains high despite the advice, it indicates a problem in the algorithm or the baseline data.
What classic mistakes must absolutely be avoided during implementation?
Certain common practices can harm the chatbot's reputation and customer trust. It is crucial to avoid using purchase history without mentioning it, which could be perceived as a secretive practice.
The bot should never guarantee sizing as an absolute certainty, as this exposes the merchant to unjustified returns if the actual fit differs. Ignoring the specificities of the fit or failing to remind the customer of exchange limitations are also frequent mistakes.
Pitfalls to avoid
Using history without explaining why it is being used.
Promising a 100% suitable size without mentioning the risks.
Neglecting the product's fit in the measurement analysis.
Failing to remind the customer of exchange conditions for specific items.
The chatbot should always advise with accuracy and humility. Recognizing its limitations strengthens the tool's credibility and prevents the customer from feeling deceived by technology that over-promises.
How does Qstomy help to implement this size recommendation?
Qstomy facilitates the integration of these recommendations by connecting the chatbot directly to your store's product catalog, available variants, and specific size guides. This connection allows the bot to access reliable and updated data in real time.
The Qstomy agent can also integrate real-time stock levels and SMS campaigns to inform the customer of immediate availability or offer restock alerts. Support rules are configured so the bot knows exactly when to transfer a sensitive case to a human, thus ensuring complete fluidity.
The Strengths of Qstomy
Seamless connection to the Shopify catalog and variants for precise data.
Integration with social content and UGC to enrich advice.
Intelligent handoff to the support team with a complete summary of the context.
Clear answers without inventing promotions or non-existent sizes.
Qstomy helps the customer move forward in their purchasing process by relying on reliable references. The bot avoids suggesting a size, reference, or availability that would need to be manually confirmed, thus ensuring a transparent and reassuring experience.
What is the checklist before launching your recommendation chatbot?
Before putting your size recommendation solution into production, it is essential to verify that all necessary elements are in place to guarantee a quality service and avoid the pitfalls mentioned earlier.
Check your data
Are the size guides complete, accurate, and up to date for all products?
Are the models' measurements or the actual fit documented?
Are the return rules correctly configured to be displayed by the bot?
Is the customer history accessible via a secure and GDPR-compliant API?
Technical challenges
Is the bot trained to detect inconsistent measurements?
Is the transfer flow to a human configured with a relevant summary?
Are the tracking KPIs activated to measure the return rate due to sizing?
Have you tested the messages with "between two sizes" scenarios?
To go further: Integrating customer service answers into an e-commerce SEO strategy useful to customers - Qstomy, Out of stock on a single size: helping the customer choose between waiting, alternative, and stock alert - Qstomy, Training an e-commerce chatbot with Shopify: using the right data without creating wrong answers - Qstomy, How to manage customer questions about in-store fittings before online purchase - Qstomy, Product seen in a short video: helping the customer find the exact item and verify what is shown - Qstomy, UGC and customer photos: using real proof to answer better without losing context - Qstomy, Name error on an order: correcting what can be corrected before the package gets stuck - Qstomy.
Enzo
September 2, 2026
