SOURCES AND METHODOLOGY
The numbers deserve
some context.
This register identifies the research, customer cases and platform rules referenced on WatsBridge. Each entry explains what the source measures and which limits matter when reading a commercial claim.
Original references checked on 30 October 2026. Email comparison references checked on 1 November 2026. Figures belong to the named sources and populations; external results do not describe WatsBridge performance.
Find the reference behind a claim.
- Baymard Institute — cart abandonment statistics
- GoKwik / Kwik Engage — Xplosive Ape
- WhatsApp Business — 6thStreet
- Fang, Yuan, Zhang, Donati and Sarvary — CESifo Working Paper 12201
- Gorgias — Pepper customer story
- Zendesk — Customer Experience Trends 2026
- Salesforce — State of the AI Connected Customer, seventh edition
- WhatsApp Business — Business Messaging Policy
- Reactivation campaign involving 360 customers
- Mailchimp — measuring email opens and clicks
RESEARCH SYNTHESIS
Baymard Institute — cart abandonment statistics
Current public statistics page, checked 30 October 2026.
Baymard reports a 70.22% average abandonment rate calculated from 50 studies. The underlying studies cover different periods and ecommerce settings. This describes the scale of abandonment; it is not the share a messaging tool can recover.
Baymard’s separate checkout-improvement estimates concern changes to checkout usability. They should not be presented as the expected effect of a WhatsApp reminder.
Related WatsBridge content: Cart recovery
PROVIDER-PUBLISHED CUSTOMER CASE
GoKwik / Kwik Engage — Xplosive Ape
UK retailer; the deployment began in October 2024. The reporting window is not precisely stated.
The provider reports 22.36% of abandoned carts recovered and GBP 17,500 in recovered sales. Its account describes more than 2,000 carts and messages sent for 90% of them.
The precise recovery denominator, attribution window and independently verified incremental effect are not supplied. No randomised holdout is described. The figures therefore remain a reported customer example, rather than a market average or a WatsBridge result.
The publication also uses “ROI” for revenue divided by messaging spend. We do not treat that ratio as net profit.
Related WatsBridge content: External recovery example
PLATFORM-PUBLISHED CUSTOMER CASE
WhatsApp Business — 6thStreet
Comparison described for 2024; fashion retail in Gulf markets.
The case reports six times as many abandoned-cart conversions compared with the push notifications and emails used in the comparison. WhatsApp explicitly labels the results self-reported and says individual results will differ.
The public account does not provide a general recovery percentage or demonstrate that the same multiple applies to another retailer. The figure is a channel comparison within the reported setup.
Related WatsBridge content: Cart recovery context · WhatsApp vs email comparison
RANDOMISED FIELD RESEARCH · WORKING PAPER
Fang, Yuan, Zhang, Donati and Sarvary — CESifo Working Paper 12201
June 2026 version; pre-sale experiment conducted in September–October 2023.
The pre-sale chatbot experiment covers 44,614 consumers on a cross-border retail platform. Compared with no live pre-sale assistance, the AI condition increased sales value by 16.3% and conversion by 21.7% relative to the control mean.
These are relative changes, not percentage-point increases. Tables 4 and 5 report the main estimates. Appendix C compares human and hybrid service: the AI-only condition does not establish statistically significant superiority over human assistance.
This working paper concerns that platform and experimental population. It does not test WatsBridge or establish a general WhatsApp sales uplift.
Related WatsBridge content: CRM and sales assistance
PROVIDER-PUBLISHED CUSTOMER CASE
Gorgias — Pepper customer story
Retail case using live chat and email; publication checked 30 October 2026.
Gorgias reports a 19% conversion rate for AI-assisted sales in the Pepper case. This is an observed rate for the assisted sales population described by the provider. It is not a 19% increase in the store’s total sales or a visitor-wide conversion rate.
The public page does not provide a randomised comparison establishing the incremental effect of the assistant. We retain it as a commercial example, separately from the field research above.
Related WatsBridge content: Sources for sales assistance
VENDOR-COMMISSIONED INTERNATIONAL SURVEY
Zendesk — Customer Experience Trends 2026
Two surveys in 22 countries, June 2025: 6,182 consumers and 5,115 business respondents.
The release reports that 86% of consumers say responsiveness and accurate resolution strongly influence purchase decisions, while 74% are frustrated by having to repeat information.
These are stated preferences from the surveyed population, not measured changes in sales. The survey does not establish a universal deadline for replying to every kind of customer message.
Related WatsBridge content: Customer expectations · Continuity across channels
VENDOR-COMMISSIONED CUSTOMER RESEARCH
Salesforce — State of the AI Connected Customer, seventh edition
Seventh-edition report and public research page, checked 30 October 2026.
The report states that 72% of customers consider it important to know when they are communicating with an AI agent. It describes expectations and attitudes; it does not establish the performance of a particular chatbot.
We use this finding to discuss clarity about automated replies and access to a person. It does not imply that WatsBridge’s optional AI Agent is already available.
Related WatsBridge content: Customer expectations and automation
PLATFORM RULES
WhatsApp Business — Business Messaging Policy
Current public policy, checked 30 October 2026; platform rules can change.
The policy covers eligibility to contact people, opt-in, requests to stop communication and the use of business messaging. It is a rules reference, not evidence of a sales or recovery rate.
A Meta-approved template’s structure and components remain distinct from local text and personalisation values inserted into permitted fields. Local preparation or preview does not establish a new Meta approval.
Related WatsBridge content: Cart recovery setup · CRM and chatbot scope
WATSBRIDGE INTERNAL CAMPAIGN SUMMARY
Reactivation campaign involving 360 customers
Public case account published 30 October 2026; exact send dates are not supplied.
The reported figures are 83.5% gross delivery, 75–85% read rates, a highest reported click rate of 32% and nine unsubscribes. Some customers had not been contacted for more than three years.
Raw exports, exact batch sizes, read/click denominators and the distinction between unique and total clicks remain undocumented in the available summary. No incremental-sales result is established. The case page explains those limits.
Related WatsBridge content: WatsBridge case study
MEASUREMENT DOCUMENTATION
Mailchimp — email opens and clicks
Documentation checked on 1 November 2026.
Mailchimp calculates opens and clicks against delivered emails. Opens depend on tracking images and can be distorted by privacy features and bots. Click reporting also requires attention to automated activity. These definitions help interpret a comparison; they do not establish a WhatsApp market average.
Related WatsBridge content: Comparing channels and results
How we use a percentage.
A reported conversion rate, a relative improvement and an increase in percentage points describe different things. For example, an illustrative 20% relative increase on a 5% starting rate produces 6%, a rise of one percentage point.
For recovery, the denominator might be all abandoned baskets, eligible baskets or customers who received a message. We keep those distinctions visible and do not invent a missing calculation base.
Attributed revenue is linked to an interaction under a chosen rule. Additional revenue requires evidence about what would have happened without that interaction. Neither figure is net profit without accounting for the relevant costs.
Vendor-published cases remain labelled as such. We do not turn them into independent averages, use the named brands to imply a partnership, or present their results as our own.
