Skip to content
Evidence and sources

Citations

Trust

97% of consumers read reviews online. 41% always read them when browsing for businesses.

Source

BrightLocal, Local Consumer Review Survey 2026 · 1,002 US adults

BrightLocal surveyed a representative panel of 1,002 US adults through SurveyMonkey for its Local Consumer Review Survey 2026, published February 2026. 97% said they read reviews online, and 41% said they always read reviews when browsing for businesses (29% in 2025). It measured self-reported reading habits, not a review checked before every business choice. It did not measure an increase in enquiries or ranking.

Back to Trust →

Service context: review management

Capture

53% of mobile visits are abandoned when a page takes over three seconds.

Source

Google / SOASTA

Google and SOASTA reported in 2017 that 53% of mobile site visits are abandoned when load time exceeds three seconds. It concerns load time and abandonment, not a promised conversion gain from a website project.

Back to Capture →

Service context: website design management

Capture

~1% of visitors fill in a contact form. The rest wanted to ask a question, book a time, or get a price, and couldn't.

Source

Formstack form-conversion benchmark, 650k organisations

Formstack’s 2015 report describes approximately 1% conversion for contact-form views in a mobile-responsive subset of its form dataset. It describes 650,000 anonymized users, which may not mean 650,000 organisations. It measured contact-form views, not all website visitors, and did not ask why people chose not to submit.

Back to Capture →

Service context: website design management

Response

62% of calls to small businesses weren't answered by a person: 37.8% went to voicemail and 24.3% got no response.

Source

411 Locals, 2016 · 85 businesses over 30 days · vendor research

411 Locals reported monitoring 85 businesses across 58 industries for 30 days in 2016: 37.8% of calls reached a person, 37.8% voicemail and 24.3% no response. The 62% without a human answer includes voicemail; this vendor sample is not a universal local-business rate.

Back to Response →

Service context: answering service

Response

7× Contact a new lead within an hour and you're nearly seven times as likely to qualify it as when you wait even an hour longer. Among businesses that replied within 30 days, the average reply took 42 hours.

Source

Harvard Business Review, 2011 · 1.25M leads at 42 companies; 42-hour average from an audit of 2,241 US companies

Harvard Business Review (Oldroyd, McElheran and Elkington, 2011) drew on two datasets. The nearly sevenfold figure comes from 1.25 million leads received by 29 B2C and 13 B2B US companies: firms that tried to contact a lead within an hour were nearly seven times as likely to qualify it as firms that tried even an hour later. The 42-hour average comes from a separate audit of 2,241 US companies and counts only companies that responded within 30 days; 23% never responded. The two figures describe different populations and are not a causal promise for this service.

Back to Response →

Service context: chat and dms

Follow-Up

93% of leads that convert are reached by the sixth attempt. The average rep makes 1.3.

Source

Velocify / InsideSales call-attempt research

Velocify analysed nearly 3.5 million leads across hundreds of firms: 93% of leads that converted were contacted within six or fewer call attempts. Separately, InsideSales reported 1.3 average call attempts across 7,960 companies in 2008–2012. These are different samples and do not show that six calls will convert an arbitrary lead.

Additional original sources: Velocify, How Many Calls to Make · InsideSales, 2014 Lead Response Report

Back to Follow-Up →

Service context: lead follow up

Close

28% Coached salespeople win 28% more of their deals. Only 26% get coached even weekly.

Source

CSO Insights 2017 · Salesforce State of Sales 2024

A CSO Insights summary reports organisations with a dynamic, formalised coaching process within a wider enablement framework had win rates 28% above average; this is an association, not a causal gain for each coached seller. A separate Salesforce survey says 26% receive one-on-one coaching at least weekly. The populations and methods differ.

Additional original sources: CSO Insights coaching report summary · Salesforce, sales productivity research

Back to Close →

Service context: sales conversion coaching

Close

93% of homeowners say an instant estimate influences who they hire.

Source

Housecall Pro, 2025

Housecall Pro surveyed 1,040 US homeowners or household decision-makers in October 2025. A separate publisher page states that 93% said instant estimates affect hiring decisions; the exact questionnaire wording and response options are undisclosed. Stated preference cannot establish sales lift or a return from bespoke automation.

Additional original sources: Housecall Pro publisher page stating the 93% result

Back to Close →

Service context: business process automation

Trust

16% higher modeled six-year value for referred customers in a German bank study.

Source

Schmitt, Skiera & Van den Bulte, Journal of Marketing, 2011

Schmitt and colleagues compared 5,181 referred with 4,633 non-referred German bank customers acquired in 2006, observing behavior for about 33 months and modeling six-year value. The 16% difference is an association after controls, not proof that asking for referrals increases introductions or value.

Back to Trust →

Service context: referral program

Follow-Up

98 conversations reopened from about 500 six-month-dormant real-estate leads in one vendor SMS pilot.

Source

SterlingAI real-estate reactivation case study

An anonymous real-estate team and SterlingAI selected about 500 leads dormant at least six months for an SMS pilot. The vendor reports 98 reopened conversations and 23 interested prospects, not completed sales. This is an unaudited self-report with no comparator; it cannot predict a local business’s win rate.

Back to Follow-Up →

Service context: database reactivation

Keep

51% of reported very poor experiences were followed by consumers reducing or stopping spending.

Source

Qualtrics XM Institute, What Happens After a Bad Experience, 2024

Qualtrics surveyed 28,400 consumers across 26 countries in Q3 2023. Across reported very poor experiences, 51% were followed by consumers saying they reduced or stopped spending; the base is not all surveyed consumers. This is self-reported behavior, not measured escalation effectiveness. Separately, Knox and van Oest followed 20,000 new customers of one retailer for 2.5 years and found recorded recovery partly countered complaint-associated churn. The client owns service resolution; Leadio routes concerns.

Additional original sources: Knox & van Oest, complaint recovery research record

Back to Keep →

Service context: review management

Keep

7.4% higher modeled five-month retention likelihood after two extra emails in a US car-wash field experiment.

Source

Wang, Lu & Shi, randomized car-wash email study, 2022

Wang, Lu and Shi randomized 4,077 new US car-wash subscribers; the treatment group received two additional emails 15 days apart. Their model found 7.4% higher likelihood of retention over about five months. The authors also discuss fatigue and differences by customer tier. This is an email finding in a subscription setting, not an SMS result or former-customer win-back rate.

Back to Keep →

Service context: email sms marketing

Keep

27% of former customers targeted with win-back offers returned in one US telecom study.

Source

Kumar, Bhagwat & Zhang, Journal of Marketing, 2015

Kumar, Bhagwat and Zhang studied one US telecom provider: 14,384 of 53,729 targeted former customers returned after win-back offers. Average time since departure was 3.4 months. This observed 27% share has no untreated comparison and does not predict results for a local service business or years-dormant contacts.

Back to Keep →

Service context: database reactivation

Marketing

62% of surveyed US consumers said incorrect online business information would put them off.

Source

BrightLocal, Local Business Discovery & Trust Report, 2023

BrightLocal surveyed 1,138 US consumers in September 2023. The 62% response is stated intent to avoid a business with incorrect online information, not an audited loss or proof of ranking improvement.

Marketing

30% of surveyed US local searchers used social media during their most recent business search.

Source

BrightLocal, Consumer Search Behavior, 2026

BrightLocal asked 1,227 US consumers who had searched for a local business in the prior three months about their most recent search; 30% used social media at some point. This is channel use, not evidence that regular posts caused bookings or sales.

Service context: social media management

Marketing

57% of surveyed B2B buyers preferred email for a salesperson’s first contact.

Source

LinkedIn Sales Solutions / Ipsos, Deep Sales, 2024

An Ipsos online survey for LinkedIn Sales Solutions asked 508 B2B buyers across 15 countries in 2023; 57% preferred email for a salesperson’s first contact. This sponsored preference survey does not measure reply, meeting or sales rates, and says nothing about LinkedIn message performance.

Response

Under 3% of callers pushed to voicemail left a message, in Invoca’s platform data from calls to home-services businesses.

Source

Invoca, How Much Missed Sales Calls Cost Home Services Businesses, 2024 · vendor platform data

Invoca’s May 2024 blog post says its platform data shows less than 3% of callers who get pushed to voicemail leave a message; the same post reports 27% of calls to home-services businesses going unanswered. Invoca sells call-tracking software, and the post gives no sample size or period. It describes calls among Invoca’s home-services customers, not every business, and does not measure how many callers rang back later. The “$1,200 per missed call” sometimes credited to this post is not in it.

Service context: answering service

Response

55% of home-services businesses did not ask leads to buy or book the job on the call, in Invoca’s 2026 benchmark.

Source

Invoca, Home Services Lead Conversion Benchmarks Report, 2026 · vendor platform data

Invoca’s 2026 report uses anonymized calls tracked by its platform across nine home-services sub-industries, drawn from a wider dataset of over 70 million calls in 10 industries; all figures are averages across its customer base and the home-services call count is not stated. 55% of the businesses did not ask leads to buy or book the job. Separately, 52% of calls were answered by a person, 38% of those were leads, and 45% of those leads converted on the call. It describes Invoca customers, not all businesses, and does not compare AI with human answering.

Response

$38,010 median annual pay for US receptionists in 2025 ($18.27 an hour in May 2025).

Source

US Bureau of Labor Statistics, Occupational Outlook Handbook: Receptionists

The BLS Occupational Outlook Handbook lists 2025 median pay for receptionists as $38,010 a year and $18.27 an hour (May 2025), and says most receptionists work full time. This is wage only, not total employment cost, and it is a national median across all industries.

Capture

90% higher bounce probability when mobile page load time increased from one to five seconds in Google/SOASTA research published in 2017.

Source

Google/SOASTA, Mobile Page Speed New Industry Benchmarks, 2017

Google/SOASTA reported in 2017 that increasing mobile page load time from one second to five seconds raised modeled bounce probability by 90% relative to the one-second baseline. This does not mean 90% of visitors bounced, and it does not measure a conversion or revenue gain from a Leadio website project.

Service context: website design management

Local SEO

42% of citations in local AI answers pointed to business websites, and 28.6% to Google Business Profiles, in BrightLocal’s 2026 study.

Source

BrightLocal, Local AI Visibility Study, 2026

BrightLocal tracked about 1,300 business locations over 60 days, running about 200,000 non-branded local searches across ChatGPT, Google AI Mode and AI Overviews and collecting about 1.9 million citations. Citation share describes which sources the tools cited, not why a business was recommended. BrightLocal sells local SEO software.

Service context: local seo

Local SEO

8% of Google searches that showed an AI summary led to a click on a regular result, against 15% without one, in a 2025 Pew Research Center study.

Source

Pew Research Center, 2025

Pew analyzed March 2025 browsing data from 900 US adults who agreed to share it: 68,879 Google searches, 12,593 of which showed an AI summary. Users clicked a link inside the summary on 1% of those visits. It describes this panel’s behavior, not every search or every business.

Service context: local seo

Social media

73% of consumers in Sprout Social’s 2025 survey expect a brand to respond on social within 24 hours or sooner.

Source

Sprout Social Index, 2025

Sprout Social surveyed 4,044 consumers who follow at least five brands on social, across the US, UK, Canada and Australia, in September 2024. The same share, 73%, said they would buy from a competitor if a brand did not respond. These are self-reported expectations, and Sprout sells social media software.

Additional original sources: Sprout Social Index 2025, methodology

Service context: social media management

Cold outreach

0.3% is the spam-complaint rate Gmail tells senders never to reach, with 0.1% advised as the safer ceiling.

Source

Google, Email sender guidelines

Since February 2024, Google has required senders to Gmail to authenticate with SPF or DKIM and keep reported spam below 0.3%. Senders of more than 5,000 messages a day must also publish DMARC and support one-click unsubscribe for marketing mail. Yahoo applies matching rules, and Microsoft began routing non-compliant high-volume mail to Junk in May 2025, with rejection to follow.

Service context: cold outreach

Cold outreach

10 days is the deadline, in business days, for honoring an opt-out under the US CAN-SPAM Act, which makes no exception for business-to-business email.

Source

Federal Trade Commission, CAN-SPAM Act compliance guide

The FTC’s compliance guide says commercial email must use accurate header information and subject lines, include a valid physical postal address and offer a working opt-out that is honored within 10 business days. Penalties apply per violating email.

Service context: cold outreach

Trust

78% of US consumers were asked to leave a review for a business in the past 12 months; 28% say they would “always” write one when asked.

Source

BrightLocal, Local Consumer Review Survey 2026 · 1,002 US adults · vendor survey

BrightLocal surveyed a representative panel of 1,002 US adults through SurveyMonkey for its Local Consumer Review Survey 2026, published 11 February 2026. 78% said they had been asked to leave a review in the past 12 months, and 28% said they would "always" write a review if asked. These are self-reports, not measured review rates. The same page gives two different figures (65% and 83%) for how many people who were asked went on to write one, so Leadio does not cite either. BrightLocal sells local SEO software.

Trust

47% of US consumers say they won’t use a business with fewer than 20 reviews.

Source

BrightLocal, Local Consumer Review Survey 2026 · 1,002 US adults · vendor survey

BrightLocal surveyed a representative panel of 1,002 US adults through SurveyMonkey for its Local Consumer Review Survey 2026, published 11 February 2026. 47% said they won’t use a business with fewer than 20 reviews, and 9% said they would use one with five or fewer. This is stated preference, not measured behaviour, and it covers local businesses in general. BrightLocal sells local SEO software.

Trust

10 companies received FTC staff warning letters on December 22, 2025 about possible violations of the Consumer Review Rule.

Source

Federal Trade Commission press release, December 22, 2025

FTC staff sent letters to 10 companies based on consumer complaints and information from the companies. The FTC says the letters are not formal determinations that the recipients violated the rule. The rule (16 CFR Part 465) took effect on October 21, 2024.

Trust

$4.2 million paid by online retailer Fashion Nova in 2022 to settle FTC allegations that it suppressed reviews rated below four stars on its own website.

Source

Federal Trade Commission press release, January 25, 2022

The FTC alleged that from late 2015 to November 2019 Fashion Nova automatically posted four- and five-star reviews and held lower-rated ones for approval that never came. The settlement was under the FTC Act, before the 2024 review rule existed. It concerns hiding reviews on a business’s own website, not which customers were asked for reviews.

Response

60× Firms that tried to contact a lead within an hour were more than 60 times as likely to qualify it as firms that waited 24 hours or longer.

Source

Harvard Business Review, 2011 · 1.25M leads at 29 B2C and 13 B2B US companies

Same HBR article and companion dataset as the nearly sevenfold figure: 1.25 million sales leads received by 29 B2C and 13 B2B US companies. "Qualify" means having a meaningful conversation with a key decision maker. It is 2011 web-lead data, not Facebook lead ads, and the multiples describe those firms, not a forecast.

Response

45% of leads who reached a person converted on the call, in Invoca’s 2026 home-services benchmark.

Source

Invoca, Home Services Lead Conversion Benchmarks Report, 2026 · vendor platform data

Invoca’s 2026 report uses anonymized calls tracked by its platform across its home-services customers, drawn from a wider dataset of over 70 million calls in 10 industries; the home-services call count is not stated. 52% of calls were answered by a person, 38% of those were leads, and 45% of those leads converted on the call. It describes Invoca customers, not all businesses, and does not compare AI with human answering.

Response

73% of calls over 30 seconds to home-services businesses were answered by a person in Invoca’s 2026 platform data, against 65% of calls over 15 seconds and 52% of all calls.

Source

Invoca, Home Services Lead Conversion Benchmarks Report, 2026 · vendor platform data

Invoca’s 2026 report averages anonymized calls tracked by its platform across nine home-services sub-industries, within a wider dataset of over 70 million calls in 10 industries and 7 marketing channels; the home-services call count is not stated. 52% of all calls were answered by a person, 65% of calls over 15 seconds and 73% of calls over 30 seconds, with sub-industry answer rates from 32% to 74%. It covers Invoca customers and several marketing channels, not Local Services Ads specifically. Invoca sells call-tracking and AI agent software.

Response

80% of consumers said they were willing to use AI-powered customer service, while two-thirds still preferred speaking with a person, in Five9’s 2026 survey.

Source

Five9, 2026 Business Leaders Customer Experience Report (press release) · 3,000 consumers, US, UK, Germany · vendor survey

Five9, which sells contact-center software including AI agents, released the survey on 24 June 2026. Hanover Research surveyed 3,000 consumers and 600 CX and contact-center decision-makers in the US, UK and Germany in April 2026. It asked about customer service in general, not calls to small businesses or after-hours calls, and does not measure hang-up rates.

Response

31 in 100 calls to dental practices hung up before reaching an agent, in Patient Prism’s analysis of 11.5 million calls in 2025.

Source

Patient Prism, Dental Patient Access Report, 2025 data · 11,552,668 calls, 8,280 locations · vendor platform data

Patient Prism, which sells call analytics to dental practices, analyzed 11,552,668 patient calls to 8,280 dental practice and DSO locations in calendar year 2025. Its funnel shows 31 of every 100 calls hung up before reaching an agent. It covers dental practices only; the page does not break the figure down by cause and does not measure AI answering.

Response

Up to 80% of callers chose texting over voicemail, among Numa’s own customers.

Source

Numa, 22 Business Phone Statistics · vendor, upper bound from its customers

Numa sells phone answering and texting for businesses. Its post says the share of callers who text varies by industry and that "our customers have seen up to 80 percent of callers choose texting over voicemail", attributed to its own 2020 data. It is an upper bound from Numa customers, not an average or an independent study.

Response

48% of calls Americans received on their phones were picked up in the first half of 2019.

Source

Hiya, State of the Phone Call H1 2019 · call data including spam, vendor research

Hiya analyzes more than 13 billion calls a month globally. Its H1 2019 report says "only 48 percent of calls Americans receive on their phones are picked up", against a background of robocalls; the figure includes spam calls and is not specific to business callbacks.

Response

$395 a month for 100 minutes of Ruby’s live virtual receptionist service (50 minutes for $250), billed per minute and rounded up, as published in October 2026.

Source

Ruby, Pricing, and Per-minute vs per-call virtual receptionist pricing, 2026 · vendor list price

Ruby’s pricing page lists virtual receptionist plans of 50 minutes ($250), 100 minutes ($395), 200 minutes ($720) and 500 minutes ($1,725) a month. Its per-minute vs per-call comparison page says it bills per minute, rounded up, and gives both 30-second and 60-second increments in two copies of the same table. Overage rates and what counts toward minutes are not stated. These are list prices that change; they say nothing about booking rates.

Response

$2.00 a call on Smith.ai’s AI receptionist Pro plan ($150 a month for 75 calls; extra calls $2.50; free plan of 25 calls), as published in October 2026.

Source

Smith.ai, AI Receptionist pricing, 2026 · vendor list price

Smith.ai’s AI receptionist pricing page lists a free plan (25 calls a month, extra calls $3.00), Pro at $150 a month with 75 calls ($2.00 each, extra $2.50; larger tiers at $1.80 and $1.67), and Enterprise from $500. Billing is per call, not per minute; known spam is filtered before answering and up to 10% of calls can be removed as spam each cycle. Month to month with 30 days’ notice. List prices that change; no booking outcomes are stated.

Response

30.0% of private-industry employer compensation costs went to benefits in June 2026; wages and salaries were the other 70.0%.

Source

US Bureau of Labor Statistics, Employer Costs for Employee Compensation, June 2026

The BLS Employer Costs for Employee Compensation release for June 2026 reports private-industry wages and salaries averaging $32.82 an hour (70.0% of employer costs) and benefits $14.07 (30.0%). It is a share of total compensation across all private-industry workers, not a markup on a receptionist’s wage, and is cited only to correct the misread "29.4% on top of wages" figure.

Follow-Up

39% lower likelihood of ever contacting a lead that was texted before phone contact; leads texted only after first contact converted at more than twice the average contacted lead’s rate.

Source

Velocify, Text Messaging for Better Sales Conversion, 2012 · ~3.5M lead records, 400+ companies · vendor research

Velocify analyzed almost 3.5 million lead records from more than 400 randomly selected Velocify customers, all generated in the first half of 2012. Texting before phone contact decreased the likelihood of ever contacting the lead by 39%; leads texted only after initial contact had the highest conversion (charted at 112.6% above average). These are lead records from a wide variety of industries worked by sales teams, not callers who had just phoned a business, and the study does not describe controls for lead quality.

Follow-Up

60% average response rate across 163,212 HVAC estimate follow-up campaign launches on Hatch’s software; the best campaign reached 90.06% with 5 texts and 2 emails over 7 days.

Source

Hatch, We Analyzed 163K HVAC Estimate Follow-Up Campaigns, 2024 · vendor platform data

Hatch, which sells follow-up software, published (23 Sep 2024) an analysis of 163,212 estimate follow-up campaign launches by HVAC businesses on its platform; campaigns needed at least 100 launches and 100 responses to be included, and all were set to start two days after the estimate. "Response rate" is not defined beyond customers responding, so it counts replies of any kind, not booked or signed jobs. It is HVAC only, has no untreated comparison group, and does not measure close rate. The same article repeats an untraced "8–12 touches to close 80% of deals" line from an outside source, which Leadio does not use.

Close

26.9% of 53,389 recorded in-home sales consultations ended in what Siro judged a commitment (Midwest 31.1%, West 27.2%, Northeast 24.3%, South 23.3%; Windows & Doors 29.1%, Decks & Patios 16.0%).

Source

Siro, In-Home Insider: Benchmarks, 2026 · 53,389 recorded consultations, Apr–Aug 2026 · vendor data

Siro, which sells conversation-intelligence software for in-person sales, analysed 53,389 in-home sales consultations recorded on its platform between April and August 2026. Its close rates "reflect Siro’s own judgment of whether a conversation advanced toward a sale, not a signed contract or revenue figure", and Siro says to treat a two-point gap between groups as noise and a twenty-point gap as real; it also says outcome labels skew toward under-counting wins. National and regional figures use the full set; trade and conversation-level figures use a geographically balanced sample. It describes companies that record appointments on Siro, over five months, and its comparisons are correlations, not proof that changing one behaviour changes a close rate.

Close

30.9% of recorded in-home appointments never confirmed who the decision-makers were; they closed at 23.0%, against 36.7% where all decision-makers were confirmed.

Source

Siro, In-Home Insider: Kitchen Table, 2026 · vendor data

Siro, which sells conversation-intelligence software for in-person sales, analysed 53,389 in-home sales consultations recorded on its platform between April and August 2026. Its close rates "reflect Siro’s own judgment of whether a conversation advanced toward a sale, not a signed contract or revenue figure", and Siro says to treat a two-point gap between groups as noise and a twenty-point gap as real; it also says outcome labels skew toward under-counting wins. National and regional figures use the full set; trade and conversation-level figures use a geographically balanced sample. It describes companies that record appointments on Siro, over five months, and its comparisons are correlations, not proof that changing one behaviour changes a close rate.

Close

44% of concessions in recorded in-home appointments were unprompted; they closed 18.8% of the time, against 23.3% for concessions given in response to the homeowner. 73.7% of conversations included at least one concession (1.85 on average).

Source

Siro, In-Home Insider: Kitchen Table, 2026 · vendor data

Siro, which sells conversation-intelligence software for in-person sales, analysed 53,389 in-home sales consultations recorded on its platform between April and August 2026. Its close rates "reflect Siro’s own judgment of whether a conversation advanced toward a sale, not a signed contract or revenue figure", and Siro says to treat a two-point gap between groups as noise and a twenty-point gap as real; it also says outcome labels skew toward under-counting wins. National and regional figures use the full set; trade and conversation-level figures use a geographically balanced sample. It describes companies that record appointments on Siro, over five months, and its comparisons are correlations, not proof that changing one behaviour changes a close rate.

Close

15.1% close rate when the homeowner wanted to get more or compare quotes, the lowest-closing objection; "can’t afford to pay" closed at 35.5% and "sticker shock" at 32.2%, against a 25.6% median.

Source

Siro, In-Home Insider: Kitchen Table, 2026 · vendor data

Siro, which sells conversation-intelligence software for in-person sales, analysed 53,389 in-home sales consultations recorded on its platform between April and August 2026. Its close rates "reflect Siro’s own judgment of whether a conversation advanced toward a sale, not a signed contract or revenue figure", and Siro says to treat a two-point gap between groups as noise and a twenty-point gap as real; it also says outcome labels skew toward under-counting wins. National and regional figures use the full set; trade and conversation-level figures use a geographically balanced sample. It describes companies that record appointments on Siro, over five months, and its comparisons are correlations, not proof that changing one behaviour changes a close rate.

Close

22.1% stall recovery rate when the seller secured a concrete next step, the highest of the responses Siro measured, though it closed only 17.7% on the call; the urgency-and-discount reflex closed at 21.7%, while reframing financing and reducing commitment risk both closed above 32%; top-tier reps converted nearly 1 in 5 stalls, bottom-tier about 1 in 9.

Source

Siro, In-Home Insider: Kitchen Table, 2026 · vendor data

Siro, which sells conversation-intelligence software for in-person sales, analysed 53,389 in-home sales consultations recorded on its platform between April and August 2026. Its close rates "reflect Siro’s own judgment of whether a conversation advanced toward a sale, not a signed contract or revenue figure", and Siro says to treat a two-point gap between groups as noise and a twenty-point gap as real; it also says outcome labels skew toward under-counting wins. National and regional figures use the full set; trade and conversation-level figures use a geographically balanced sample. It describes companies that record appointments on Siro, over five months, and its comparisons are correlations, not proof that changing one behaviour changes a close rate.

Close

76% of recorded in-home appointments included "wants to make an informed decision", associated with a close rate 20.4 points lower; "lower the total price" came up in nearly half and was associated with +0.5 points.

Source

Siro, In-Home Insider: Homeowners, 2026 · vendor data

Siro, which sells conversation-intelligence software for in-person sales, analysed 53,389 in-home sales consultations recorded on its platform between April and August 2026. Its close rates "reflect Siro’s own judgment of whether a conversation advanced toward a sale, not a signed contract or revenue figure", and Siro says to treat a two-point gap between groups as noise and a twenty-point gap as real; it also says outcome labels skew toward under-counting wins. National and regional figures use the full set; trade and conversation-level figures use a geographically balanced sample. It describes companies that record appointments on Siro, over five months, and its comparisons are correlations, not proof that changing one behaviour changes a close rate.

Close

75 min median in-home appointment (quartiles 46 / 75 / 113); appointments of 113 minutes or more, against 46 or less, were 2.35x as likely to include a real close attempt, 1.79x as likely to include a financing conversation and 6.3x as likely to reach deeper discovery.

Source

Siro, In-Home Insider: Benchmarks, 2026 · vendor data

Siro, which sells conversation-intelligence software for in-person sales, analysed 53,389 in-home sales consultations recorded on its platform between April and August 2026. Its close rates "reflect Siro’s own judgment of whether a conversation advanced toward a sale, not a signed contract or revenue figure", and Siro says to treat a two-point gap between groups as noise and a twenty-point gap as real; it also says outcome labels skew toward under-counting wins. National and regional figures use the full set; trade and conversation-level figures use a geographically balanced sample. It describes companies that record appointments on Siro, over five months, and its comparisons are correlations, not proof that changing one behaviour changes a close rate.

Keep

Three-quarters of surveyed consumers said they would unsubscribe from a business’s promotional texts if they received too many; 74% said they would be more likely to respond to a text if they knew a real person was on the other end.

Source

Podium, Local Business Messaging Trends 2021, via Small Business Trends · 1,004 consumers, US and Australia · vendor survey

Podium, which sells business messaging software, surveyed 1,004 consumers in the US and Australia for its 2021 Local Business Messaging Trends report; figures are relayed by Small Business Trends (27 Jul 2021), which says "three-quarters" rather than an exact 75%. Stated preferences, not measured unsubscribe or reply behaviour; the original report was not opened.

Business Brain

Over 60% fewer hallucinated responses from the best retrieval-augmented chatbots than from non-retrieval models, in automatic and human evaluations on general-knowledge conversation tasks.

Source

Shuster, Poff, Chen, Kiela & Weston (Facebook AI Research), Retrieval Augmentation Reduces Hallucination in Conversation, Findings of EMNLP 2021 · peer-reviewed

Shuster and colleagues compared retrieval-augmented dialogue models with non-retrieval baselines on knowledge-grounded conversation tasks (Wizard of Wikipedia, CMU_DoG). The "over 60%" reduction in hallucinated responses is from automatic and human evaluation, stated in the paper's introduction (not the abstract). The paper's 70%/85% figures are knowledgeability gains, not hallucination reduction. It did not test small-business calls or chat.

Business Brain

19% of working hours spent "trying to track down information needed to complete tasks" by interaction workers at large organizations, in McKinsey Global Institute's 2012 analysis.

Source

McKinsey Global Institute, The social economy, 2012 · analysis built on IDC data

McKinsey's figure describes "interaction workers" at large organizations and rests on proprietary McKinsey data plus IDC's 2009 estimates (footnote 60). It is not a small-business measure. The widely repeated "1.8 hours a day / 9.3 hours a week (McKinsey)" is not in this report.

Business Brain

1.8%–24.2% of document summaries contained content not in the source, across the 108 models on Vectara's hallucination leaderboard (updated September 22, 2026).

Source

Vectara, Hallucination Leaderboard (HHEM), 2026 · vendor benchmark

Vectara asks each model to summarize short documents using only facts in the document, then scores factual consistency with its own HHEM model. It measures summarization, not question answering, and the dataset is not public. The range runs from 1.8% (antgroup/finix_s1_32b) to 24.2% (mistralai/ministral-3-3b-2512). Secondary sites publish conflicting "Vectara" figures; cite only the GitHub table with its date.

Business Brain

$650.88 in damages, plus interest and fees, ordered against Air Canada after its website chatbot wrongly said bereavement fares could be claimed retroactively.

Source

BC Civil Resolution Tribunal, Moffatt v. Air Canada, 2024 BCCRT 149 · decision, Feb 14, 2024

A British Columbia small-claims tribunal rejected Air Canada's argument that its chatbot was "a separate legal entity", found the airline "did not take reasonable care to ensure its chatbot was accurate", and ordered $812.02 in total ($650.88 damages, $36.14 interest, $125 fees). The correct policy was on another page of the same site. It is not binding precedent outside that tribunal.

Local SEO

6th is where a high Google star rating (4–5) ranks among 187 Google Maps ranking factors scored by 47 local search experts, behind primary category, proximity to the searcher, keywords in the business name, address in the city searched and business hours.

Source

Whitespark, Local Search Ranking Factors 2026 (published November 2025) · expert survey

Whitespark asked 47 local search experts to score 187 factors for their impact on the local pack and Google Maps (also local organic, conversion and AI visibility). Top ten Maps scores: primary category 227, proximity 225, keywords in business title 223, physical address in city of search 213, open at time of search 189, high numerical Google rating 181, address showing (not a service-area business) 176, additional categories 173, quantity of native Google reviews with text 170, map pin placement 165. Recency of reviews ranks 11th (164) and a sustained influx of reviews 14th (154). This is expert opinion, not measured ranking data, and Whitespark sells local SEO software and services.

Local SEO

8,186 businesses in 200 cities analysed by Sterling Sky and Places Scout, which concluded that “the number of reviews you’ve gotten this month matters more than your total number of reviews.”

Source

Sterling Sky / Places Scout, What Gets You Ranking for “Near Me” in 2025 · agency study, correlational

Sterling Sky (an agency) and Places Scout captured the top 10 local pack results for five "near me" searches (plumber, HVAC, locksmith, garage door repair, lawn care) in 100 large, 50 mid-sized and 50 small US cities, 8,186 businesses in all, published 5 November 2025. Findings are correlations: recent reviews mattered more than total count, reviews with text more than rating-only, and hiding the address correlated with lower rank. It also reports one dental client that got 60+ reviews a month, received none for 18 days, and saw rankings drop. Single-client tests are anecdotes, not proof of cause.

Local SEO

1st to 7th is how far one business fell in Sterling Sky’s account after reports led Google to remove keywords from its business name.

Source

Sterling Sky, Do Keywords in Business Names on Google Maps Impact Ranking? (2023) · agency case examples

Sterling Sky (Joy Hawkins, 2023) reports a test in which adding a keyword to a restaurant’s Google Business Profile name raised its ranking for that keyword, and describes three cases where Google removed keywords from competitors’ names after reports, after which their rankings dropped (one drug rehab business from 1st to 7th). Google’s guidelines say unnecessary information in a business name could get the profile suspended. These are agency case examples, not a controlled study.

Social media

82% vs 45% 82% of ad executives believed Gen Z and Millennial consumers felt positive about AI-generated ads; 45% of those consumers said they did.

Source

IAB with Sonata Insights, The AI Ad Gap Widens, Jan 2026 · 505 consumers, 104 executives

IAB and Sonata Insights surveyed 505 US Gen Z (16–27) and Millennial (28–43) consumers who engaged with ads, and 104 US ad executives at companies spending at least $1 million a year on media. Gen Z were more negative than Millennials (39% vs 20% very or somewhat negative). It measures attitudes to AI-generated ads in general, not to a business owner’s own AI clone, and not older buyers.

Social media

73% of Gen Z and Millennial consumers said knowing an ad was made with AI would increase or make no difference to their likelihood to buy.

Source

IAB with Sonata Insights, The AI Ad Gap Widens, Jan 2026 · 505 US consumers

Same IAB survey. Self-reported purchase intent, not measured behaviour. IAB also reports clear disclosure as the third-highest driver of attention after high-quality visuals and funny content, and that fewer than half of advertisers using generative AI always disclose.

Social media

No significant difference in information retention, engagement or trust between a real person and their hyper-realistic avatar, in a 290-participant online experiment.

Source

Schiefelbein, Hammond & Hevner, Information Systems and Neuroscience (Springer), 2025

Schiefelbein, Hammond and Hevner (University of South Florida) compared videos delivered by a real person, by their hyper-realistic avatar, and by the avatar with disclosure, in a B2B context; DOI 10.1007/978-3-031-71385-9_23. They also report a significant (p = 0.003) difference in information retention between the avatar and avatar-with-disclosure videos. Separately, Schiefelbein told Bay News 9 (1 Feb 2024): "When you don't disclose and someone finds out, they feel fooled." The "nearly 85% unhappy" figure in that coverage is a reporter paraphrase with no published table and is not used.

Social media

54–61% unaided brand recall across four AI-disclosure label styles, against 54% for the unlabeled ad, in a MediaScience test with 900 US participants reported in June 2026.

Source

MediaScience AI Labeling Impact Study, as reported by Marketing Dive, June 2026 · secondary

MediaScience, with MediaPet and the Ehrenberg-Bass Institute, tested four label styles (text in first 3 seconds, text at seconds 4–6, continuous text, continuous icon) against an unlabeled control. Per Marketing Dive (12 Jun 2026): recall 54% control and icon, 60–61% text labels; recognition 87–92% vs 88% control; positive brand attitude 44–51% vs 51% control (icon lowest). We could not open the original report or eMarketer’s coverage (403). Paid video ads, not organic owner videos.

Social media

71.6% of small business owners surveyed had recorded a video for their business and decided not to post it; the most-named reason was that it did not look professional enough (53.9%).

Source

HeyGen, The State of AI Avatars 2026, Sept 2026 · 1,000+ owners, 93 countries · vendor survey

HeyGen surveyed 1,000+ small business owners and solopreneurs (solo to 99 employees) across 93 countries; method, weighting and question wording are not published. Reasons for not posting sum to over 100%, so multiple answers were allowed: did not look professional enough 53.9%, did not like how they looked 48.6%, how they sounded 35.4%, changed mind on content 33.8%. HeyGen sells AI avatar software.

Growth System

1.08% / 7.9% of revenue: the average advertising spend (Small Business Trends) and 2018 average marketing spend (Web Strategies) that the SBA’s budgeting page reports. The page makes no 7–8% recommendation.

Source

U.S. Small Business Administration, How to Get the Most From Your Marketing Budget, 2019 · government page citing secondary data

The SBA blog post cites Small Business Trends for 1.08% of revenue spent on advertising (retail about 4%, restaurants 1.93%) and Web Strategies for 7.9% average marketing spend in 2018. Both are other publishers’ averages, not SBA research or advice. The widely repeated "SBA recommends 7–8% of revenue" does not appear on the page (checked 9 Oct 2026).