Last reviewed 2026-08-05

Methodology, sources, and metric verification

Figures on this page are modelled, not audited client results. Inputs (call volume, ticket size, close rate) are stated next to every number so you can substitute your own. We do not publish named-client outcomes, aggregate ratings, or testimonials we cannot evidence.

For the technology side of the same disclosure — which AI models each system runs on, a models-by-industry load chart, and how those models are grounded, guardrailed, and evaluated rather than fine-tuned — read which AI models we use and how they're trained.

Metric definitions

Every metric we publish, defined

Speed to lead

The elapsed time between a prospect raising their hand (form, call, DM) and the first substantive reply reaching them.

Formula
first_reply_timestamp − inbound_timestamp, measured per lead and reported as a median.
Assumption
An automated reply counts only if it asks a qualifying question or offers a booking action — an autoresponder that says 'we got your message' is not a reply.

Benchmark sources: Harvard Business Review, MIT / InsideSales.

Missed-call recovery rate

The share of unanswered inbound calls that are converted into a booked job through an automated text-back conversation.

Formula
booked_jobs_from_missed_calls ÷ total_missed_calls, per calendar month.
Assumption
Requires a phone system that reports missed calls with caller ID, and a jurisdiction where transactional SMS reply to an inbound caller is permitted.

Benchmark sources: Google / Ipsos small business studies, Twilio Messaging research.

Booked rate

The share of qualified conversations that end with a confirmed appointment on the calendar.

Formula
confirmed_appointments ÷ qualified_conversations.
Assumption
Qualification uses the same three questions before and after install, so the denominator is comparable.

Quote follow-up close rate

The share of sent quotes that convert to signed work after an automated multi-touch follow-up sequence.

Formula
signed_quotes ÷ sent_quotes over a fixed 30-day window from send date.
Assumption
The window is fixed so a longer sales cycle does not inflate the post-install figure.

Review velocity

New public reviews earned per month across your primary review profile.

Formula
new_reviews_this_month, counted on the profile itself rather than on requests sent.
Assumption
Requests are sent to every completed job with no filtering or gating by expected sentiment.

Benchmark sources: BrightLocal, Whitespark.

Cost per booked job

All-in monthly cost of the system divided by the jobs it books.

Formula
(system_cost + tool_subscriptions) ÷ booked_jobs_attributable_to_the_system.
Assumption
Only jobs where the system made first contact are attributed; jobs from referral or repeat customers are excluded.

Payback period

How long the system takes to return its own cost in gross profit.

Formula
total_setup_and_tool_cost ÷ (incremental_jobs_per_month × average_ticket × gross_margin).
Assumption
Uses gross margin, not revenue. Where a margin is not stated on a page, the model assumes the trade's typical margin and says so inline.
How we model

Modelling, attribution, and correction rules

What our numbers are

Every figure on this site is one of three things, and we label which: a metric definition (how a thing is measured), a modelled scenario (arithmetic run on stated inputs), or a cited third-party benchmark with a link to the original research.

We do not publish measured outcomes for named clients, and we do not emit review or rating structured data. If a number is not traceable to inputs on the page or to a source in the list below, treat it as absent — and tell us, because it should not be there.

How a modelled scenario is built

Each case study starts from operating inputs a business owner can check against their own books: monthly inbound volume, the share currently unanswered, average ticket, and gross margin. Those inputs are printed next to the result.

We then apply one behavioural rate — a recovery rate, a booked rate, or a close rate — drawn from the conservative end of the cited research rather than a vendor best case. The arithmetic is written out line by line so any reader can substitute their own numbers.

  • Inputs are stated, never implied.
  • One behavioural assumption per model, taken from the low end of the cited range.
  • Payback uses gross profit, not revenue.
  • No compounding, no year-two projections, no 'up to' figures.

Attribution rules

A job counts toward a system only when the system made first contact or performed the follow-up that produced the booking. Repeat customers, referrals, and inbound calls answered by a human are excluded from system attribution even when the system was live.

Where a metric can be gamed — review counts, for example — we define it on the outcome (reviews visible on the profile) rather than on the activity (requests sent).

Review cadence

This methodology and the source list were last reviewed on 2026-08-05. Sources are re-checked when a cited study is superseded, and metric definitions change only with a note here explaining what changed and why.

Corrections

If you find a figure that does not reconcile with its stated inputs, or a source that no longer supports the claim attached to it, we will correct or remove it. Accuracy matters more to us than a stronger-looking number.

Source list

Research we cite

Numbers appear alongside their inputs on every illustrative build, so you can substitute your own figures.

Check the math against your own numbers

Pick your industry, read the model, and decide from the arithmetic rather than a claim.

Ready to install?
All systems, $49/mo — cancel anytime.
Get access →