Methodology

Numbers without methodology are marketing. This page documents how we define the metrics behind claims on Auto-Respond pages, and how we source any hard statistic we publish. Our rule is simple: a number on our site is either computed from a defined internal metric in your own dashboard, or it is attributed to a named public study. If you spot a figure that is neither, treat that as a bug and tell us.

Start your 7-day free trial if you want to see the dashboards that compute these metrics for your own account.

What we measure

Two core metrics, computed per account in your dashboard, define what we mean when we talk about speed and results:

Response time. The interval between an inbound lead webhook arriving at Auto-Respond and the first AI-generated reply being sent on the appropriate channel (Thumbtack message, Yelp message, Facebook message, SMS, voice call, etc.).
Booked appointments. Calendar events accepted as jobs by the contractor within seven days of the originating lead, as defined below.

We do not publish aggregate first-party conversion percentages or ROI multiples as if they were measured across a customer base — we have no such audited dataset, and stating one would be false advertising. Where a page needs a conversion or speed-to-lead statistic, it cites named public research (see below). We also do not measure or claim "leads generated" — that is the platform's job (Thumbtack, Yelp, etc.), not ours. Our scope starts the moment a lead enters our system.

How we define response time

Response time is the wall-clock interval, measured to the second, between two events:

T0: The inbound lead webhook is received by Auto-Respond. For Thumbtack and Yelp this is the official platform webhook timestamp. For Facebook Lead Ads it is the moment the lead form payload is delivered. For Google LSA it is the moment the lead notification reaches our integration.
T1: The first outbound message generated and sent by our AI on the corresponding channel.

Response time = T1 − T0. The metric does not include any latency that occurred before T0 (i.e., time the lead sat inside the originating platform before being delivered). We cannot observe that latency, so we exclude it from our claims rather than estimate around it.

How we define a booked appointment

A booked appointment is a calendar event that meets all four conditions:

Created on the contractor's connected Google Calendar or Microsoft 365 calendar through our integration.
Linked to a customer record sourced from a tracked inbound lead.
Confirmed by the contractor (i.e., not auto-cancelled, not flagged as a misfire).
Scheduled within seven days of the originating lead arrival.

We deliberately exclude "warm replies," "interested but not booked," and "callback requested without confirmed time." Those are softer signals and we choose not to count them as conversions, even though doing so would inflate our headline numbers.

Public sources we cite

When a page needs a conversion or speed-to-lead statistic, it draws from named, public research and attributes the source inline. The studies we rely on:

Speed-to-lead (5-minute rule). Waiting from 5 to 30 minutes to respond makes you 100x less likely to contact a lead and 21x less likely to qualify it. (MIT — Oldroyd, Lead Response Management Study, 2007.)
First-hour decay. Contacting a lead within an hour makes you about 7x more likely to qualify it than an hour later, and roughly 60x vs 24+ hours. Of 2,241 U.S. companies audited, only 37% responded within an hour and 23% never responded at all. (Harvard Business Review, "The Short Life of Online Sales Leads," 2011.)

We attribute each figure to its actual source and do not reword the numbers. We never present these as our own measured results, and we never invent a multiple of our own to sit alongside them.

What we will not state

We do not publish first-party percentages, cohort sizes, or ROI dollar figures presented as measured across a customer base, because we do not have an audited dataset to support them. That means no invented "X% of our customers booked," no "our March cohort of N leads," and no "$N,NNN/month net" claims. Where the honest answer is qualitative, we keep it qualitative — "more booked jobs," "recovers calls that would otherwise be missed," "a fraction of the cost of a receptionist" — rather than dress it up as a statistic.

Platform coverage

Auto-Respond ingests leads from Thumbtack, Yelp Request a Quote, Facebook Messenger, Facebook Lead Ads, Instagram DMs, Google Local Services Ads, Angi, Nextdoor, Porch, and Houzz. The dashboard metrics you see reflect the platforms where we have native integrations and where data flows through our system end to end. We do not extrapolate to platforms we do not integrate with.

What your dashboard excludes

When you read your own response-time and booked-appointment metrics in the dashboard, the following are excluded so the numbers reflect real customer activity:

Test events. Calls and messages you generate while testing your own setup.
Internal traffic. Leads from staff or related parties, identified by phone number, email, or IP heuristics.
Channels we do not natively integrate with. If you manually copy-paste leads from a platform we do not connect to, those leads are not tracked.

Why we stay qualitative or cite a source

Per-business variance is large. Two HVAC contractors at the same Thumbtack lead volume can see meaningfully different results from the same auto-responder configuration, depending on service area density, average ticket size, competitive intensity, and seasonality. Publishing a single first-party percentage would imply a precision and a dataset we do not have.

So we do one of two things. Either we describe the benefit qualitatively ("more booked jobs," "faster than replying by hand"), or we cite a named public study with its real figure attached. We do not split the difference by inventing a number and calling it ours.

Update cadence

The public sources we cite are dated in the surrounding sentence (e.g., "HBR, 2011"), so the vintage of any statistic is always visible. We review the claims on our pages periodically to confirm each one is either a defined dashboard metric or an attributed public figure.

If you find a statistic on our site that is neither qualitative nor attributed to a named source — that is a content bug. Tell us and we will fix it.

Where this methodology is referenced

This page is linked from:

The comparison table footnotes on our Thumbtack auto-responder page.
FAQ answers on the AI receptionist pillar.
Hero stat citations on the lead response automation pillar.

Frequently asked questions

Why do you publish ranges and qualitative claims instead of single percentages?

Per-business variance is high. A roofer in Phoenix and an HVAC contractor in Boston do not see the same lift from the same change in response time. Single point estimates would imply a precision nobody has. Where we cite a hard number we attribute it to a named public study; otherwise we keep claims qualitative ("more booked jobs," "faster than manual") rather than invent a figure.

What counts as "response time" at Auto-Respond?

Response time is measured from the moment we receive an inbound lead webhook (from Thumbtack, Yelp, Facebook, Google LSA, Angi, Nextdoor, Instagram, etc.) to the moment our system sends the first AI-generated reply. It does not include the time the lead sat in the originating platform before being delivered to us, which we cannot observe.

How do you define a "booked appointment"?

A booked appointment is a calendar event created on the contractor's connected Google Calendar or Microsoft 365 account, with a confirmed slot and a customer record attached. We do not count "interested replies" or "warm leads" — only calendar events that the contractor has accepted as actual jobs.

Which statistics on your pages are from outside research?

Any conversion or speed-to-lead figure on our pages is attributed to a named public source — for example MIT (Oldroyd, 2007) or Harvard Business Review (2011). We do not present invented first-party percentages, cohort sizes, or ROI multiples as if they were measured results.

What do you do when you do not have data for a claim?

We stay qualitative. If we cannot back a number with a defined internal metric or a named public study, we describe the benefit in words ("books more jobs," "recovers calls that would otherwise be missed") instead of stating a figure we cannot support.

Want to see the dashboard yourself?

Start your 7-day free trial and run the same response-time and conversion measurements on your own lead flow.