Is your automation actually working? The numbers to watch

Is your automation actually working? It is if it moves four kinds of numbers: revenue it booked, cancels it saved, AR days it recovered, and hours it gave back to the office. That is the entire test, and it is the one we apply to every stack we audit. Everything else on the dashboard (sends, opens, clicks) is diagnostic, useful for finding out why a sequence stopped performing but useless for deciding whether it should exist. Most owners we meet track the diagnostics and skip the outcomes, which is how a company ends up proud of a 42% open rate on a win-back series that has not actually won anyone back since last spring.

What is the metrics ladder?

Every sequence has the same funnel, and each rung answers a different question:

• Sent and delivered: is the plumbing working? A delivery rate under roughly 97% on email or a sudden dip on SMS means a list or carrier problem, not a marketing problem.
• Opened and clicked: is the message landing? Directionally useful, but treat opens with suspicion since Apple's privacy features inflate them; clicks and replies are the honest engagement signals.
• Replied: is it starting conversations? For lawn care, replies are often the real conversion event: a reply to a quote follow-up is a sales call waiting to happen.
• Booked: did money move? A program sold, a balance paid, a visit rescheduled, a cancel reversed. This is the only rung that justifies the sequence's existence.

The discipline is reading the ladder top-down only when the bottom rung disappoints. Booked revenue fell: was it delivery, engagement, or close? That is the entire diagnostic method.

What does booked revenue per sequence mean?

Attribute a sale to a sequence when the customer bought the thing the sequence offered within a defined window of receiving it: 30 days is a fair default for upsells, longer for win-backs. This requires your automation platform to see RealGreen sales data, which is an argument for running automations on a synced database rather than a disconnected email tool; the sale posts in RealGreen, the sync carries it back, and the sequence gets its credit without anyone keying anything. Perfect attribution is not available and chasing it is a trap. Consistent attribution is available, and consistency is what lets you compare March to August and sequence A to sequence B.

What is the headline number for each kind of sequence?

Revenue sequences: booked dollars per send

Upsells, cross-sells, prepay, quote follow-up. Illustrative math, plug in your own: a quote follow-up cadence touches 70 stale estimates a month; if 11 close at a $480 average program, that is roughly $5,300 a month the cadence gets credit for, minus whatever a human follow-up habit was already closing (see cannibalization below).

Save sequences: save rate

Cancel-saves and at-risk interventions are measured as saves divided by attempts, tracked by cancel reason. If 25 households a month hit the cancel moment and the sequence keeps 5 on the books, your save rate is 20%, and the annualized revenue is those 5 times your average program value, a number that quietly rivals your best acquisition channel. Measuring it requires logging the attempt even when it fails, which offices forget; the setup details are in our cancel-save post.

AR sequences: recovered days

For dunning, the headline is days-sales-outstanding on the receivables the ladder touches, plus percent of balances cleared without a human call. If average time-to-paid on past-due invoices drops from 38 days to 24 after the ladder goes live, the sequence is funding itself in cash-flow terms regardless of what any email metric says.

How do you tell incremental from cannibalized?

The uncomfortable question: would some of these customers have bought, paid, or stayed anyway? Yes, some would. Prepay discounts taken by people who prepaid every year regardless are the classic case. You do not need an academic answer; you need an honest one. The practical tools: compare against your own history (what did prepay uptake look like the season before the sequence existed), hold out a small random slice for your biggest-dollar sequences (send to 90%, compare the 10%), and ask the office who on the booked list was already in motion. If a sequence's claimed revenue survives a haircut for the obvious would-have-boughts, keep it; if the entire number evaporates, the sequence is decorating sales that were happening anyway, and its send slot should go to something else.

What does the monthly one-pager look like?

One page, every sequence on one row: name, audience size, sends, replies, headline outcome number (booked dollars, save rate, or AR days), and a one-line note from whoever owns it. Formatting is nothing; the ritual is everything. The page gets reviewed once a month next to your operating numbers, and any sequence with a headline number of zero for two straight months gets fixed or killed. Keep it separate from the owner's weekly business scorecard (that job belongs to the 12 numbers every RealGreen owner should check weekly); this page answers one narrower question: is the machine we built earning its keep? Where the page lives is up to you: synced RealGreen data can feed pretty much any reporting tool with an API, a Zapier connection, or a native connection, and if assembling it by hand is the blocker, this roll-up is the kind of thing our reporting module produces automatically off the sync.

One caution in the other direction: do not measure automations to death in their first month. Sequences with long cycles (win-backs, prepay) need a full season before the verdict is fair, and killing a slow-burn sequence at week six is its own failure mode. Patience for outcomes; impatience for plumbing.

The takeaway: sends, opens, and clicks are diagnostics; booked revenue, save rate, recovered AR days, and hours returned are the verdict. One monthly page with a headline number per sequence tells you what to scale, what to fix, and what to kill, and it takes less time to maintain than the illusions it replaces.

Set up the scoreboard

1. List every live sequence and assign each one its headline metric: dollars, save rate, or AR days.
2. Define the attribution window per sequence and write it down so nobody relitigates it monthly.
3. Wire sold-program and payment data from RealGreen back to your automation platform, or accept that you are measuring engagement, not outcomes.
4. Build the one-pager and put it in a standing monthly slot beside your operating review.
5. Add a holdout to your single biggest-dollar sequence this season and settle the incrementality question with your own data.