Metrics and the weekly review
Twenty-four numbers, one hour every Friday. Half of them are targets rather than measurements today, because Austen has not sold a pilot yet and has little usage history. The job of the first three months is to replace every target on this page with something you actually counted.
The Friday hour
- Open /admin/costs (or run php artisan credits:usage --models once the pilot branch is merged) and write down the cost per finished article.S7
- Update the pipeline table: accounts touched, replies, calls, pilots sold.
- Update the delivery table: articles delivered, articles late, brief rounds.
- Update the money table: cash in, recurring revenue, gross margin on the month.
- Pick the one number that looks worst and write one sentence about what you will do differently next week.
Do it at the same time every week. An hour that moves is an hour that stops happening.
Pipeline: your side
| Number | Target, months 1 to 3 | Target, months 4 to 12 | Worry when |
|---|---|---|---|
| Accounts first-touched this week | 15 | 10, because renewals take the time | Under 5 for two weeks running |
| Gap reports sent | 8 a week | 6 a week | Fewer than half the accounts you touched. It is the touch that works and it is nearly free |
| Reply rate, whole 5-touch sequence | 10% of accounts | 15%, as the copy gets better | Under 5% after 60 accounts. The problem is the segment or the trigger, not the volumeS40 |
| Discovery calls booked | 3 a week | 4 a week | Under 1 a week for a month |
| Discovery to demo | 50% | 60% | Under 30%: the qualification gate is not being used |
| Demo to pilot | 25% | 35% | Under 20% after 10 demos: the demo is wrong, most likely because the tracked-changes moment is missing |
| Pilots sold this month | 1 to 2 | 2 to 3 | Zero in any month after month 3 |
| Pilot to retainer or subscription | Not measurable yet | 1 in 4 to a retainer; 1 in 2 to somethingS39 | Under 1 in 5 converting to anything. The pilot is being delivered as a project, not sold as a start |
Every figure in this table is a planning assumption. None of it is measured on Austen and none of it comes from a vendor benchmark either, because vendor outbound statistics are published by companies selling outbound tools. Overwrite them after 100 accounts and 10 demos.S40
Delivery: the customer's side
| Number | Healthy | Worry | Why it matters |
|---|---|---|---|
| Articles delivered this month | Matches what was sold | Any shortfall at all | This is a delivery business. One missed batch on a publisher campaign ends the account |
| Articles delivered late | 0 | Any | Late is the only failure mode that is entirely yours |
| Brief approval rounds per article | 1 to 2 | 3 or more | Three rounds means the kickoff call did not establish the voice or the audience. Fix the kickoff, not the briefs |
| Articles rejected after delivery | 0 to 1 per pilot | 2 or more | Either the acceptance criteria were not agreed, or the brief approval is being rushed |
| Tracked changes accepted by the customer | Most of them | Almost none | If a customer declines every edit, the editor and the brand voice disagree. Fix the style guide. Austen shows the count in the review but does not store it, so note it when the edits are applied |
| SEO and GEO scores on delivered articles | Averages steady or rising | Either average falling two months running | Austen's own grades, shown on every article against a 90 target most articles do not reach: in the 60 days to 30 September 2026, 11 of 35 reached 90 on SEO and 3 of 35 on GEO.S6 Watch the trend, not the target. A falling average is an early warning about brief quality |
| Average delivered word count | Near 1,500 | Above 2,000 | Editing is priced per word. A 2,200-word average moves the pilot margin from 77% to about 67%S10 |
| Your hours per pilot | 8 to 12 | Above 15 | It is the binding constraint on the whole plan.S40 Measure it honestly on the first three |
Money
| Number | Where it comes from | Base case, month 6 | Base case, month 12 |
|---|---|---|---|
| Cash collected this month | Stripe | $10,988 | $20,272 |
| Recurring revenue | Subscriptions plus retainers | $5,988 | $12,772 |
| Pilots live this month | Pilot status pages | 2 | 3 |
| Subscriptions live | Stripe | 8 Starter, 4 Pro | 18 Starter, 10 Pro |
| Retainers live | Stripe | 2 | 4 |
| Model cost per finished article | /admin/costs | Target under $1.50; estimate is $0.79S9 | Same |
| Editing cost per 1,000 words | Proofed's own books | Assumed $35; must be measuredS10 | Same |
| Gross margin on the month | Revenue minus model, editing and payment fees | Target 75%+ | Target 78%+ |
The one number that can go wrong silently. The article allowance is enforced, but credits are measured and never enforced, so a Pro account running heavy optimisation, images, translations and derivatives around its 50 articles is invisible unless somebody looks.S7 That is what the first item in the Friday hour is for. Set an allowance from measured data before you ever turn CREDITS_ENFORCE on, and never turn measurement and enforcement on together.
The four numbers that rewrite this guide
- Real model cost per finished article, from /admin/costs, available now. Replaces the estimate table in Chapter 4.
- Real editing cost per 1,000 words, from Proofed's books. It is the single largest uncertainty in the whole guide and it moves the pilot margin between 77% and roughly 95%.
- Real reply rate by segment, after 100 accounts. Replaces the pipeline targets above and tells you which of the six segments to stop working.
- Real hours per pilot, after three. It decides how many retainers you can carry, which is the constraint the plan runs into around month 9.
A CRM that will actually get used
Do not buy one yet. Use a spreadsheet with one row per account and these columns: company, segment, priority, trigger, trigger URL, trigger checked date, contact one title, contact two title, touch one to five dates, reply yes or no, call date, demo date, outcome, next action, next action date. Sort by next action date every Friday. Move to a real CRM when you have more than 40 live conversations, not before, because the setup time buys you nothing at this stage.
The target CSV in Chapter 7 is already in this shape. Import it, add the tracking columns, and start.
What this chapter is not
- Not measurements. Every target here is a planning figure chosen to be conservative. Austen has almost no customer data and none is invented.
- Not vendor benchmarks. No published outbound statistic is quoted, because the ones available are published by companies selling outbound tools and none of them was measured on these segments.
- Not a dashboard specification. Twenty-four numbers in a spreadsheet, read once a week by one person, beats a dashboard nobody opens.