Reporting
Every month, the reports. In between, ask anything.
A question / how it travels
- You ask
- In plain words. In your dashboard, by email, or on the call.
- It runs
- As a query against your joined data, not against a summary of it.
- You get
- The figure, its sample, its period, its confidence, and how it reconciles.
- Turnaround
- Quick for what the data already answers. Hours or days for a new cut. Never never.
Ask in your own dashboard, or just ask us. Either way the answer is computed from your own numbers, joined across every system you run, with the method attached. Not an opinion, and not a slide.
01Why your reporting cannot do this
The dashboards you have answer the questions someone anticipated. Yours are not on them.
A chart is fixed when it is built. The question you actually have spans four systems that have never been joined, so it is not on any screen you own. It goes unanswered, or it gets answered with a vendor’s opinion and a confident slide. The fix is not another chart. It is joined data underneath, so a new question can be asked at all.
The question crosses the seam
“Which keywords produced customers last quarter” needs the ad account, the website, the CRM and the spend ledger at once. Each vendor holds one of them. None of them can answer, and each one’s report looks fine.
So it is answered by opinion
The agency says the cheap channel is working. Your intake lead says the leads are bad. Both are reading a quarter of the picture, and the meeting ends where it started.
The join is the work
Once the Outcome Ledger exists, every question of this shape becomes a query. The expensive part was done once; after that the answer is a matter of hours, not a new project.
The method page teaches the join in full. This page is about what you can ask once it exists.
02What gets connected
Every system you run, joined. Including the ones kept by hand.
Each source contributes the one thing it knows. None of them can answer the question on its own; joined, they answer most of them.
- 01always
Ad platforms
What was spent, on which campaign, keyword and audience, by day.
- 02always
Analytics and tag manager
What fired on the site, once or more than once, and what it was told to count.
- 03always
The website and its forms
Who enquired, when, and the click that brought them.
- 04always
The CRM, including its stage history
Every stage each enquiry passed through and when. Not just where they are now.
- 05always
The spend warehouse
Certified cost, landed nightly, so no month is priced on a provisional number.
- 06where it exists
Call and text
Which calls came from which ad, and what followed.
- 07where it exists
Email sends
What your staff and your sequences sent, to whom, and whether it moved anything.
- 08always
The spreadsheets your team keeps by hand
The counts the business has actually run on, often for years.
Your spreadsheets, kept
The sheet your team has maintained by hand for years is not thrown away. It becomes a source.
The hand-kept count is usually one of the things we are hired to fix: typed by hand, unaudited, no change history, and its meaning drifts. The reconciliation is where that shows, in the open, with the people who keep the sheet in the room.
- 01
It is loaded, not replaced
Every tab and every month goes in next to the system data, with its own labels. Nobody has to learn a new name for a number they have counted every Friday for five years.
- 02
It is reconciled, cell by cell
Our count is set against theirs for every period the sheet covers. Where the two agree, that agreement is itself a check on both. Where they differ, the difference is explained, not averaged away.
- 03
It is queryable
“What did the sheet say in March, and what does the system say now?” is a question with an answer. The history your team built by hand becomes something you can ask about.
03The kinds of questions
The questions a dashboard was never built for.
Generic by design. Yours will be more specific, and the shape is the same whether you sell to consumers or to other businesses.
Spend
- Which keywords produced customers last quarter, rather than leads?
- Cost per held consult, estimate or demo by channel, and how it has moved over twelve months.
- Does the cheap channel actually work, measured at the sale rather than at the form?
- What are we paying for the segments that never buy?
- Which ad set the platform calls best is best once the outcome is counted?
Funnel
- Where between enquiry and payment does the loss actually happen, this year versus last?
- No-show rate by how far ahead the appointment was booked.
- How long does a proposal sit before it is signed or lost, by deal size?
- Which lost reasons are growing, and in which segment?
- Did the booking rate fall because of demand, capacity or follow-up?
Reconciliation
- Do our numbers match the sheet the team keeps by hand, and where exactly do they diverge?
- What actually happened in March? Not the headline: the stage-by-stage account.
- Do the CRM, the finance records and the sheet agree on how many customers we signed?
- What did the agency report, and what does the joined data say about the same period?
Operations
- Speed to first contact by day of week and by coordinator.
- How many enquiries are overdue for a follow-up right now, and how old is the oldest?
- Which sequence step gets the reply, and which gets the unsubscribe?
- Which pages or forms on the site started under-performing, and when?
None of these are answered in real time. They are answered in hours or days, from your data, instead of never.
04How an answer arrives
Computed, not generated. Every answer carries the same eight things.
This is the part that matters. An answer you cannot check is an opinion with a number in it. These are the rules the answer is held to, and they are the same rules the monthly report is held to.
Anatomy / what comes back
- The figure
- Computed by a query, not written from memory.
- Sample
- How many enquiries, deals or days it rests on.
- Period
- Which dates, in your local calendar.
- Definition
- What counted as the outcome, by the agreed written definition.
- Confidence
- Proven, suggestive, untestable or refuted. Stated, not implied.
- Reconciliation
- How it compares to the number you already keep, and why they differ.
- Limits
- What the data cannot say, named rather than skipped.
- The query
- Kept, so the same question re-run next month gives a comparable answer.
R1
Computed, not generated
Every number traces to a query that ran against your joined data and can be run again. If a number cannot be re-run, it is not reported.
R2
Definitions live in one place
What counts as a held meeting, an estimate given or a retained client is written down once and used everywhere: the monthly report, the weekly grid and the answer to your Tuesday question all share it. Change the definition and every number moves together, in the open.
R3
Never a bare zero for something unmeasured
A sparse table does not mean nothing happened. The answer says “none since the feed broke” or “not tracked before June”, never “only a handful ever”. Unknown stays unknown rather than being forced into a bucket.
R4
Numerator and denominator on the same scope
Never total spend divided by channel-attributed outcomes. The first wrong number in most reports is a mixed denominator, and it is the one that makes the cheap channel look good.
R5
Frozen, then revised in the open
Each day’s figures are frozen once, then re-checked as late outcomes land. The first value survives: a revised number shows what it was and when it changed, so last month’s report still means what it said.
R6
The agreed definition governs
Where our count and the one you already keep disagree, you get the reconciliation and the reason, not whichever number flatters someone. Neither side changes a definition unilaterally; a change means re-striking the baseline.
05The recurring reports
Four reports on the retainer. Each one has an audience and a question.
The questions above are answered on demand. These arrive on a schedule, from the same joined data and the same definitions, so the number in the report and the number in the answer never disagree.
- 01
Executive
- For
- The owner
- Cadence
- Monthly
Is the marketing working? Cost per enquiry, booked, held and won by business unit, against last month and last year, with revenue on matured cohorts, what changed, and the data-quality notes that qualify it.
- 02
Marketing
- For
- Your marketing lead and your ad agency
- Cadence
- Every two weeks
Where should the next dollar go? Channel, campaign and keyword traced to held and won; network mix over time; what the platforms are being told and whether the outcome feed is healthy; the brief for the agency call.
- 03
Operations
- For
- Your intake or sales manager
- Cadence
- Weekly
Is the follow-up holding? The KPI grid that replaces the hand-kept sheet, speed to first contact, no-show by booking lead time, backlog age, overdue follow-ups, workload by coordinator, lost-reason coverage.
- 04
Development & systems
- For
- The owner
- Cadence
- Monthly
What was built, and is it holding? What shipped this month and what it changed, the systems running behind it, requests by channel and how fast they turned round, whether the scheduled jobs and emails actually went out, and who is using what. The month of engineering, in plain language, with nothing to take on trust.
06Your dashboard
A place to read it, and to ask the next question.
The reports arrive on their schedule whether you open anything or not. The dashboard is there for the moment you want more than the report in front of you.
Read the reports
Every recurring report, rendered from the same stored numbers that produced it, with each figure carrying its sample, its period and how it reconciles. Nothing is a static image of a chart somebody exported.
Build a new one by asking
Describe the cut you want in plain words. It is generated against your joined data, you see the query behind it, and if it is worth having every month it becomes one of your recurring reports.
Ask a question outright
For the question that is not a report. Read-only, answered from the same joined data under the same definitions, with the method attached and the limits named.
Access is yours and it is read-only: asking a question never changes a number, and nothing you ask writes back to your CRM or your ad accounts.
07Where AI fits
An agent runs the queries. An engineer directs it. The data was already joined.
That is the whole of it. The speed comes from the join and the tooling existing before you ask, not from a model guessing at your numbers.
Division of labour / engineer, agent, and what neither of them is
What the agent does
Writes the query, runs it against the joined data, pulls the sample and the period, checks the result against the definitions in config, and drafts the answer with its evidence attached. Thousands of small read-only scripts, each written for one question.
What the engineer does
Reads your question, decides what it actually asks, reviews the query and the result, and sends the answer. Nothing reaches you that a person has not read against your data.
What it is not
A model answering from a summary of your business, or a confident paragraph with no query behind it. Every figure traces to a query that can be re-run against your data, and anything that leaves as a report has been read by a person first.
If you have been pitched AI reporting before and it came back vague, this is the difference: the model is writing queries against data that is already joined and already has agreed definitions. It is not guessing at your business from a description of it.
08On the retainer
This is the reporting half of the fractional CMO retainer.
The four reports and the questions between them come with the seat. The reporting engine that makes it possible is built once, as part of the implementation, and anchored on the services page against what the same engine costs built in-house or through an agency.
On the reference engagement the staff spreadsheet was reproduced cell by cell before it was automated. The specifics are in the case study. Read it there.
However you ask — in the dashboard, by email, or on the call — the answer is computed from your joined data under agreed definitions, and every figure can be re-run.
Start
Start with the teardown. The questions come after.
The teardown reads your ad accounts, analytics and website and tells you whether what you are being told is true. The reporting on this page needs the CRM joined, which is the diagnostic and the build. The call is thirty minutes to decide whether to begin.