Disclosure: This article may contain affiliate links. We earn a commission at no extra cost to you if you purchase through our links. We only recommend tools we genuinely believe in.
Last December, a client I had worked with all year asked me a simple question on a call: "So, what did we actually get this year?" I had the answer, spread across twelve monthly reports, a hundred Slack threads, and my own memory. I did not have it in one place, in one page, in words he could forward to his partner. I spent the weekend rebuilding the year from scattered files, and the report I finally sent was late, defensive, and padded with charts to hide the fact that I had not planned it. He renewed anyway, but the conversation that should have been a victory lap felt like an exam I had not studied for.
That is what a year-end report is for. It is not a summary. It is the one document that turns twelve months of work into a renewal decision the client can make with confidence. Done right, it takes an afternoon, because the year is already documented in your monthly reviews. Done wrong, it takes a weekend, because you are reconstructing a year from memory. This post is the system I use now, and it plugs straight into the monthly retainer review I already run, and the case study habit that gives the report its proof.
This is a playbook, not a results report. The figures below are planning templates to adapt to your own numbers, not measured outcomes from my clients. Your numbers will differ because your clients, your channels, and your fees differ.
The Problem: The Year-End Report Most People Send Is a Data Dump
The default year-end report is a 14-page PDF of every chart the platform can export, in the order the platform exports them, with a "thank you for a great year" on the last slide. The client opens it, sees a wall of graphs, and closes it. Three things go wrong:
- It starts with data, not with the plan. The client agreed to a goal in January, usually in plain words: more qualified leads, lower acquisition cost, a launch that lands. A report that opens with impressions and click-through rates never answers whether the goal was met, because it never restates the goal.
- It hides the misses. Every year has a quarter that went sideways, a channel that underperformed, a test that failed. A report that skips those reads as spin, and the client knows it. The moment they spot the omission, they stop trusting the wins too.
- It has no ask. A report that ends with "thanks for a great year" wastes the one moment the client is actually looking at the whole year. The report should end with a clear recommendation for next year, with a price, because that is the renewal conversation, and it is easier to have it in December than in January when the budget is already set.
The fix is a one-page report with four parts, built from documents you already have, drafted with AI, and checked by you before it goes out. One page, because the client will actually read one page. Four parts, because those are the four questions the client is asking.
The System in One Pass
The stack reuses the tools from the retainer review and case study posts. No new platform is required, and no part of this system sends anything to the client on its own.
| Stage | Tool | What happens |
|---|---|---|
| Gather | The twelve monthly retainer reviews, plus results emails | One folder per client, one source doc per month, pulled into a single year file |
| Score | The January plan, next to the December actuals | Each goal marked met, missed, or moved, with the reason in one line |
| Draft | Claude or ChatGPT, with the year file and the case studies attached | A one-page draft in the four-part structure below |
| Check | You, with the source docs open | Every number traced to a source, every miss kept in, nothing invented |
| Send | Email, with the one-pager attached, plus a short appendix | Sent in early December, with a call offered the same week |
Notice what automation does not do here: it does not decide what the year meant, it does not choose which misses to include, and it does not set the price for next year. It drafts the page from the notes you already kept. You decide the story, and you send it.

Step 1: Gather the Year in One File, in One Hour
Start with the twelve monthly retainer reviews. If you have been running them, the year is already written. If you have not, this step is the reason to start in January, and this December you will do a shorter version from your invoices, your ad accounts, and your sent emails. Either way, the output is one file with five sections, and it takes about an hour per client.
- The January plan. The goals agreed at the start of the year, in the client's own words where possible. One or two lines per goal. If the plan changed mid-year, note when and why, because a goal that moved is not a goal that was missed.
- The monthly results. One line per month: the headline number, the main thing shipped, and anything that broke. Twelve lines, not twelve reports.
- The wins with proof. The three or four moments the year actually turned on: a launch, a campaign that worked, a fix that stopped a leak. Each one gets a link to the case study if you wrote one, because a claim without a link is just a claim.
- The misses. Every goal that was missed, and every quarter that went sideways, in plain language. This is the section most people skip, and it is the section that makes the report believable.
- The money. What the client paid across the year, in one line, and what the work returned, in the same units the plan used. If the return cannot be measured cleanly, say so, and say what you measured instead.
Feed that file to the model, not your raw exports. A model handed twelve platform exports will write a data dump. A model handed one tidy year file will write a report.
Step 2: Score the Year Against the Plan
The heart of the report is a small table, and it is the part the client reads first. Four columns, one row per goal, no adjectives in the score column.
| Goal from January | Result by December | Score | Why, in one line |
|---|---|---|---|
| Grow qualified leads from ads | Leads up, cost per lead down | Met | Creative testing workflow from March compounded all year |
| Launch the new offer in Q2 | Launched in Q3 | Moved | Client delayed the product, launch slipped a quarter |
| Cut acquisition cost by a fifth | Cost flat, volume up | Missed | Budget went to volume instead of efficiency, agreed in June |
| Build the email follow-up system | Live since September | Met | Follow-up sequence now runs on every new lead |
Three rules for this table. First, the score column holds one word only: met, missed, or moved. A goal that moved is not a failure, and it is not a success either; it is a decision, and the report says when it was made. Second, every missed goal keeps its reason in the open. The client already knows the quarter went sideways. Naming it first is what makes the rest of the page credible. Third, the numbers are planning figures from your own records, traced to a source doc, never rounded up in the retelling. If a number cannot be traced, it does not go in.
Step 3: Draft It With AI, Then Check Every Number
Give the model the year file, the scorecard table, and the case studies, and ask for a one-page draft in four parts: what we set out to do, what happened, what it cost and returned, and what I recommend next. The prompt I use is short, and it has two hard rules in it: every number must come from the year file, and anything the file does not support must be flagged, not filled in.
The draft comes back, and then you do the work the model cannot. Open the source docs and check every number, one by one. This takes twenty minutes and it is the whole job. The most common error is a number that is true for one month and presented as true for the year, usually because the model found the best month and treated it as the average. The second most common is a win attributed to the wrong quarter. Fix them, and rewrite the draft in your own words. A report in your voice, with your numbers checked, reads as yours. A report pasted from the model reads as a template, and the client can tell.
Step 4: End With the Ask, Not With Thanks
The last part of the page is the recommendation for next year, and it is the reason the report exists. One paragraph, three lines: what you recommend, what it costs, and when it starts. This is the renewal conversation, and it belongs here, while the year is fresh, not in January when the client has already set the budget without you.
The recommendation comes from the scorecard, not from a menu. A missed goal that is still worth chasing becomes next year's first project. A goal that was met becomes the case for expanding the same work, and that is where the upsell ladder from last month's post fits: one rung, one price, one start date, offered after a win. A goal that moved becomes a question, not a pitch, because the client may not want it anymore. If you have no recommendation, say so plainly and say why: the honest line is "the current scope is the right scope for next year," and it costs you nothing to write.
Step 5: Send It in Early December, Then Offer the Call
Send the report in the first week of December, as an email with the one-pager attached and the monthly detail as a short appendix. The email itself is five lines: the year in one sentence, the scorecard in one sentence, the recommendation in one sentence, the attachment, and an offer of a call this week. The client reads the email, opens the page, and books the call, or replies with a yes to the recommendation, which is the same thing.
Never send the report in the same message as an invoice, and never send it between Christmas and New Year, when nobody reads anything and the renewal quietly slides into February. If the client does not reply in a week, send one follow-up, then stop. Silence after a year-end report is usually a busy December, not a no, and the call offer stays open into January.
What Breaks, and the Guardrail for Each
| Failure | Warning sign | Guardrail |
|---|---|---|
| The report is a data dump | Fourteen pages, charts in platform order, no goal restated | One page, four parts, the scorecard table on it |
| The misses are hidden | Every goal marked met, no quarter mentioned | Every miss in the open, with the reason in one line |
| A number cannot be traced | The model wrote a figure you cannot find in the source docs | If it cannot be traced, it does not go in |
| The ask never happens | The report ends with "thanks for a great year" | The last part is always the recommendation, with a price |
| The report arrives too late | It goes out after Christmas, or with the January invoice | Send in the first week of December, with a call offered the same week |
Lessons
- The year-end report is not a summary. It is the renewal conversation, written down, and it is easier to have in December than in January.
- A one-page report gets read. A fourteen-page export gets closed. The scorecard table is the page.
- The misses are the credibility. A report that names them is believed. A report that hides them is not.
- The year file is the real system. If your monthly reviews exist, the report takes an afternoon. If they do not, December is when you feel it, and January is when you start.
- AI drafts the page from the notes you kept. You check every number, you decide the story, and you make the ask. That split is the whole workflow.
