The quarterly retrospective for solo consultants

The quarterly retrospective for solo consultants — a 90-minute exercise that compounds

You run your practice for three months. Some deals close; some don't. Some engagements go well; some go sideways. You think about doing a quarterly review. You don't. The quarter ends, the next one starts, and the only thing that's actually changed is which clients you're working with.

Solo consultants almost never run a real retrospective on their own practice — even though most of them recommend retrospectives to their clients. The reason isn't laziness. It's that the prevailing retrospective frameworks are designed for teams. Sprint retros. Start/stop/continue. Action items distributed across owners. None of that maps cleanly onto an operation of one, and the friction of trying to translate it is enough to make the whole exercise feel pointless.

This article is the alternative — a 90-minute solo retrospective that produces three artifacts and one decision. It's the cheapest, highest-leverage 90 minutes you can spend in a quarter, and the difference between consultants who run it and those who don't compounds visibly across years. AI does the pattern-finding; you make the decisions.

This is a Category 1 task — the kind of work that requires your judgment, where AI helps but doesn't replace you. (For the broader frame on which tasks should go to AI in the first place, start here.) AI's role in the retrospective is to surface patterns from your pipeline data that you'd miss on your own. Your role is to decide what to change.

Why team-retro frameworks don't work for solo operators

The standard retro is built for a group. It uses voting, blameless framing, distributed action items, and a facilitator. None of that applies when you're the team of one. The risk of running a team retro by yourself is that you produce a list of ten things to "consider" — and then implement none of them, because there's no facilitator pushing for accountability and no team noticing whether the actions actually got done.

The solo retro inverts this. Instead of a list of changes, you produce one. Instead of distributed accountability, the constraint is the discipline. Instead of pattern-finding via group brainstorm, you let AI find the patterns in the data and you decide what to do about them.

The output of a good solo retrospective is one sentence: the single change I'm making in the next quarter, and the Monday-morning version of how I start. If your retro produces five changes, the retro failed — five changes implemented poorly is worse than one change implemented well.

The 90-minute structure

Three artifacts, one decision, in 90 minutes:

  1. Pull the data. ~30 minutes. The retrospective is partly the discipline of collecting data you've never collected.
  2. Pattern read. ~20 minutes. AI surfaces the patterns; you read them.
  3. ICP refinement. ~15 minutes. Based on the patterns, what changes about who you're trying to serve?
  4. The one change. ~25 minutes. The single decision for next quarter.

The whole thing happens in a single sitting. Don't break it up — the value comes from holding the data in working memory across all four steps. A walk between steps is fine; a multi-day retro is a retro that produces nothing.

Step 1 — Pull the data (30 minutes)

You'll need rough numbers on every prospect who entered your pipeline in the last 90 days:

  • Name, source (where did they come from), outcome (won / lost / still open)
  • For won deals: scope, price, engagement length
  • For lost deals: roughly why (no budget, wrong fit, ghosted, lost to competitor)
  • A note on which engagements you'd take again versus which were a slog

If your CRM is your inbox — which it probably is — this takes 30 minutes of going through email threads with a spreadsheet open. That's not wasted time; it's the hardest part of the retro. The reason most solo operators don't do retrospectives isn't because they don't believe in them — it's because the data isn't easy to pull.

A rough format:

Prospect | Source | Outcome | Scope/Price | Notes
ABC Corp | Referral from Jane | Won | $14k / 3wk | Smooth; would take again
XYZ Inc  | Newsletter | Lost | -- | Budget came in lower than my floor
...

Eight to fifteen rows is enough. If you have fewer than eight, your retro is too early — wait for more pipeline data.

Step 2 — Run the pattern-finder prompt (20 minutes)

This is where AI earns its keep in this exercise. The prompt:

You're my retrospective coach. Here's my pipeline data for the last
90 days. Tell me what you see.

Output:

1. **The shape of my quarter** — won/lost/open counts and total
   booked revenue. 2-3 sentences on what this looks like.
2. **The pattern in my wins** — what do my closed deals have in
   common? Industry, size, source, problem-type. Be specific.
3. **The pattern in my losses** — same question, for the deals that
   didn't close. What do they share?
4. **The biggest surprise** — the one finding I'd not have noticed
   on my own. Could be a positive (a source overperforming) or a
   negative (an industry that looks like fit but isn't).
5. **The question I should be asking myself** — based on this data,
   the most important question I should sit with for an hour. Not
   advice. The question.

Be honest. If the dataset is too small to draw conclusions, say so.

My data:
[PASTE]

My current ICP and offering, for context:
[PASTE]

The "question I should be asking myself" line is what turns this from a report into a useful exercise. AI is decent at pattern-spotting in business data and unhelpful at giving advice; this prompt plays to the strength.

A real output for an early-career consultant might surface that three of four lost deals were "too early" — prospects who described their situation as "we're growing fast and want help thinking about systems." Those calls converted at 0%, but they cost an average of 75 minutes each across prep, call, and follow-up. The pattern is invisible deal-by-deal and obvious in aggregate.

Step 3 — Refine the ICP (15 minutes)

Pattern read in hand, propose specific changes to your ICP. Add criteria recent wins shared but your current ICP doesn't capture; remove criteria the data doesn't support; sharpen anything vague.

Critical: argue against your own changes before locking them in. The prompt:

Based on the pipeline pattern read above, propose specific changes to
my ICP for next quarter.

Output:

1. **Add to my ICP** — criteria recent wins shared but my current
   ICP doesn't capture. Each new criterion as a single bullet with
   the evidence.
2. **Remove from my ICP** — criteria in my current ICP that the
   data doesn't support. Bullet each, with the evidence.
3. **Sharpen** — criteria that are roughly right but vague.
   "B2B companies" -> "B2B companies in regulated industries with
   50-300 employees." Each refinement with the evidence.
4. **Skeptical view** — argue against your own changes. What if
   I'm over-fitting to a small sample? Where would your changes
   have incorrectly disqualified a recent win?

My current ICP:
[PASTE]

Pipeline data:
[PASTE OR REFERENCE PROMPT 8.1 OUTPUT]

Section 4 is the discipline. Most consultants either never update their ICP or update it constantly. The skeptical pass prevents over-rotating on three months of data — which is a small sample, and AI will pattern-match on it whether the patterns are real or not.

Step 4 — Pick the one change (25 minutes)

Now the decision. The prompt:

Based on everything above, recommend the single biggest change I
should make to how I run my practice next quarter.

Rules:
- One change. Not three. Not "consider these five." One.
- Specific enough to be acted on Monday morning.
- Big enough to matter, small enough to actually do.
- Honest about what it might cost (not just what it might gain).

Format:
- **The change** — 1 sentence
- **Why it's the highest-leverage move** — 2-3 sentences
- **What it costs me** — what I'm giving up to do this
- **Monday-morning version** — the first specific action, doable
  in under 2 hours, that starts the change

If you can't pick one, that's a sign the data isn't actionable yet —
say that.

The constraint is the value. Five changes implemented poorly is worse than one change implemented well — and most retros end with a list of "things to consider." That list never gets executed because there's no Monday-morning action attached to anything specific.

A worked outcome

Suppose Priya's pipeline pattern read surfaced what we noted earlier — three of four losses were "too early" prospects converting at 0%. Her one-change output might read:

The change. Stop taking discovery calls with prospects who can't articulate a specific operational problem in their inbound message.

Why it's the highest-leverage move. Three of your four losses last quarter were prospects who described their situation as "we're growing fast and want help thinking about systems." Those calls are charity work — you spent ~75 minutes per call (15 prep, 30 call, 30 follow-up) on prospects whose actual problem hasn't crystallized yet. That's roughly 4 hours of your quarter on calls that closed at 0%.

What it costs you. Pipeline volume goes down. You'll feel this. The tradeoff is that the remaining pipeline converts at 3-5x the current rate. Your Q2 won-deal count is unlikely to be lower than Q1 even with fewer calls.

Monday-morning version. Add a single line to your contact form: "What's the specific problem you're trying to solve right now?" — and if the answer is two sentences or less, send a polite redirect to your blog post on diagnosing operational pain instead of a calendar link.

That's a useful retrospective. One sentence she wouldn't have generated on her own, anchored to her actual data, with a Monday-morning action. Worth 90 minutes once a quarter for the rest of her career.

Pitfalls

  • Doing the retrospective with too little data. Three deals isn't enough to find patterns. AI will pattern-match anyway. If your dataset is small, admit you're operating on intuition and revisit next quarter.
  • Letting "too early" be a category instead of a question. When AI tells you that "too early" was the dominant loss reason, the next question is why are "too early" prospects finding me in the first place? That's a marketing-channel question, not a pipeline question.
  • Skipping the skeptical pass. It's the section that prevents you from chasing a three-month anomaly into a strategic mistake.
  • Picking five changes instead of one. The constraint is what makes the retro work. If you genuinely can't pick one, the data isn't actionable yet.

What to do this week

Block 90 minutes on the calendar. Use a Friday afternoon if you can — it's typically lower-stakes than a Monday morning, and the next quarter is mentally tomorrow rather than mentally now. Pull the data, run the prompts, pick the change. Do it once and you'll see whether it's worth doing every quarter going forward.

For most operators, the answer is yes — the time investment compounds because each retrospective sharpens your ICP, which sharpens your filtering, which raises your win rate, which gives you more data for the next retrospective.

Going deeper

This article distills the retrospective workflow from Chapter 8 of The Solo Consultant's AI Playbook, which adds the worked example end-to-end and pairs the retrospective with the seven other workflows the chapter pulls together.

The framing for which tasks to give AI at all (Category 1 / 2 / 3) is in the first article. The retrospective is mostly Category 1 — your judgment is the deliverable — with AI as the pattern-finder, not the decision-maker.

— Digital Kreative

Back to blog