One recommendation, even when the analysis is inconclusive

Four outputs

Stage 4 — decide and recommend — is where the professional earns the fee. Stage 3 built a lens; stage 4 looks through it and says what to do. The stage produces four things: one recommendation; the reasons, tied to the criteria and the matrix; the risks, each with a mitigation; and next steps, each with an owner and a date. The rubric for the stage, R-psp-4 — a Memra rubric that mirrors the stage criteria the open sources agree on — has six lines: a single clear recommendation · linked back to the criteria and the matrix · revisits the client's goals and values · acknowledges the strongest counter-consideration · names the top risk and a mitigation · next steps with owners and dates. This module takes them in that order.

The professional's job is a judgement

The most common stage-4 failure is not a wrong recommendation; it is no recommendation. Three shapes of it: "it depends" — true of every decision and useless to the client; three options and no choice — the analysis handed back to the person who paid to have it made; and "more data is needed" as the whole answer — sometimes true, never sufficient. The client asked a question because she has to act by a date with the information she has. So the rule: recommend anyway. State what you recommend, on what basis, and what would change your mind. That third clause is what makes a judgement under uncertainty honest rather than reckless — it tells the reader exactly which fact, if it turned out otherwise, would flip the answer.

Worked example — Lakeshore's recommendation

> Lease a new press (option C). On the figures provided, it is the lowest annual cost of the three fixes — $58,000 a year, against about $67,400 to buy and about $74,400 of margin lost by outsourcing — it restores on-time delivery with a new press under a maintenance contract, it keeps both operators, and it adds no debt. It led the matrix at 4.10 against 3.05 for buying, and it led under every weighting tested. I would change this recommendation if large-format volumes were expected to fall by more than about 20 %: below that, a lease payment that does not shrink becomes the most expensive option and outsourcing's variable cost becomes attractive.

Four sentences. The first is the recommendation and nothing else. The second is the basis, in the client's numbers. The third is the link to the matrix — the total and the sensitivity, because "4.10" alone is a number, and "led under every weighting tested" is a reason. The fourth is the flip condition, and it names a measurable thing.

Now the failures, on the same case. "Each option has strengths and weaknesses, and the right choice depends on Priya's priorities" — true, and worth nothing: her priorities were fixed in stage 1. "Lakeshore could lease, buy, or outsource; we present all three for Priya's consideration" — the matrix was built and then ignored, which loses the second rubric line as well as the first. "We recommend obtaining more quotes before deciding" — a next step wearing a recommendation's clothes.

Write the clinic's

Grand River Family Dental's matrix (module 6): paper 2.20, reminder-only 3.10, full online booking 3.90, hybrid 3.75 — and the choice between the last two turned on training burden. A stage-4 sentence that respects that: Move to an online-booking system with reminders (option C). It scored highest on the two objectives the dentists named — fewer no-shows and an end to double-bookings — at about $1,800 a year against an estimated 12 % of appointments lost. I would change this to the hybrid if the front desk cannot absorb the training within a fortnight, because that is the only weighting under which the hybrid overtakes. Recommendation, basis, link, flip condition.

Write one cold

> A Guelph café's matrix puts a second espresso machine at 3.9 and extended opening hours at 3.7; the totals swap if the weekend-demand score is one point lower.

Recommend the machine; the basis is the matrix and the queue data; the flip condition is weekend demand — and the sentence says what evidence (say, four weekends of till data) would settle it. What you must not write is "both are viable".