A pitch meeting is a controlled environment. The firm chooses who sits at the table, which slides appear, which prior project gets named. Everything you see has been selected. That is not dishonesty; it is the job. But it means the meeting itself tells you very little about whether the work will land, because a firm that can do the work and a firm that can describe the work produce an almost identical hour.
What separates them is paperwork. Not paperwork as bureaucracy, but as a set of specific documents that force a provider to commit, in writing, to things that are easy to say out loud and hard to fake on paper. In a large organization, where the person who runs the selection is rarely the person who lives with the result, those documents are also the only thing that survives the handoff.
1. Find out who actually does the work, and lock them in
Professional services firms run on what the industry calls leverage: the ratio of junior staff to senior staff on an engagement. The economics depend on it. A partner's rate is high, but a partner cannot bill forty hours a week on your account and still sell work, so the model puts senior people on the pitch and on the steering calls, and puts the day-to-day in the hands of people five or ten years younger. This is normal and often fine. The juniors are frequently very good. The problem is that nothing in the pitch tells you which way the ratio falls on your job.
The document that fixes this is a key personnel clause. It names the individuals assigned to your engagement, states the minimum share of their time committed to it, and requires written notice and your approval before anyone is substituted. Ask for resumes of the named people, not firm credentials. Then ask the question that actually produces information: what else are these people assigned to during the same months, and what happens to your timeline if one of those projects slips.
A firm with real capacity answers that in specifics. A firm that is hoping to staff you after they win says something about depth of bench.
2. Make the deliverable testable, not describable
Most disputes over consulting work are not arguments about quality. They are arguments about whether the thing was finished. That happens because the statement of work, the document that defines what is being bought, was written in verbs that cannot be checked: assess, support, advise, enable, align.
A deliverable is testable when a person who was not in the room can look at it and say yes or no. "A written assessment of the claims intake process" is describable. "A process map covering every step from first notice to file close, a list of handoffs with the system of record for each, and a ranked list of at least fifteen findings with estimated effort for each" is testable. The second version also tells you something before you sign, because a firm that cannot produce that level of specificity in a proposal usually cannot produce it in delivery either.
Pair it with an acceptance procedure: how many business days you have to review, what a rejection must say, how many revision cycles are included, and what happens if the deliverable is rejected twice. Without that, acceptance happens by silence, and silence has a way of arriving right before the final invoice.
3. Check the things that have a paper trail, before the things that do not
References are worth having, but they are a curated list and you should treat them that way. The checks that carry more weight are the ones a firm cannot stage.
- Certificate of insurance. Professional liability, also called errors and omissions, covering the engagement value, with your organization listed as a certificate holder so you are notified if the policy lapses. Ask for the certificate itself, not a statement that coverage exists.
- Conflict check in writing. A signed statement identifying current and recent work for your competitors, your vendors, and any party you are in a dispute with. Large firms serve crowded markets; a disclosed conflict with a screening arrangement is usually workable. An undisclosed one found later is not.
- Subcontractor disclosure. Who else touches the work, under what terms, and whether your confidentiality and security obligations flow down to them.
- Security and data handling addendum. If the engagement touches your systems or your customers' records, this is where the controls get named. The National Institute of Standards and Technology is responsible for much of the control vocabulary these addenda borrow, so a provider that works at scale will recognize the terms without needing them explained.
When you do call the references, skip the question about satisfaction. Ask what changed mid-engagement, and how the firm handled it. Every project changes. The answer tells you how they behave when the plan stops being accurate.
4. Buy the hours in a way you can watch
Fixed fee and time-and-materials are not a preference, they are a bet on how well the scope is understood. Fixed fee moves the risk of overrun to the provider, which is why providers price it with a cushion and defend scope hard. Time-and-materials moves the risk to you, which is cheaper when the work is predictable and expensive when it is not.
Either way, what you need is visibility while the money is being spent rather than after. Three mechanisms do most of that work:
| Mechanism | What it prevents |
|---|---|
| Not-to-exceed cap on a time-and-materials engagement | An overrun you learn about on the invoice |
| Invoices with daily time entries by named person | Senior rates billed for junior hours |
| Monthly estimate-to-complete against budget | A schedule that quietly slips two months before anyone says so |
Add a written change order process: scope changes require a signed amendment with its own price and schedule impact before work begins. Firms that work with large clients expect this. It protects them too, because it gives them a clean way to say that something was not in the original deal.
5. Decide now where the knowledge lands when they leave
The expensive failure in outside expertise is not bad work. It is good work that leaves with the people who did it. Six months after the final presentation, the model is a spreadsheet nobody can update, the analysis rests on assumptions nobody documented, and the only people who understand it bill by the hour.
Fix that in the contract. Specify ownership of work product, including the underlying models, code, and working files, not just the final report. Require source files in editable formats. Build a knowledge transfer deliverable into the schedule, with named internal staff who attend it, and hold a meaningful share of the fee against its completion. Set out what the provider must return or destroy at the end, and when.
Then do the part that is yours: assign an internal owner who sits in the working sessions rather than the steering committee. The organizations that get durable value from consultants are the ones where someone on staff can explain the reasoning afterward.
None of these checks require you to out-argue a specialist on their own subject. They require you to ask for commitments in writing and notice which firms find that easy. The ones who do it without friction are usually the ones who have delivered this before.