Every firm learns why it wins and loses too late, from too few people, who are too polite to tell the whole story. There’s now a fourth source, says Steven Lewis, and it’s worth understanding before someone sells you a bad version.
A major firm lost a panel every firm wanted. The pitch team told me the client shovelled on the truth at the debrief. There was no clear value proposition, the value-adds made no sense, and nobody called the client before writing the pitch. The firm is locked out of another shot at this whale for years.
Everything you know about why your firm wins and loses comes from three places: the debrief, the client listening programme and the partners’ read. All three miss in the same way.
Three sources, one failure
The debrief is too late. It exists only once the decision is made.
Client listening is too few. I’ve seen client surveys return 15 to 20% on a small send, answers knocked out grudgingly between bites of a chicken wrap. I watched one pursuit team go into a dozen panel renewals armed with slim listening from three years earlier. These were multimillion-pound panels.
And much of the feedback that reaches you has been smoothed to a bar of soap. Clients soften what they say, especially if they see themselves drinking your conference-room flat whites in the future. Partners don’t ask the hard questions and slide the fader down on even the soft criticism they hear.
Why this matters more now
AI is making pitching cheaper. Cheaper is not better. Most of the pitch tools demonstrated to firms are a photocopier with AI-powered find-and-replace. They read your back catalogue, learn the house style and swap in the new client’s name. The photocopier approach works if two things are true: you have Dolly Parton’s back catalogue and your only problem was the cost of writing all those hits.
If your old pitches said what partners like telling their mums about the firm, not what clients needed to hear, the machine will produce the same crap faster. Producing crap faster is a medical condition. Firms building their own tools face the same trap from the inside. The training data is the firm’s own work, and the person signing off the build is often the person who wrote it. They’re building a mirror, not an engine.
The constraint on pitching was never production. It was knowing what the client needs to hear.
Asking the same client twice
Synthetic client research puts your thinking in front of modelled clients before it reaches real ones. You build a panel of AI personas (we call them Client Proxies), each configured to reason like a specific kind of decision-maker: a general counsel in financial services running 40 in-house lawyers, a CFO at a high-growth telco and the audit committee chair. We model everyone in the decision room, not just the ones partners can lunch.
Then you can do what no real client will sit still for: put the same buyer in front of all the questions you need to ask. That includes questions you couldn’t ask a real client – ‘How bad would your experience have to get before you considered changing firms?’
When we built 334 Client Proxies modelled on corporate legal buyers, we gave them two scenarios. In the first, they were choosing a firm for a high-stakes matter, one with the board’s attention. In the second, the matter was routine. You can’t do this with human respondents, because they anchor on the first question you asked. It’s the same reason you can’t show one client five value propositions and trust the fifth answer. (Assuming you’d want a real client to know you can’t decide how you’re valuable.)
On the high-stakes matter, not one chose the challenger. On routine work, more than nine in 10 were open to one. Same buyer, same challenger; only the stakes moved. And when high-stakes work did move, the incumbent had stumbled. A better pitch almost never did it.
A 2011 study of 166 chief legal officers at S&P 500 companies found real clients pointing the same way: service quality lay behind 60% of terminations, cost alone behind 7%. One synthetic research finding matching one real-client study doesn’t validate an approach, but it is a reason to take the methodology seriously. Especially when you can’t survey 166 chief legal officers yourself.
What synthetic research gets you as a marketer
Picture your next pitch strategy meeting. The incumbent has held the client for a decade, and a senior partner says to sharpen the fees. Today, you have a hunch that fees aren’t the lever, but the partner has the weight in the room. You lose. With research behind you, you have more weight. In the research above, price barely registered when buyers explained what rules out a challenger. A challenger leading with a discount is answering a question the client isn’t asking, and you can say so.
With a panel to poll, you can test your value proposition before you pitch it, instead of learning too late from the rejection. You hear from the buyer types who never return your survey. And you get the objection no client would put to a partner’s face, because a modelled general counsel has no relationship to protect.
What synthetic research isn’t
Synthetic research doesn’t predict behaviour. It shows how clients reason, not how many will act on it.
The numbers are directional. A modelled panel tells you how hard a market leans. It is not a head count of real clients, and anyone quoting you decimal places is selling precision the method doesn’t have.
Synthetic research doesn’t replace talking to clients. It makes the chances you get to listen more valuable, because you arrive with better questions.
What synthetic research does well is giving you the reasoning: the objections clients have and the language they use. It tells you the proof you need to show if clients are to believe you are the pragmatic, commercial, deep bench you say you are.
Three questions before you buy
Synthetic research attracts big promises, especially from build-a-bot websites that spin up digital personas from a few details. Ask any supplier three things. What can’t your method tell me? How is the panel built to match my clients? What do you do when your method is wrong? A supplier who can’t answer the first question has told you everything.
Steven Lewis is the founder of Taleist Agency in Sydney, which runs synthetic research for professional services firms. He has worked in business development inside a top-tier Australian law firm. Read the study at taleist.agency/goliath.
Members only: if you want to add this reading to your CPD, please log in and complete the evaluation
Log in
