Every proposal methodology sold in the last twenty years rests on a claim of the same shape. Winning proposals share a characteristic. Ours teaches you the characteristic. The characteristic changes depending on who is selling: an executive summary that leads with outcomes, a compelling risk narrative, a named business case, the right pricing structure.
We set out to find which one the evidence supports. The honest answer is that it supports none of them, and understanding why is considerably more useful than another confident claim would have been.
The headline we could not write
The article we intended to write was “the proposals that closed had one thing in common.” To write it responsibly we needed a study that took proposal documents from a substantial set of professional services wins and losses, coded what was actually in them, controlled for the conditions that existed before the document was written, and identified a document-level characteristic that independently predicted the outcome.
This matters because the claim is made constantly, usually with a number attached, and the number almost never survives contact with its own methodology. Understanding what the good research actually measured is the difference between a defensible position and a confident guess.
What the win/loss data actually measures
Three studies come closest, and each one is instructive about a different limitation.
Shaik, Sridhar, Sriskandarajah and Mittal studied a global on-site services provider using 4,574 sales opportunities across 23 countries between 2010 and 2021, of which 2,821 resulted in submitted bids. The outcome is observed win or loss in archival CRM data, not a survey asking salespeople why they thought they won, which already places it above most of what circulates. Their model covers opportunity size, relationship type, strategic account status and service characteristics, and finds that stronger customer relationships significantly increase the probability of winning.
That sounds tantalisingly close to the one thing. It is not. Relationship is one predictor among several, and the dataset contains structured opportunity information rather than proposal copy. The authors note that their data omit factors including detailed pricing and salesforce incentives, and propose future work incorporating unstructured material such as emails and internal notes. You cannot convert that finding into “winning proposals build relationships,” because the relationship largely existed before anyone opened a document.
Rezazadeh supplies the largest professional services dataset we found: 25,578 closed opportunities from a global B2B consulting firm across healthcare, energy and financial services, spanning January 2015 to August 2019. It starts with 20 CRM variables and expands them into 137 model features, then validates against a further 846 closed opportunities gathered over three months.
The important result for our purposes is structural rather than substantive. The dataset does not collapse to one obvious winning characteristic. Predicting the outcome requires seller, account, geography, project and value features working together. The method succeeds precisely because it is multivariate. And once again, nothing inspects the submitted document.
Singh, Marinova and Singh get closest to content. Their study analyses actual email communications during B2B contract negotiations, augmented with interviews, salesperson survey data and archival performance records, then checks the mechanism in an experiment with professional salespeople. The underlying dissertation reports 43 contract negotiations observed over two and a half years, with contract award as the outcome rather than salesperson opinion.
What they find is not one phrase or tactic. Complementary influence tactics can increase buyer attention and the likelihood of a contract award, while poorly combined tactics degrade attention and actively disadvantage the seller. The result is an interaction between tactics, which is close to the opposite of a universal formula.
The nearest thing to convergence
There is something resembling agreement across the literature, but it sits one level above proposal technique.
Palmatier, Dant, Grewal and Evans synthesised 94 studies covering roughly 38,000 relationships. They concluded that seller objective performance is influenced most strongly by relationship quality among the relationship constructs tested, and that relationship marketing works better where relationships matter more, notably in service offerings and business markets, and when the relationship is person to person rather than firm to firm. For a boutique professional services firm, that is an unusually well-aimed finding.
Put that alongside the Shaik result, where stronger relationships significantly raised win probability on actual competitive bids, and relationship quality becomes the strongest candidate for a genuinely convergent factor.
The most convergent finding in the literature is about the state of the relationship before the document was written, not about anything inside it.
Which is an awkward conclusion for anyone selling proposal training, and a clarifying one for anyone deciding where to spend effort.
Adaptation beats formula
A second research stream argues the operative principle is not relationship alone but adaptation, and it comes with a methodological warning worth internalising.
Franke and Park combined 155 samples representing more than 31,000 salespeople. Adaptive selling increased self-rated, manager-rated and objective performance. Customer orientation increased only self-rated performance. That gap between what people believe about their own effectiveness and what shows up in outcomes is exactly where most proposal folklore lives.
McFarland, Challagalla and Shervani examined 193 matched buyer and seller pairs and found support for matching different influence tactics to different buyer characteristics, rather than applying one method universally. What persuades depends on whom you are persuading.
Verbeke, Dietz and Verwaal synthesised 268 studies covering 292 samples, 79,747 salespeople and 4,317 organisations. Their multivariate model identifies several independent correlates of performance, including selling-related knowledge, adaptiveness, role ambiguity, cognitive aptitude and work engagement. Crucially, those relationships are moderated by measurement method, research context and sales type. Even the largest synthesis available refuses to produce a single lever.
Three things that follow from this
- Be suspicious of self-reported win reasons. The customer orientation result shows how cleanly a belief about effectiveness can fail to appear in objective performance. Most published win/loss insight is self-reported.
- Universal formulas contradict the matching evidence. If the effective tactic depends on the buyer, a template that applies the same persuasive structure to every reader is optimising the wrong thing.
- Context moderates almost everything. Sales type and measurement method change the size and sometimes the direction of effects, which is why a statistic borrowed from a different selling motion tells you very little.
The study nobody has run
The gap is specific enough to describe precisely, which also makes it a real opportunity for any firm holding a decent proposal archive.
What is missing is a study that takes the submitted documents from a substantial set of professional services wins and losses, codes their content systematically, controls for the pre-existing conditions that the CRM data shows to matter (relationship strength, account status, deal size, competitive context), and then asks whether any document-level characteristic independently predicts the outcome.
Every element of that already exists separately. Shaik has the bid context. Rezazadeh has the opportunity variables. Singh has the language analysis. Nobody has combined them, and without the controls the whole exercise is worthless, because a firm with stronger relationships also writes more confident proposals and wins more often. Untangling those three requires the controls, not more anecdotes.
What to do instead
The absence of a universal answer is not the absence of guidance. It relocates the guidance.
If relationship quality is the most convergent predictor and it largely predates the document, then the proposal's job is narrower than the methodologies claim. It is not there to manufacture conviction from nothing. It is there to avoid squandering the position you already hold, and to make the decision easy for someone who is already inclined to say yes but has to justify it internally.
If adaptation outperforms formula, then the highest-value work happens before drafting: establishing which buyer you are writing for, what they already believe, and which question is genuinely unresolved for them. That is a discovery problem rather than a writing problem, and it is why we argue elsewhere that you should start where the buyer's reasoning stops rather than defaulting to a house structure.
And if context moderates nearly everything, then the only defensible source of a rule about your proposals is your proposals. The general literature can tell you which factors are plausibly worth measuring. It cannot tell you what works in your category, at your deal size, against your competitors.
Common questions
Does this mean proposal quality does not matter?
No, and the evidence does not support that reading either. The Singh study found that how influence tactics combine measurably affected contract awards, so language does move outcomes. What is missing is evidence for one document-level characteristic that works across contexts.
Why do so many vendors publish a single winning factor?
Usually because the underlying data is self-reported win/loss interviews, where the buyer or seller supplies a tidy reason after the fact. The customer orientation finding, strong on self-rated performance and absent on objective measures, shows how far that can drift from what actually happened.
Is relationship strength just a proxy for incumbency?
Partly, and the research is careful about this. Shaik models strategic account status separately from relationship type, which helps, but the underlying concern is real: firms with strong relationships differ from firms without them in many ways at once. That is precisely why the controlled document study is worth running.
How many proposals would we need to analyse to learn something real?
Enough to hold the pre-existing conditions constant while varying the document. Several hundred closed opportunities with CRM history attached is a reasonable starting point. The constraint is rarely volume. It is whether the outcome and the account context were recorded consistently enough to trust.