Trust is a promise made, delivered, and repeated. It is also how the years clock gets built, in increments small enough to keep.
I have been offered fancier definitions than that, usually by someone selling a workshop, and I have never needed one. A statement goes out, the delivery matches it, and the cycle runs again. Every repetition deposits a little belief into the account, every miss withdraws more than the deposit was worth. The balance decides whether a stranger believes the next thing you say.
A Trust-Promise Pair is the smallest honest transaction a business can have with a person: one specific promise, and a delivery that matches it exactly. Repeat that across every page, every email and every support ticket, a thousand times over, and the accumulated record becomes a reputation. That reputation is the thing the Trust Algorithm is actually trying to measure.
The mechanism runs at two scales at once. At the micro scale, one person receives one specific promise, watches it get kept, and trusts the business slightly more than they did ten minutes earlier. At the macro scale, thousands of those small transactions pile up into Reputation, Reputation reads as Authority, and Authority is what rankings, referrals and recommendations are made of. The macro is built from the micro. There is no other supplier.
What a Trust-Promise Pair Actually Looks Like
A pair is not vague, not aspirational, and not designed to be true eventually or true on average. It is a claim narrow enough to be checked.
An article opens by saying it will explain something in five minutes, and five minutes later the reader understands the thing. Not the thing plus three adjacent topics they were quietly required to read first, just the thing. A contact page says enquiries get a reply within 24 hours, and the reply arrives within 24 hours, every time. Not within 48 hours usually, and not within 24 business hours with the weekend silently excluded. A calculator says it will show a person what they can afford, they put their numbers in, and it shows them, honestly, with no email capture standing between the question and the answer. A guide says it will walk through a process step by step, and no reader hits step three only to discover they needed something from step seven.
None of this is revolutionary, and no money changes hands in any of it, but something does. The person on the other side now believes the business a little more than they did, not because a grand claim was made, but because a small claim was kept.
The biggest pair I ever ran was two words long. At BookPrint, the book printing company I built, the promise was bookshop quality. It sounds a bit daft until you know that self-published books have a reputation for looking self-published, and that most people can tell just by picking one up. We could keep the promise because the machinery could keep it. Our printers ran oversized 13 by 19 inch sheets, which fit four copies of a standard 6 by 9 inch book where the ordinary New Zealand sheet fits two. The same book, at half the cost. The inside paper came from Austria, lightweight but bulky, exactly what a premium paperback uses. The cover printer was a Canon, chosen because its toner held laminate where other machines let it peel at the crease. Peeling laminate is the fastest way to make a book look like amateur hour. And we said no to nearly everything that was not a book, which is the part everyone skips. A promise you can keep is mostly a capability you already have, described accurately.
One customer told me her existing printer had quoted cheaper than I had, on literally the same machine I owned, and she went with us anyway, because we were the book specialist. That is what a stack of kept promises buys. It outweighs price.
Why This Matters for the Algorithm
The Trust Algorithm does not care about a logo, a byline or a testimonial wall. It is trying to estimate one thing, which is whether people trust a source enough to pass its work along to other people.
People share what they trust. They do not share what feels like marketing, and they do not share the article that promised a solution and delivered a pitch. When a business closes pair after pair, the shares begin: a link emailed to a colleague, a page dropped into a work chat, a name mentioned over a beer. Each share is a small credibility vote from someone with nothing to gain, and votes like that are the only kind the algorithm counts with much confidence.
Someone lands on a page through search, the page promises to solve one specific problem, and it does, with no padding and no upsell riding along. That person knows nothing else about the business, but they are slightly more inclined to trust it than they were, and when the same problem crosses a colleague's desk, they pass the page on. The colleague has the same clean experience and does the same thing. Now there is a widening circle of people who have each had one honest transaction, and from the outside, the sum of those transactions reads as authority. More visibility creates more chances to keep the promise, which creates more shares, which creates more visibility. The loop feeds itself.
A broken promise runs the loop in reverse. The page promises one thing and delivers another, nobody shares it, the ranking stalls, and fewer people ever get the chance to be disappointed. That is the algorithm doing its job properly.
Authority, in algorithmic terms, is exactly that: not what a business claims about itself, but the accumulated evidence of promises kept, counted through the behaviour of the people on the receiving end.
The Bar Is Shockingly Low
Most businesses break pairs constantly without noticing, because nobody inside the building reads the promises the way a stranger does.
Subscribe for weekly insights, says the form, and then the sales emails arrive twice a week. Book a free consultation, says the button, and then 45 of the 60 minutes are a pitch. We put clients first, says the homepage, while the pricing is structured to maximise commission. Templates you can use immediately, says the resource library, from behind an email gate. Each of these made a specific promise and delivered something adjacent to it, and adjacent is broken.
Take a restaurant, a hypothetical one, that promises locally sourced ingredients. The produce is local and the meat is local, but the seafood arrives frozen from Thailand, so the promise was two-thirds kept. Two-thirds kept is a strange quantity, because the diner who notices does not complain to the manager. They mention it once, quietly, to their partner, and the restaurant drops off the recommendation list without ever learning why.
Or a tradesperson who says I'll call you back tomorrow morning, and calls the day after. Tomorrow morning is specific, that is what made it worth saying, and the specificity is exactly what makes the miss visible. The job goes to someone else, and the tradesperson never finds out that the first promise was the one that failed.
Nobody announces a withdrawal from the trust account. They stop converting, something felt off, and they move on. One broken promise is small, but the aggregate across hundreds of interactions compounds into a reputation for being another business that says things without quite meaning them. In a market as small as ours, that reputation travels on its own legs.
The Mortgage Calculator
I spent years in mortgage and property marketing, at mortgagehq and homes.co.nz, so let me build the hypothetical in the industry where I watched these mechanics up close.
Picture a mortgage calculator built the honest way. Five questions, three minutes, no email capture, no registration, no callback. The promise is we'll show you where you stand, and the meaning underneath it is specific. What these numbers say you can afford, not what a lender would prefer you to believe, and not the figure that maximises anyone's commission.
The person using it is anxious. They are wondering whether they are being ridiculous, whether the house is a mistake, whether they will overextend and spend a decade regretting it, and they want a direct answer. Most calculators refuse to give one. They are sales tools wearing a tool costume, hedged with disclaimers and gated behind contact details, engineered to route an anxious person toward somebody whose job is to close.
The honest calculator answers the question. It shows its working, lets the person adjust the assumptions and run it again, and asks for nothing beyond the numbers needed to do the maths. The absence of capture is the whole design decision, and it is the one every stakeholder meeting would try to reverse.
Now reason it forward. Some of those people will eventually ring the firm, and I would expect those calls to feel different. The trust transaction already happened, the caller has already concluded the firm was not routing them anywhere, and the conversation starts at implementation rather than persuasion. I cannot hand you a conversion percentage for a calculator I have described into existence, but the logic only points one way. Small promise, kept cleanly, standing alone.
Scale and Compounding
Every surface you own is making a promise, whether you wrote it deliberately or not. The homepage implies it can explain the business in a minute, the pricing page implies transparency, the support address implies a reply, the invoice implies no surprises. Each kept promise is individually tiny, but run the count: fifty interactions in a month is six hundred in a year, and a reputation is nothing more than the running sum.
The strangest evidence I have that promises compound came from my lawn-mowing round. I started it at nineteen, after quitting university, and I paid for flyer drops, ten thousand at a time. Lawn mowing and a picture on the front, section clearing and one-off jobs on the back. Long after I stopped mowing, the calls kept coming. The longest was six or seven years after I finished, from someone who had kept the flyer on their fridge or in a drawer that whole time. One piece of paper, one specific offer, quietly making its promise for the better part of a decade. I thought it was insanity then, and I still do.
A website is the same object with better distribution. It sits there making its promises to every visitor at every hour, and whether it compounds trust or drains it depends entirely on whether those promises are kept.
The Volume Trap
Most businesses cannot run this system, and the reason is one contradiction sitting in the middle of their marketing.
The content calendar demands volume: post daily, feed the algorithm, twelve blog posts a month, five social updates a week, the machinery of content marketing grinding away. Trust-Promise Pairs demand something else, because every published piece is a promise that this will be worth someone's time, and twelve posts a month is twelve promises. Keep all twelve and the trust compounds. Keep eight and break four, and the withdrawals outweigh the deposits, since a miss costs more than a keep earns. A business cannot feed both, and most have never noticed they are choosing.
It is quarterly review time, the marketing team presents, and the slide says twelve posts, forty-eight social updates, three white papers, engagement up 15 per cent. Everyone is pleased, the budget rolls over, and the cycle repeats. Nobody in the room asks whether the twelve posts kept their promises, because volume is easy to measure and quality is not, so volume quietly becomes the proxy for everything. Leadership rewards it, the team supplies it, and the whole operation optimises for the Beast while the trust account slides into overdraft.
The businesses that win long term publish less and keep more. Their dashboard looks worse in Q3, and the compounding shows up in conversion, retention and word of mouth two or three quarters later. That is precisely too late for the review that sets the budget. The Beast pays immediately, pairs pay slowly, and most businesses cannot see past the next review.
The Audit
Go through your website, every page, every form, every email sequence, and ask three questions of each.
What promise does this make? Be specific about it. We're here to help is not a promise, it is wallpaper. This guide shows you the three questions to ask an accountant before choosing one, that is a promise, narrow enough to be checked. If you cannot find a specific promise on a page, the page is broken before anything was kept.
Does it keep that promise? Read the page the way a stranger would, with no memory of what it was meant to say. If your form promised no capture, does it capture. If you promised a reply inside 24 hours, when did the last one actually go out. Every no here is an open withdrawal, running daily.
Is the promise specific enough to be kept or broken at all? A vague promise is teflon, nothing can be measured against it, which is exactly why nervous businesses write them. This will help you is unfalsifiable. This will save you at least an hour of admin a week can be checked by anyone with a calendar, and checkable is the entire point.
Wording matters more here than most people expect. At BookPrint nobody was allowed to use the word price, only cost, because price sounds like some bullshit you made up to make money, and cost sounds like the cost of the work. One word, and the promise underneath the number changes from trust me to check me.
A thirty-page site gives you ninety answers, which sounds like a project and is honestly more like an afternoon. I would expect you to surface the same two findings every audit surfaces: one grand vague promise repeated on every page, we deliver excellence or one of its cousins, and one specific promise broken constantly, usually the 24-hour reply that no single person owns. There will also be pages promising nothing at all, the culture page, the mission statement nobody reads, and those are noise. Delete them, then fix your broken promises in order of how many people each one touches, starting this week rather than eventually.
The NZ Business Culture Angle
New Zealand business culture ran on handshake trust for generations. A deal was a deal, a person's word carried weight, and the reference check happened at the pub. That system worked because everyone stood close enough to collect on a broken promise.
A handshake does not travel through a website. Nobody can read a face in a search result, and the online market is full of strangers making identical claims in identical fonts, so the old signal is simply unavailable. Trust-Promise Pairs are the digital replacement for the handshake: small, specific, direct, checkable. Keep your word on something small, and a stranger will extend you credit on something larger.
In a country of five million, the compounding runs faster than almost anywhere, because the degrees of separation are short. Someone uses your site, has a clean experience, mentions it at the pub, and the next person arrives already half-trusting. Small countries run on word of mouth rather than ad budgets, and word of mouth only works if every interaction holds up when it gets retold.
The longer telling of BookPrint, the lawn round, and what they taught me about promises is in Marketing Curious: Working the Noise. This page is a rendering. The seed is the source. The book is the story of building it.
Read next:
- The Algorithm: How the Trust Algorithm actually works
- Brand: Building brand that survives scrutiny
- The Diagnostic: Audit your own Trust-Promise Pairs