Skip to main content

Five industries. Forty-five examples. Nowhere near the whole list

What we build is the same underneath: the same scoring, the same routing, the same discipline about what a machine decides and what a person decides. What is never the same is where your data lives, what you are allowed to do with it, and how you are allowed to make contact.

That changes by industry, and again between Europe and the US. Getting that layer wrong is how these projects fail quietly.

These are the five industries we work in most, and the moments inside each are examples rather than a menu. Most of what we build is not on this page, because it started from one company's problem rather than from a list.

So read the one closest to you. If the thing costing you money is not here, that is normal, and it is where most of our work comes from. 25 minutes is usually enough to work out what it is and whether we can build it.

01 Wholesale

The customer who quietly stopped ordering, and nobody noticed

They ordered every three weeks for six years. The last order was in April. It is now September. The person who owned them left in March, the account was reassigned inside two weeks to somebody now carrying 240 others, and it has not been called since. That is not neglect. That is what reassignment means.

Your system is excellent at telling you what was ordered and has no opinion about what was not. Everything in the building watches orders. Nothing watches for the absence of one. And this is the most recoverable revenue you have, because the customer already bought from you and nobody has to be convinced of anything.

What we build. A dormancy rule set against each customer's own rhythm rather than a blanket 90 days, because a weekly customer and a quarterly customer are two different problems. It ranks the quiet ones by what they earned you rather than what they spent, since a large low-margin customer and a small high-margin one are not the same loss.

What changes. On Thursday your inside sales team gets 12 customers who have gone quiet, with the last order, the mix and who used to own them attached, in time to plan next week's calls. Not Monday morning. Monday morning is when the orders come in.

Written for20 to 200 peoplemore than 1,000 customer records with a few hundred still orderingorder intake running through the ERP and a shared mailboxany figures on this page are illustrative

Recognise it? 25 minutes is enough for both of us to know whether it is worth building.

Request a Fit Call

Nine moments from the same building

It is not the only one.

The customer who quietly stopped ordering

A dormancy rule set against each customer's own rhythm, not a blanket 90 days.

Most agencies open with "so, what CRM are you on." In this industry the honest answers are "we are not," "there is a module nobody fills in," and "we bought Salesforce in 2019 and stopped paying for it in 2021." For our purposes those are the same answer.

How it works

  1. what you do have is Prophet 21, Eclipse, SX.e, NetSuite, Exact, AFAS or Business Central, and six years of order lines in it nobody has ever exported
  2. the order history gives you the rhythm, the margin and the product mix
  3. delivery records give you the sites, which is how you find the second branch nobody mentioned
  4. your webshop knows what the ERP never sees: who logged in and stopped, what they abandoned in a basket, and what they searched for that you do not stock, which is the only place in the building you can read demand you are not serving
  5. the shared mailbox tells you which customers are live, and we read it at account level rather than at person level, because in the Netherlands that goes to your works council before it is switched on and in Belgium it sits under CAO 81
  6. we only introduce a CRM if the work genuinely needs one

What changes

You get the system without a migration project you never asked for.

The first month. We read the order history out of your system, however it comes out, and sit with whoever knows the customers to agree what "gone quiet" means for your business, which takes an afternoon and decides whether any of this works. The first list is with inside sales three weeks after we have the data, not three weeks after you sign.

What it costs and what you are committing to. Fixed written scope at a fixed price, agreed before anything starts. The tuning weeks after go-live are inside that price, and they end when the test in your scope passes, which you call, not us. If you would rather we kept running it after handover, that is monthly and you can stop at the end of any month. No minimum term and no setup fee arriving behind the first invoice.

What we need from you, for the first build and not for all eight. A read-only export of the order lines, the customer file and the product groups, three years back, plus who owns which customer. That usually comes from whoever administers the ERP rather than from you, and their lead time is the honest variable in the three weeks, not ours. We sign whatever confidentiality and data terms you use before we touch anything, and we do not need your mailbox in month one.

When this is not worth doing. If 20 customers are most of your revenue and one person genuinely speaks to all of them, do not buy this. Buy key account management, and we would say so. And if you already run inside sales off a call list worked every week, you need a better list, which is a smaller job and we would price it as one.

The first run finds 200, not 12. Most are quiet for good reasons and you clear that backlog once, which nobody enjoys and everybody underestimates. After that it is two or three a week, forever, which is the number a person can act on.

The rule over-alerts in week one and inside sales quietly stops opening the list by week five, which is why we tune at week five rather than at handover. The export drops the delivery address about half the time, so multi-site customers look like one account until somebody notices the second branch. And the person who can settle what "gone quiet" means turns out to be away for the two weeks that matter, which is worth knowing before we start.

You already know some of your customer base has gone quiet. What you do not know is which 12, and what they were worth. That is a 25 minute conversation, not a project.

02 Manufacturing

By the time the RFQ reaches you, it was already decided

The drawing is frozen. The model already has somebody's part number in it. The approved vendor list was agreed 9 months ago in a meeting you were not in. You are quoting against a specification written around a competitor, and you will be told you lost on price.

You are not losing at the quote. You are losing 18 months earlier, at the point where an engineer picks a part and stops looking. Nobody is watching that point, because it is not in your system or the RFQ, and it is not sales' job or marketing's job, so it is nobody's. The design freeze and the purchase order are 18 months apart, and you only ever see one of them.

What we build. A watch on the moments that open a design window at your accounts and your non-accounts: a new platform or product programme, a line expansion, a permit application, a capital grant, a relocation, an equipment fleet reaching end of life, and a rule with a date that forces a redesign. Then a map of which companies each one lands on, which is the hard part. Then a ranking by how long the window is still open.

What changes. You are in the conversation while the specification is still being written, instead of arriving once the RFQ is out and price is the only variable left.

Written for50 to 300 peopleselling components or equipment into other people's designsany figures on this page are illustrative

Recognise it? 25 minutes is enough for both of us to know whether it is worth building.

Request a Fit Call

Nine moments from the same building

It is not the only one.

The new platform you heard about when the RFQ arrived

Watch your accounts' product programmes, engineering hires and prototype activity, and get to the engineer while the model is still open.

This is the hard half, and it is where most of this work quietly fails. The rule itself is public. Which of your accounts it lands on is not, and that is the part worth paying for.

How it works

  1. a rule applies to companies with a specific characteristic, a process, a substance, a throughput, a piece of equipment, a state line, and that characteristic is almost never a field in any database you can buy
  2. so we derive it, from four kinds of public record that exist in both our markets under different names
  3. what a site is permitted to do: in the US that is remarkably open, with EPA's ECHO carrying around 800,000 facilities searchable by industry code, and a state air permit naming equipment before it is installed; in Europe the same records exist but are indexed by address rather than by company, so you watch territories and match back
  4. who already holds a certification: UL Product iQ and the NRTL listings on one side, IAF CertSearch and the schemes that are genuinely complete on the other, and positive evidence only, because no register in either market is complete enough to tell you who does not hold one
  5. who has already committed capital: subsidy and incentive registers in both markets, and in Belgium the filed annual accounts, which show the year's investment in a way no private US company ever discloses
  6. who tells you what they do: association membership, equipment references on their own site, and job postings for process, maintenance and commissioning engineers, because the industry code is self-declared and stale in every country we have worked in, NACE or NAICS, and it is wrong in both

What changes

You are talking to the companies the rule lands on, rather than to everyone with the same industry code. And when a date moves, which in the US it routinely does, you know which 40 accounts to re-brief instead of all 400.

The first month. We work out which rules and standards create a purchase in your sector, which is a much shorter list than it looks. The watch on those rules is live inside the first month. Then we build the applicability list, which is the hard part. In the US the first ranked set lands around week four, because the facility data is open. In Europe it is closer to eight, because the same records arrive as legal text indexed by address. We tell you which you are in on the call.

What it costs and what you are committing to. Fixed written scope at a fixed price, agreed before anything starts. The tuning weeks after go-live are inside that price, and they end when the test in your scope passes, which you call, not us. Running it after handover, if you want us to, is monthly and cancellable at the end of any month.

What we need from you. Someone technical who can say what applies to whom, and the last 20 deals you won and the last 20 you lost, told to us by the person who was in the room, because the field in the CRM will not survive contact with the question. If you have a works council, a heads-up in week one: anything showing who followed up and who did not can need their consent, and that runs fine in parallel but not from week five.

When this is not worth doing. If you sell 40 parts a year to six customers you have known for 20 years, you do not need a system. Keep taking those six people to lunch, and we would say so. This starts paying when the number of firms that could specify you is larger than the number your commercial team can hold in their heads.

The first applicability list has a hit rate, and in week four that number is usually uncomfortable. That is the point: we measure it by running the list backwards against accounts you have already won, then spend weeks five to ten moving it. You get a list a salesperson should work in week four and a list you would defend to your board around week ten.

Permit registers change schema. A US state retires a search tool and folds it into a document portal. The Dutch planning transition split records across a legacy archive and a new system, and municipal practice is still uneven. So each feed has a freshness check that alerts us before it alerts you.

And dated copy rots. A rule you briefed a customer on in March can have its compliance date pushed by two years in May, which has happened, and if you do not catch it you have told 40 accounts something that is no longer true.

Somebody's design freezes this quarter with your part in it or without. Nothing about that waits for you to be ready.

03 Construction

Half the estimating goes on work that was already decided

The invitation arrives, the estimator prices it, and it goes out. It was never coming to you. The relationship was formed 18 months ago, the specification names somebody else's system, and the shortlist existed before the documents did.

This is not a failure of effort. Your pipeline has always been portals and relationships, and neither tells you which opportunities were live and which were theatre. So the estimator prices everything, because declining without a reason is worse than losing. Bid cost is the only line in your commercial budget that nobody measures and everybody pays.

What we build. A score against the shape of what you genuinely win, from your own bid history, so a no-bid has a reason attached that survives a conversation with the person who wanted to bid. And a watch on the early records, so the work you should be winning is visible while it is still being shaped rather than when it is being priced.

What changes. Your estimator spends the week on 4 bids you can win instead of 9 you cannot, and starts appearing on lists before they close.

Written for20 to 200 peoplespecialist or subcontract workbidding outside your home patch, your usual client type, or bothany figures on this page are illustrative

Recognise it? 25 minutes is enough for both of us to know whether it is worth building.

Request a Fit Call

Nine moments from the same building

It is not the only one.

The bid you were never going to win

Score opportunities against the shape of what you actually win, and let the estimator decline early with a reason he can defend.

By the time a package goes out to bid you are already late, and so is everyone bidding against you.

How it works

  1. the client's capital plan is public long before anything is designed, in council papers, a housing association's portfolio strategy, or a public body's advance notice of what it intends to buy
  2. the land moves before the design does, and who bought the plot is a matter of public record
  3. the permit record proves a project is real, and it says something different in every country: in the Netherlands the notice gives you the address and the works and not one party, so the names come from elsewhere; in Flanders the file itself goes online during the public inquiry, architect included; in the US it names the owner and the contractor of record, and is sold back to you as a database
  4. then a hard event says the ground is about to move, a permit going irrevocable, a start-of-work notification, a notice of commencement
  5. the bid platform is the last of these and not the first, though its own advance notices and market consultations run months ahead of the tender they trail
  6. and the parties leak the rest themselves, in projects-in-preparation pages and in who they are hiring, by region, before they announce anything

What changes

You know the project's name, and the names on it, before there is a bid list to be left off. Which of these you get depends on where you work, and we tell you that on the call rather than after the invoice.

The first month. We take one patch, not five. We establish what you genuinely win from your own bid history, which is almost always narrower than your capability statement. We map who is active there and who they already use. The first list, with a reason against every name, lands three to four weeks after the bid history is in usable shape. If it is PDFs, spreadsheet tabs and one estimator's memory, add a week, and it usually is.

What it costs and what you are committing to. Fixed written scope at a fixed price. The tuning weeks after go-live are inside that price, and they end when the test in your scope passes, which you call, not us. Running it after handover, if you want us to, is monthly and cancellable at the end of any month.

What we need from you. Your bid history however scrappy, and half a day with whoever estimates.

When this is not worth doing. If what stands between you and the work is a licence class, a recognition category or a bonding limit, that is a qualification problem and no list fixes it. Sort that first. And if you work one patch and know everyone in it, the build that pays is a different one: your lapsed contractors, the agreements you are on and have never been called against, and the clients whose supply chain is replaceable. Same method, different first month, and we would tell you which on the call.

Bid history arrives as PDFs, spreadsheet tabs and an estimator's memory, with no consistent outcome field, and normalising it is a week nobody budgets. Then the first score disagrees with the estimator about a bid he wanted, and if that argument is not had in week five the score gets ignored quietly for the rest of the year.

Public data quality is uneven and we would rather say where. Dutch planning data is the best of the three in Europe, Flemish is workable, Walloon is fragmented per commune, and Luxembourg is effectively manual. In the US it depends which counties you work in and whether somebody already sells the aggregate.

You will price next quarter's bids either way. The only question is whether the estimator knows which ones were decided before they arrived.

04 Staffing and Recruiting

Two hundred old clients. Some of them are hiring this morning.

You placed for them in 2021, or 2019. The relationship did not end, it just stopped having a reason to continue. Their careers page has 4 openings on it today, one of them open for 30 days, and nobody in your building knows, because nobody's job is to look.

This is not a database problem. Your system knows exactly who you placed and what it was worth. What it cannot know is what happened at those companies afterwards, because nothing outside your own record has ever been pointed at it. It will answer any question you ask it. It only knows what you told it.

What we build. A watch on every company you have ever invoiced, their careers pages, the boards, the aggregators, joined back to your placement history, so an opening is not a lead, it is your opening at your client. Routed to the recruiter who owned the relationship, with what you placed, what it billed and who you dealt with attached.

What changes. A recruiter opens Monday with 6 old clients hiring now and a reason to call each one, instead of a list of companies somebody told them to warm up.

Written for10 to 60 recruitersfive years of placement history in an ATSperm, contract or bothany figures on this page are illustrative

Recognise it? 25 minutes is enough for both of us to know whether it is worth building.

Request a Fit Call

Nine moments from the same building

It is not the only one.

The lapsed clients who are hiring right now

Watch every company you have ever invoiced, joined back to your placement history, flag the role that has been open 30 days because that is the one their own team cannot fill, and route matches to the recruiter who owned the relationship with what you placed and what it billed attached.

Bullhorn if you are on Bullhorn, Carerix, OTYS, Mysolution or Recruitee if you are not. The point is the same and so is the work. The data is not the problem. Nothing is asking it anything.

How it works

  1. your placement history tells you which roles you genuinely win, which is narrower and more specific than your website claims
  2. your billing history tells you which clients were good, which is not the same as the ones you remember fondly
  3. your own placed candidates tell you where you put people, which is the first half; where those people are now is the half your system has never been told, and finding out is both the work and the warmest route into a company you have never sold to
  4. and there the record stops, because an ATS is a closed world
  5. the automation module you already pay for fires when your own fields change, an assignment end date, a last-activity date, which is useful and is not this
  6. this fires when the outside world changes, and joins that back onto your own record, which is the part nothing in the box does

What changes

The system you already pay for starts volunteering things instead of waiting to be searched.

The first month. We pull the placement history out of your ATS, whichever one you are on, and agree the match rule with a recruiter who will tell you the truth about which clients were good, which matters more than the data does. The first watchlist is live three weeks after we have API access. That access is issued by your ATS vendor, not by you or us, and on some tiers it is a commercial conversation first, so the request goes in on day one because it is the long pole.

What it costs and what you are committing to. Fixed written scope at a fixed price. The tuning weeks after go-live are inside that price, and they end when the test in your scope passes, which you call, not us. Running it after handover, if you want us to, is monthly and cancellable at the end of any month.

What we need from you. ATS access and one honest recruiter.

One thing we check first. If nobody has closed out a job order with a reason in two years, half of this does not work yet and we will say so in week one rather than in month three. Placements and invoices are almost always clean, because finance made them clean. Everything a recruiter typed is a coin flip. We build on the first and treat the second as a hint.

One line on scope. We build the commercial layer: which of your clients is hiring, which of your own placements to call. We do not build anything that scores a candidate for a hiring decision, because that turns your ATS into a regulated product on both sides of the Atlantic and it is not what you asked for.

When this is not worth doing. If you are all perm, under about 80 companies you have invoiced, and nobody has left in two years, one recruiter with a calendar reminder beats anything we would build. This starts paying the second time a desk changes hands.

Coverage is honest rather than total. Clean where the client runs a modern ATS, which is most of a US book and a good half of a European one, a licensed feed for the rest, and a named list in week one of the clients we cannot cover reliably. A watchlist you think is complete and is not is worse than one you know has holes.

The first version surfaces too much, and a recruiter shown 30 things ignores all 30, so we tune at week five. Careers pages get rebuilt and the feed for that client dies silently, so each one has a freshness check that alerts us before it alerts you.

Four of those companies are hiring this week. You will find out from us, from a competitor's placement, or not at all.

05 Software

You moved upmarket, and the pipeline did not come with you

The board asked for bigger deals. You hired two AEs and pointed them at a market nobody in this building has ever mapped. The self-serve signups still arrive and they are still small. The list of companies worth a real sales cycle exists only in the founder's head.

Meanwhile your product knows something nobody has asked it. The accounts that renewed, expanded and referred have a shape. It is sitting in the usage data, the won-deal notes and the support history, and it has never been used to decide who to sell to next.

What we build. Software, not a subscription. The pattern your best accounts share, derived from your own product data and your won deals rather than from an industry code and a headcount filter. Every company that matches it, mapped, because nobody sells that list. And the signals you already emit landing in the queue your team already works in rather than a new surface nobody opens. You own the code at the end of it, and the day you want to take it in-house you can.

What changes. Your AEs work one ranked list built out of what your own product already knows, instead of a bought list and a dashboard.

Written fora B2B software company moving upmarket8 or more AEsproduct usage that lands somewhere queryable, or an engineer willing to make itany figures on this page are illustrative

Recognise it? 25 minutes is enough for both of us to know whether it is worth building.

Request a Fit Call

Nine moments from the same building

It is not the only one.

You moved upmarket and the pipeline did not come with you

Derive the pattern your best accounts share from your product data and your won deals, then map every company that matches it.

Everyone selling you this has an opinion about it. Ours is written down before anything is built, because in the EU it now has to be.

How it works

  1. a model reads the things a rule cannot, the support thread, the call record, the security review, the note somebody typed at eleven at night, and it is better than any rule you could write at saying what is going on in an account
  2. so the model proposes, with its reasoning attached and the evidence linked, never as a number with no argument behind it
  3. a written policy decides what it may do on its own, which at the start is close to nothing
  4. anything that spends a person's hour or reaches a customer goes through a queue somebody clears in four minutes
  5. the prioritisation starts as a rule you can read in one line, and once you have a few hundred won and lost outcomes it becomes a small model fitted on your own results, still deterministic at inference, still able to say which feature moved the score, and better than a threshold somebody guessed in month one
  6. what it never becomes is a language model's opinion, because you cannot audit that and you cannot reproduce it
  7. every extraction gets scored against a set we label by hand before it goes live, and anything the model is unsure of goes to a person rather than into the queue
  8. and we do not write your emails, because AI-written opening lines are part of why your reply rate fell

What changes

You can say out loud, to your board or to an auditor, why any account was prioritised, who decided it, and exactly what the system was allowed to do without asking a person.

The first month. We get product events somewhere they can be scored. If they already land in a warehouse, a replica or a CDP, that is read access and half a day of an engineer. If they do not, instrumentation comes first, which is real engineering work on your roadmap and not ours, and we tell you which of the two you are in on the first call rather than in week two. We also need write access into whatever your AEs work from, because a score that stays in a spreadsheet changes nothing.

What it costs and what you are committing to. Fixed written scope at a fixed price. The tuning weeks after go-live are inside that price, and they end when the test in your scope passes, which you call, not us. Running it after handover, if you want us to, is monthly and cancellable at the end of any month. You own the code.

What we need from you. Somebody who can say what a meaningful signal means in your product, which nobody outside your company can do. And whoever signs the data processing agreement, because we will be reading your customers' usage data and you should make us prove how we hold it before you let us near it.

When this is not worth doing. This is a question of attention, not volume. If your AEs carry fewer than about 20 live accounts each and open every one of them daily, they will beat any rule we write and you should leave them alone. Below about 10,000 a year in contract value this should never reach a human, and the right build is lifecycle email and a better self-serve flow.

The scoring model goes stale every time you ship. You change the onboarding flow in March and a pricing tier in June, and every threshold is now measuring something that no longer exists. A stale model does not error. It quietly over-alerts, the AEs stop opening it inside six weeks, and that is the only real telemetry you will ever get.

That is why the tools you have already bought stopped working. It was never the software; nobody owned the part that decays. So we read your changelog and retune when you ship, and the code is in your repo, so the day you want this in-house you can take it.

The list your AEs work on Monday will come from somewhere. Right now it comes from whoever shouted loudest on Friday.

06Not on this list

Not here?

These five are where we have the most repetition, not the edges of what we can do. Underneath, every moment on this page is the same shape: something changed at a company, somebody in your business should have known, and nothing was watching. That shape does not care what you sell.

A logistics operator: the customer whose volume halved over four months and nobody asked why. A clinic group: the referrer who stopped referring, 11 months ago. A waste processor: the industrial customer whose tonnage dropped when they changed a process, and nobody asked what changed.

Where the data comes from

One thing about where the data comes from.

Every signal on this page has a source we can name, and we write the lawful basis down before we build rather than after somebody asks. The first message a person receives says where we got their details and how to stop.

The rules are not the same on both sides of the Atlantic, and that is most of the work. Email: the US lets you send cold and opt out, the Netherlands mostly does, Belgium does not without an opt-in. Calls: about a dozen US states need every party's consent and one of them charges 5,000 dollars a call when you get it wrong, so we announce every recording everywhere and stop arguing about it. Anything touching a shared mailbox goes past your works council in the Netherlands and sits under CAO 81 in Belgium. And website visits resolve to a person in the US, where that is lawful, and to a company in the EU, where it is not without consent.

If a signal cannot be gathered lawfully in a market you sell into, we say so on the call and build the one that can. That is a shorter list than people expect, and knowing which is which is most of what you are buying.

Fit Call

Twenty-five minutes, and you will know.

Not a demo, not a pitch, not free consulting. Sometimes the next step is nothing. We hold calls in European and US hours, and two of us are on every call. One of the two is your contact from then on, and the work is done by all three of us.

Request a Fit Call

or write to contact@clarityrev.com

25 minutesNo charge