I have spent twenty-five years on the selling side of enterprise data, which means I have also spent twenty-five years watching buyers make the same handful of mistakes, over and over, and pay for them. Some of those mistakes made me money, and I still wished the buyer had known better, because a client who overpays or under-buys is a client who churns, and a churned client is worse for me than a fair deal ever was. So here is the guide I wish more of my buyers had read first. It is written by the person across the table, it is mostly on your side, and I will be honest at the end about the part that isn’t.
1. You build on a field that isn’t there
The first and most expensive mistake happens before price ever comes up. You license a dataset because it contains the field you need, you confirm the field exists in the schema, and you never ask the one question that matters: is it actually populated where you operate? A field can be present, labeled required, and empty across half the country. Coverage is wildly uneven; a single county can hold more records than dozens of states combined, and a rural jurisdiction may have almost nothing. If your product depends on a field that is full in the demo and hollow in your target markets, you will find out in production, after you have already built around it. Ask for a field population report, broken out geographically, before you sign. Do not build a house around a load-bearing beam that turns out not to exist.
2. You don’t know who you’re actually buying from
In data, the company selling to you is very often not the company that collected the data. There is a short list of true sources who gather records directly, and a longer crowd of resellers who license from them and mark it up. Buying from a reseller is not automatically wrong, but you should know that you are doing it, because it changes your price, your latency, and your exposure. A reseller’s source can realign underneath them mid-contract, and suddenly the data you standardized on is arriving from somewhere new. Ask, plainly: do you collect this yourself, or license it, and from whom. If they won’t answer, that is itself an answer.
3. You get the license scope wrong
The number on the contract is not really pricing the data. It is pricing what you are allowed to do with it: which datasets, which applications, how many divisions, and above all your redistribution rights. Buyers get this wrong in both directions. They under-license, sign for a narrow use, and then cannot legally do the thing their roadmap needed, forcing an awkward renegotiation from a weak position. Or they never realize that redistribution, the right to pass the data or its derivatives to their own customers, is the single most valuable term in the document, and they give it away or buy it blind. Watch the language itself, too: much of it was written before the modern internet and still carries modeling and redistribution restrictions that no longer match how anyone works. Contracts get amended and products get bolted on, but the old clauses rarely get rewritten. Read the grant of license as if it were the product, because in a licensing deal it is.
4. You lock the layout into the contract
Most vendors attach the field layout to the contract as an exhibit, and most buyers let them. If you can, don’t. Ask instead that the layout be governed by a documented production process, with changes handled as production change orders rather than contract amendments. The reason is simple, and to anyone who has built software it is a little strange: a layout is a thing that is supposed to change. You do not amend your agreement with a software vendor every time they ship an update, yet in data that is exactly what people do, freezing into a legal document the one part of the arrangement that is guaranteed to move. When a new standard field appears, an unparsed version of a company name, say, you should simply receive it, the way you receive any other improvement, not renegotiate paper to get it. No one can seriously argue that a single new field is worth the fully burdened cost of drafting an amendment, on either side.
Vendors do need to define their deliverables in the contract. But the right unit is the category, not the field. You are licensing tax, deed, mortgage, and the proprietary arrangement of that data under those headings; you are not licensing a particular column in a particular position, any more than you contractually specify the hour your updates arrive. There is a legal grain to this, too: the raw facts in a dataset are not themselves ownable, and what the law actually protects is the original selection and arrangement of them. The arrangement under the categories is the product. A field is not. So define the categories in the agreement, let the layout live in the production documentation, and you spare both sides the post-contract amendment waterfall, the slow drip of oh, we forgot this and oh, we discussed that and it isn’t in here, that eats months and goodwill.
And a warning that follows from the first mistake in this guide: putting the layout in the contract does not mean those fields will be populated. The two have nothing to do with each other. A contract exhibit is a promise about structure, never about coverage, which is one more reason the field population report never belongs in a contract, it is always changing. Reserve amendments for what actually changes the deal, a whole new category of data, foreclosure, commercial, something substantive, and route the rest as production change orders. A field here or there is production. A new category is product. Lock the categories, free the layout, and you will thank yourself at every renewal.
5. You run a bake-off you can’t judge
When buyers feel unsure, they reach for a bake-off: line up three vendors, run the same test, pick the winner. It feels rigorous. It is often theater. The buyer running the test is, almost by definition, evaluating the very expertise they are outsourcing because they lack it, frequently by hiring a third party who lacks it too. The bake-off measures whatever the test happened to measure, which is rarely the thing that breaks you in production, and it lets everyone point at a leaderboard instead of owning a judgment. If you must test, test against your actual use in your actual markets, and be honest with yourself about what the test cannot see.
6. You cut data to save a little now
I have watched a buyer decline a set of fields to shave a small amount off the price, and I have watched the same buyer come back two years later, desperate for exactly those fields, after they had already chosen a different vendor over the gap. Storage is cheap. The data you don’t think you need is cheap to carry and ruinously expensive to reconstruct after the fact, because by then the relationship, the context, and sometimes the original source are gone. If a field is even plausibly on your roadmap, take it now. The few thousand dollars you save by cutting it is the most expensive money you will ever pocket.
7. You treat a rent like a price
Enterprise data is not priced like a good. It costs almost nothing to reproduce and has no competitor across the street, so what you are paying is closer to a rent than a price: value-based, scoped by rights, and quoted as a range. Buyers who miss this burn their energy grinding the headline number down a few percent while ignoring the terms that actually govern their outcome, the scope, the term, the redistribution rights, the renewal mechanics. The number is the least negotiable and least important part of the deal. Spend your leverage on the rights and the flexibility, not on shaving the rent.
8. You go over the rep’s head
This is the costliest mistake in the guide, and it is the one buyers are proudest of. The move is to treat the whole thing as a price war: hide your budget, set vendors against each other, run down the product, threaten to walk or to just build it yourself, and when the rep won’t fold fast enough, escalate over their head to someone senior, on the theory that the real boss will cut the real deal. It is exactly backward. The rep is the most motivated advocate you have inside that company, because their compensation is built on closing your deal, which means their interest is aligned with getting you to yes at a number that clears. Go around them and you do not reach a more generous authority. You reach a busy executive who has no idea who you are, no incentive to care, and who prices from a very different place.
In my experience senior executives essentially never grant the discount the buyer is picturing. If anything they price higher, because they are not salespeople, they do not price deals for a living, and they systematically overvalue their own product, convinced it is simply better in a way that even they can never quite make hold up. The rep was quietly translating that overvaluation into a number you could live with. Remove the rep and you get the overvaluation raw. The hardball buyer who is sure he is being shrewd talks himself out of his own best ally and into paying more.
Watch, too, how eager the rep is to set up that meeting. A rep who fights to keep you out of the corner office is protecting you; a rep who cheerfully walks you in already knows the number is about to go up, and knows it pays them, because their commission rides on the figure that clears. When your own advocate is glad to grant your request, the request is against your interest. And the meeting is a trap even if you play it perfectly, because you cannot walk into a vendor’s chief executive and plead poverty; telling a public company’s CEO that you are broke and unproven would be absurd. The only posture available to you is strength: we just raised forty million, we are going to be enormous, get in early. But strength is exactly what a vendor prices up. You have been maneuvered into the one move that raises your own price, and you feel powerful the whole time.
The story in your head and the story across the desk are not the same story, either. You are picturing a partnership, a believer, an eventual acquirer. The executive is thinking: another reseller with fresh money doing something possibly interesting, I will take my slice, let them go prove it in the market, and if it works, which it usually does not, I will look at buying them then. You believe you are being courted. You are being metered.
And if the executive, flattered by the partnership pitch, routes you to corporate development, you have not won; you have wandered into the most expensive free deal in the building. Corporate development does not onboard data users, because that is not what the function is for; it hunts partnerships and acquisitions. It may even hand you the data at no charge, and paper it, in exchange for a right of first refusal to acquire you later. Then it passes you to the same solutions engineer who serves the paying accounts, except now no salesperson is doing the real work, the scoping, the deliverables, the patient here-is-what-you-actually-need. Production ends up playing the role sales is supposed to play, for an account that is not paying, and resents it accordingly. So the buyer who went around the rep to get it for free ends up under-supported, quietly disliked by the people now stuck serving him, and less successful than the client who simply paid. You bought the asset and forfeited the outcome.
There is a deeper reason this style feels natural to some buyers and foreign to others, and it has nothing to do with who anyone is; it is about the market they learned to buy in. Economies run on different amounts of trust in strangers, and the ones with thinner impersonal trust breed a more adversarial, extract-what-you-can script, while higher-trust markets assume the counterparty is a repeat player worth cooperating with. The pure arbitrage buyer, the one who treats every dataset as a commodity to be squeezed to the floor, is running the first script. It works on true commodities. It fails badly on a relationship good you will depend on for years, because it burns the one thing that actually protects a buyer who cannot benchmark the price: the trust of the person representing you.
The whole guide in one idea
Everything you buy in this category is a rent on something you cannot make yourself, sold by someone whose real product is the relationship around it. Buy it like a haggle and you will win the pennies and lose the dollars: cut the field you needed, sign the license that boxed you in, and alienate the one person whose job was to get you a fair number. Buy it like the start of a relationship, ask the boring questions early, spend your leverage on the terms, and keep your rep on your side, and you will pay less for more, every time. The person across the table wrote this. Believe the part where it is on your side, and stay awake for the part where it isn’t.
On why some markets breed a more adversarial buying script: Joseph Henrich, The WEIRDest People in the World (2020), and Francis Fukuyama, Trust (1995), on generalized versus kin-based trust and how it shapes economic behavior. On why the arrangement, not the raw fields, is the protected product: US copyright protects only the original selection and arrangement of a data compilation, not the underlying facts (Feist Publications v. Rural Telephone Service, 1991), and the US recognizes no separate database right. On why the number is a rent rather than a price, see the companion piece, "Why Nobody Sweats the Spread."