The $99,000 Invoice: What AI Recruiting Vendors Won

The pattern is painfully common. An HR leader approves a $48,000 annual subscription for an AI recruiting platform. The business case is built around that number. The CFO signs off. The ROI calculations use it as the denominator. Nine months later, the invoice arrives.

$147,000.

The additional $99,000 was all technically in the contract. Integration costs that weren't anticipated. API overage charges nobody mentioned during the demo. Premium support fees listed in Appendix C, page 47, in eight-point font. Implementation consulting that "might be needed" turned into "absolutely required." And a bias audit mandated by new legislation that nobody mentioned was coming.

Gartner's research on AI implementation costs tells the same story: organizations routinely underestimate total project costs by 2-3x. What looks like a $50,000 investment becomes $150,000 by the time you count everything.

The response from vendors when buyers complain? "These are industry-standard costs. Every enterprise client pays this." As if the buyer is the problem.

I wish I could tell you this story is unusual. It isn't. In my work building an AI recruitment platform, I've had hundreds of conversations with HR leaders, and the pattern is depressingly consistent: vendors quote license fees, buyers build business cases around license fees, and then reality arrives in the form of invoices that make CFOs question whether HR knows what they're doing.

Some research puts it at 85% of organizations misestimating AI project costs. Sounds abstract until you're the one explaining to your CFO why you're 200% over budget. What those statistics don't capture is that smart, experienced people are getting played by a sales process specifically designed to hide the real numbers until it's too late to walk away.

This guide is my attempt to fix that. Not with vendor marketing language, not with sanitized "best practices," but with the truth about what AI recruiting technology actually costs—and the games vendors play to hide it.

The Anatomy of a Vendor Quote

Here's something most buyers don't realize: that license fee they quote you? It's often calculated to land just under your procurement threshold for executive approval.

I've sat in rooms where this math is done. If your company requires VP approval for purchases over $50,000, you'll get a quote for $48,000. If the threshold is $100,000, the quote will miraculously land at $95,000. This isn't paranoia—it's sales strategy. The goal is to get the deal signed before the real costs become visible.

In a comprehensive analysis of enterprise AI implementations, McKinsey found that software licensing typically represents only 25-35% of total first-year costs. For AI recruiting specifically, the ratio can be even more extreme. I've seen implementations where licensing was 18% of year-one spend.

Let me walk you through a typical deal pattern that Aptitude Research and other HR tech analysts have documented. A mid-market company—manufacturing, retail, healthcare, it doesn't matter—approves a $60,000 annual license. Clean number. Easy to budget. Easy to defend.

Here's what happens.

Implementation quoted as "typically 2-4 weeks" takes eleven. Nobody lied— there were legitimate complications with legacy data. But every week past the original estimate costs $1,500, and by week eleven the buyer has spent $18,000 on implementation alone. The sales rep's response: "Implementation timelines are estimates, not guarantees." True. Also completely unhelpful.

The integrations are worse. "We integrate seamlessly with Greenhouse," the demo promised. What "seamlessly" actually means: "We have an API. Your IT team will need to build the connection, or you can pay us $12,000 to do it." Then HRIS integration—another $8,000. Then LinkedIn Recruiter data syncing—$6,000 more. Each one "optional," each one essential for the system to actually work.

Candidate data is a mess—fifteen years of accumulated records with duplicates, typos, and missing fields. The vendor's data team spends three weeks cleaning it at $5,000 per week. Nobody mentioned data hygiene during sales. Why would they? It doesn't close deals.

The license includes "basic training"—a two-hour webinar. For eight recruiters to actually learn the system, you need "advanced training." That's $2,000 per person. Another $16,000. Then recruiters start actively resisting the new system, so you hire a change management consultant for $10,000. Then you realize somebody needs to actually manage the platform day-to-day—run reports, troubleshoot issues, maintain the integrations. That costs roughly $12,000 in allocated staff time in year one.

Finally, NYC Local Law 144 goes into effect and suddenly you need a bias audit. The vendor offers to provide one: $20,000. And after the third time your team waits 72 hours for a response to a critical issue, you upgrade from "standard" to "premium" support. Another $12,000. Which, I suspect, is exactly what the vendor was counting on from the start.

Final year-one cost: $183,000. Three times the quoted license fee.

That's not three times the quoted price. That's three times the quoted price plus the emotional cost of explaining to your CFO why you're 205% over budget.

Something needs to be said.

I run an AI recruiting company. When I describe the tactics vendors use to obscure costs, I'm describing an industry I'm part of. I've sat in pricing meetings where we debated whether to bundle implementation fees or list them separately—knowing that bundled pricing looks cheaper even when the total is identical. I've watched our sales team emphasize features that close deals rather than features that deliver value. I've seen the spreadsheets where we calculate exactly how much to charge before a buyer needs VP approval.

I'm not writing this to pretend OpenJobs is different from every other vendor. We have the same incentives. We face the same pressures. Some of what I'm about to describe, we've done ourselves—or been tempted to do.

I'm writing this because I've watched this game long enough to believe someone should explain the rules out loud. If this helps you negotiate better with my competitors, good. If it helps you negotiate better with me, also good. I've decided I'd rather compete in a market where buyers understand what they're buying. That market is harder, but it's more honest.

And since I'm being honest: I've made mistakes too. Last year, we quoted a mid-sized logistics company $40,000 for implementation. We came in at $58,000. Their legacy system had dependencies we hadn't properly scoped, and untangling it took three weeks longer than planned. I could tell myself it wasn't our fault—their documentation was terrible—but that's exactly the excuse I criticize other vendors for making. We should have done better discovery. We should have built in more buffer. We didn't. The client was frustrated, and they had every right to be.

I tell you this not from moral high ground, but from the same mud as everyone else—trying to do better.

The Dirty Tricks

Contract analysis firms that specialize in HR technology have documented the patterns. Everything that costs money is in the contract—just buried. API rate limits and overage charges in section 4.7, sandwiched between definitions of "Authorized Users" and "Service Level Objectives," written in language that requires a law degree to parse.

Not every vendor is out to screw you. Some genuinely believe their quotes are accurate. They've been trained to sell that way, and most of them have never seen what happens after the contract gets signed.

But some vendors know exactly what they're doing. These contracts are masterclasses in what I call strategic opacity—not lying, exactly, but arranging truths in ways designed to be overlooked. Here are the tactics I've seen most often:

The "Unlimited" Trap

"Unlimited users!" the vendor says. "Unlimited job posts! Unlimited candidates!"

What they don't say: the platform has usage caps on AI features. Every resume screened by AI counts against your monthly allocation. Every chatbot conversation. Every candidate matching operation. Go over your allocation, and you're paying 2x-3x per-unit overage fees.

G2 and TrustRadius reviews are full of these stories: "unlimited" platforms that cost an extra $40,000 in overage fees during a hiring surge. "They never once mentioned usage limits during the sales process. When I complained, they pointed to page 23 of the terms of service."

The "Integration-Ready" Lie

Nearly half of HR leaders say integration issues stopped them from using AI tools they'd already bought. Not because the technology doesn't exist— it does—but because "integration-ready" doesn't mean what you think it means.

When a vendor says "integration-ready," they mean they have APIs. When you hear "integration-ready," you think it'll work with your existing systems. What actually happens: your IT team spends three months building custom connectors, or you pay the vendor $20,000 to $100,000 to do it for you. I've watched this exact scenario play out dozens of times.

A mid-sized retailer paid $8,000 for Workday integration, $12,000 for a custom career site connection, and $6,000 for LinkedIn Recruiter sync. That's $26,000 on top of their $60,000 license—43% more than the headline price—just to connect their existing systems.

A healthcare network paid $45,000 for SAP SuccessFactors integration, plus $8,000 per year in ongoing maintenance. The sales rep had called SAP integration "straightforward."

An enterprise financial services firm paid $120,000 for Oracle HCM integration with real-time bidirectional sync. The original quote for integration? $35,000.

The Tier Bait-and-Switch

Vendors love to demo the enterprise tier while quoting the professional tier.

During sales, you see the beautiful analytics dashboard, the advanced bias monitoring, the white-glove customer success manager. "All included!" the sales rep says. What they don't clarify: all included in the tier that costs twice what they quoted you.

The tier you can actually afford gets you a dashboard with five reports, no bias monitoring, and support tickets that go into a queue. By the time you realize this, you've signed a two-year contract.

The Compliance Time Bomb

This one is newer, and it's catching a lot of buyers off guard.

As of February 2025, the EU AI Act requires companies to eliminate "unacceptable" AI systems and comprehensively train all employees using AI. HR tools used for recruiting are classified as high-risk AI systems with strict compliance requirements.

NYC Local Law 144 requires annual bias audits. Illinois, Maryland, and other states have enacted or proposed similar regulations.

Many vendors are silent about compliance costs during sales. Why? Because compliance is boring, and boring doesn't close deals. But compliance isn't optional. Annual bias audits run $10,000-$50,000. GDPR data protection impact assessments run $10,000-$30,000. Documentation, consent management, and candidate rights procedures add another $15,000-$40,000 in the first year.

I've talked to buyers who had no idea these costs were coming until their legal team saw the contract.

What It Actually Costs

A CFO once told me that the hardest part of approving HR technology wasn't the first-year cost. It was knowing that the first-year cost was a lie.

"Every vendor shows me year-one numbers," he said. "Nobody shows me year four. And year four is when I find out whether this was a good investment or a slowly compounding mistake."

He was right. So let me give you what vendors won't: a realistic picture of costs across time.

The first three months are where your budget dies. This is implementation hell—integration overruns, data cleanup surprises, training costs that triple because your team needs more hand-holding than you expected. If you're going to go over budget, this is when it happens. Plan for it.

Years one through three are more predictable, but they compound in ways people forget. Your license will increase 5-10% at each renewal—it's in the contract, probably in small print. You'll need ongoing training for new hires. Your admin overhead never goes away. Compliance costs recur annually. None of this is surprising; it's just easy to ignore when you're excited about a new platform.

Years three to five are where the real questions emerge. Is the AI model still performing, or has it drifted? Is the vendor still innovating, or are you stuck on an aging platform? What would it cost to switch? These are questions nobody asks during procurement. But they're the questions that determine whether your total cost was 2x or 5x what you originally budgeted.

Now, let me show you what this looks like at different company sizes—because a $150,000 implementation at an enterprise means something very different than a $150,000 implementation at a startup.

Enterprise (1,000+ Employees)

First-year investment: $150,000-$500,000+

If the vendor quotes you $100,000 for an enterprise license, plan to spend $250,000-$400,000 in year one. That's not pessimism; that's pattern matching across dozens of implementations.

The breakdown typically looks like: Platform licensing at 30-40% of total. Implementation services at 20-25%. Integration development at 15-20%. Training and change management at 10-15%. Compliance at 10-15%.

Enterprise buyers have one advantage: negotiating power. Volume-based discounts can reduce licensing costs by 15-30%. Dedicated customer success resources are often included. Multi-year contracts can lock in pricing. Use that leverage.

Mid-Market (200-1,000 Employees)

First-year investment: $75,000-$200,000

Mid-market companies are in the worst position. They're too big for the startup-friendly pricing tiers and too small for enterprise negotiating power. They need enterprise-grade features but can't afford enterprise implementation support.

If this is you, my advice: be ruthlessly focused. Don't try to implement every feature. Pick one or two use cases that will deliver clear ROI, nail those, and expand later. The mid-market companies that succeed are the ones that resist the temptation to do everything at once.

SMB and Startup (Under 200 Employees)

First-year investment: $15,000-$75,000

Smaller organizations have options that didn't exist three years ago. Platforms like Talentprise, Breezy HR, and LinkedIn Recruiter Lite are designed for simpler tech stacks and smaller teams. Self-service implementation is actually feasible at this scale. Native integrations work because your systems are newer and more standardized.

The trap for SMBs: buying more than you need because the enterprise sales pitch made you feel inadequate. You don't need predictive workforce planning. You need to fill twelve roles this quarter without hiring another recruiter. Buy accordingly.

When More Expensive Means Better Value

Before the failure stories, a contradiction: sometimes the most expensive option is the best value.

I've seen $200,000 implementations deliver better ROI than $50,000 ones—not because the technology was four times better, but because the higher price forced organizational attention. When the CEO has to personally sign off on a $200,000 purchase, they pay attention to whether it's working. They ask questions in quarterly reviews. They hold people accountable. When it's buried in HR's discretionary budget at $50,000, nobody notices it failing until the contract renewal comes up and someone asks, "Wait, are we still paying for that?"

The cheapest path isn't always the safest. Sometimes the right move is to spend more money than you think you should—not because you need the features, but because you need the visibility.

Three Failures Worth Studying

Every case study you read in vendor marketing materials is a success story. Nobody publishes their failures. But failures teach more than successes, so let me tell you about three implementations that went wrong.

The Healthcare Network That Doubled Their Budget

This is actually a success story in disguise—but it didn't feel like one at the time. Healthcare AI implementations are particularly well-documented because of HIPAA compliance requirements.

The pattern: Regional healthcare network. 8,000 employees across twelve facilities. 400 annual hires, mostly nurses and medical technicians—roles where a bad hire doesn't just cost money, it can cost lives. Budget approved: $180,000, based on the vendor quote.

By month four: $220,000 spent. The vendor had quoted the professional tier; the team discovered mid-implementation that they needed enterprise-grade compliance features for HIPAA—$40,000 upgrade. Epic HR module integration required custom middleware that nobody mentioned during sales—$67,000. The consulting firm handling HIPAA documentation took three months longer than scoped—$28,000 in additional fees. And the 200 hiring managers who'd been using paper processes for fifteen years were actively hostile.

CFOs call meetings to discuss "project viability." Everyone in those rooms knows what that means.

The documented choice: kill the project and explain a $220,000 loss, or finish it and maybe explain a $350,000 success. Either way, you're explaining something. Organizations that push through tend to get results.

Eighteen months later in documented healthcare implementations: time-to-hire drops from 32 days to 10. First-120-day retention up 18%. Credential fraud incidents—a serious problem when you're hiring people who claim nursing licenses—nearly eliminated. ROI calculations come out to 150-200%.

But the first year ages people five years. The research doesn't recommend the experience—just finishing what you start.

The Fintech That Abandoned Ship

This pattern is well-documented in HR tech analyst reports, and the mistakes are avoidable.

Series C fintech. 350 employees, growing fast, flush with cash. They want to build "Google-caliber recruiting infrastructure" despite hiring maybe 80 people a year. Enterprise AI recruiting platform—$120,000 annual license. That's $1,500 per hire just for the platform.

Then: integrate everything. ATS, HRIS, background check provider, technical assessment platform, video interview tool. "We need a unified data layer." Total integration cost: $85,000.

Six months in: $230,000 spent and the system still isn't fully operational. The Greenhouse integration breaks every time Greenhouse pushes an update. The HackerRank connection drops candidate scores into the wrong fields. The three-person recruiting team—three people, for 80 annual hires—spends more time on Slack with vendor support than talking to candidates.

Leadership changes. The project gets killed. Back to manual processes and a simple ATS. Estimated waste: $300,000 including internal time. The $85,000 integration infrastructure runs zero data through it.

This pattern shows up in startup postmortems repeatedly. Enterprise-grade isn't a quality level. It's a scale assumption. If you're hiring 80 people a year, you don't need enterprise infrastructure. You need a spreadsheet that works and a recruiter who answers emails quickly. These companies buy a Ferrari to drive to the grocery store, then complain that the Ferrari requires too much maintenance.

The Retail Chain That Couldn't Change

This pattern is the most frustrating, because everything works exactly as designed. Except for the people. McKinsey's research on technology adoption failure documents this repeatedly.

National retail chain, 400+ locations, 25,000 employees. Every fall, they hire 5,000+ seasonal workers for the holiday rush. Corporate HR wants to systematize the chaos—too many bad hires, too much turnover, too many complaints from customers about employees who clearly don't want to be there.

AI recruiting platform specifically designed for high-volume hourly hiring. Implementation goes smoothly. Technology works. Pilot stores show a 23% improvement in seasonal hire quality scores. Everyone at headquarters is optimistic.

Then: rollout to 400 store managers who have been hiring their own seasonal workers the same way for twenty years. Post on Craigslist, interview whoever shows up, hire based on whether the person seems like "a good fit." These managers don't see a problem. Their stores are fine. Why does corporate keep sending new systems to learn?

The resistance isn't loud. Nobody sends angry emails or organizes protests. It's quieter than that, and more effective. Managers mark candidates as "rejected" in the system, then hire them anyway through the old paper process. They schedule interviews through the platform, then ask completely different questions than what the system records. They enter hiring decisions weeks after the fact, making the data useless for analysis.

Prosci's change management research documents this pattern: technology adoption fails when implementers don't involve frontline users. After two years, actual AI adoption in the field is under 30%. The ROI model assumed 80%. The project gets quietly discontinued.

This pattern stays with me. The technology works. The implementation is fine. But nobody at headquarters asks the store managers what they need, or explains why the old way isn't working, or gives them any reason to care. They just send a new system and expect compliance. And they get exactly what that approach deserves.

What I'd Do With Your Budget

The HR leaders who succeed the second time around share common patterns, documented in both industry research and practitioner communities.

The approach that works: spend the first month talking to people who've been through implementations—good and bad. Make your CFO sit through vendor demos so they can't later claim they didn't understand what you were buying. Hire a contract lawyer to review pricing terms before signing anything. It takes twice as long to get started. But you come in under budget.

Here's what I've learned from studying successful implementations:

If You Have Less Than $25,000

Pick one thing. Just one. The number-one mistake at this budget level is trying to do too much.

If your biggest pain point is sourcing, buy a sourcing tool. If it's screening, buy a screening tool. If it's scheduling, buy a scheduling tool. Get really good at that one thing before adding complexity.

Self-service implementation only. If a vendor tells you that you need professional services at this budget level, they're selling you the wrong product.

Native integrations only. If the integration requires custom development, skip it. Accept some manual data entry rather than blowing your budget on connectivity.

If You Have $25,000-$100,000

You can afford a real platform, but you can't afford to screw it up.

Budget 25-30% for implementation, not the 15% vendors suggest. You will hit unexpected issues. Build in buffer.

Two integrations maximum in year one. Your ATS and one other system. That's it. More integrations can wait until the core platform is stable.

Set aside $10,000-$15,000 for change management. This isn't optional. If your team doesn't adopt the system, the system is worthless.

Negotiate hard on overages. At this budget level, a $20,000 overage charge could blow up your entire business case. Get caps in writing.

If You Have $100,000-$300,000

You're in enterprise territory. Act like it.

Get executive sponsorship before you sign anything. Implementations at this scale require organizational commitment, not just HR commitment. If your CEO or CHRO isn't visibly behind this, you're setting yourself up for the retail chain's failure mode.

Budget 15-20% specifically for compliance. At this investment level, you're probably in scope for EU AI Act, NYC LL144, and other regulations. Don't treat compliance as an afterthought.

Negotiate multi-year pricing with exit clauses. Vendors love multi-year deals and will give significant discounts. But insist on data portability guarantees and reasonable termination provisions. You don't want to be locked into a bad relationship.

Consider building internal expertise rather than relying on vendor support. Organizations with AI Centers of Excellence reduce project delivery time by 40-60%. At this investment level, hiring a dedicated AI recruiting operations person might make more sense than paying premium support fees forever.

If You Have More Than $300,000

At this level, you have options most organizations don't.

Best-of-breed might make sense. Instead of one platform that does everything adequately, you might be better served by specialized tools that each do one thing exceptionally well, connected through a custom integration layer.

Custom AI model training becomes feasible. Off-the-shelf models work, but models trained on your specific hiring data work better. The economics only make sense at scale, but at this scale, they might.

Build an AI recruiting operations team. One or two people dedicated to optimization, troubleshooting, and continuous improvement. The investment pays for itself through better utilization and faster issue resolution.

How to Not Get Played

The HR leaders who come in on budget share a common approach, documented across industry research and practitioner communities: assume everyone is lying. Not maliciously—just optimistically.

Build your whole process around verification. If someone tells you something will cost X, find three other people who've bought from that vendor and ask what they actually paid. If someone tells you integration will take two weeks, assume six. It's not cynicism. It's just... careful.

That's the right approach. Let me break it down.

Before You Talk to Any Vendor

Know your numbers first. How long does it take you to fill a role, by role type? What does each hire actually cost, including everyone's time? What's your 90-day turnover? I've watched buyers walk into vendor meetings without being able to answer these questions, and it never ends well. You can't evaluate whether an investment is worthwhile if you don't know what "worthwhile" means for your organization.

Map your tech stack before anyone tries to sell you on "seamless integration." Every system that touches recruiting. Every integration point. Every data flow. When vendors tell you their platform connects easily with Workday or Greenhouse, you need to know whether that's true for your specific environment— not the idealized environment in their demo.

Talk to IT and Legal before you fall in love with a platform. Not after— before. I've seen deals die in legal review because nobody thought to ask about GDPR implications until month three of the sales process. IT can tell you which integrations are realistic. Legal can tell you which compliance requirements apply. Let their input shape your requirements, not just react to your selection.

During the Sales Process

Demand itemized pricing. This is non-negotiable. When a vendor gives you a bundled number—"$75,000 all-in"—that's where surprises hide. Make them break it down: license, implementation, integrations, training, support, compliance, data migration. Every line item. If they push back, if they say "we don't typically do that," that tells you something important about what they're hiding.

Ask them, directly, about typical overruns. Here's how I'd phrase it: "In your last ten implementations, how many came in over the initial estimate, and by how much?" A vendor who says "our implementations always come in on budget" is either lying or has never actually looked. A vendor who says "honestly, about 60% run 15-20% over, usually because of data cleanup or integration complexity"—that's someone who's done this enough times to know the truth. I'd rather work with someone who's honest about the mess than someone who pretends it doesn't exist.

Talk to customers who left. Every vendor will give you references. Of course the references are happy—that's why they're references. But the customers who churned know something the happy customers don't. Ask the vendor: "Can you connect me with a customer who decided not to renew?" Watch their reaction. If they refuse, find former customers yourself on LinkedIn. A quick search for the vendor name plus "former customer" or "moved away from" usually turns up someone willing to talk.

Get everything in writing. Everything. If a sales rep tells you something is included—"oh, that's part of the package"—ask them to show you exactly where in the contract it says that. If it's not in the contract, it doesn't exist. I've watched buyers get burned because they trusted verbal assurances that vanished the moment the signature dried.

Negotiation Points That Matter

These are the clauses that will save you—or cost you—tens of thousands of dollars. Don't skip this section.

Overage caps. Remember Sarah's $40,000 overage surprise? That happens when you sign a contract with unlimited overage exposure. Don't. Negotiate maximum overage rates. Better yet, negotiate automatic tier upgrades— if you hit 80% of your allocation, you move to the next tier at a predetermined price rather than getting hit with per-unit charges.

Fixed-price implementation. Time-and-materials implementation is how vendors make their real money. Every week of delay becomes profit for them. Push for fixed-price with defined deliverables. If the vendor refuses— if they say "we can't guarantee timelines because every implementation is different"—that's them telling you they're not confident in their own estimates. Believe them, and price accordingly.

Renewal terms with escape hatches. Vendors love multi-year deals and will give you 15-25% discounts for signing one. Take the discount, but insist on off-ramps. A clause that lets you exit after year one with 90 days notice if certain performance metrics aren't met. A price cap on renewal increases. Something. You don't want to be stuck paying year-three prices for a product that stopped working in month eight.

Data portability. This is the one everyone forgets until it's too late. If the relationship ends, what happens to your data? Can you export it? In what format? At what cost? I've watched organizations stay with terrible vendors for years because switching would mean losing five years of candidate history. Get the export terms in writing before you sign.

The Real Advice

The wisdom from HR leaders who've been through this process clusters around a few key insights, documented in practitioner communities and industry research.

First: assume the vendor is lying. Not maliciously—most of them believe what they're saying. But the incentives are misaligned. They get paid when you sign. You get value when the implementation works. Those aren't the same thing.

Second: don't be embarrassed about asking questions that feel dumb. Too many buyers are so worried about looking unsophisticated that they don't ask basic things like "what does this cost include?" and "what happens if we go over our allocation?" Performing competence instead of actually being competent.

Third: the invoice is coming. Not might come—is coming. The hidden costs aren't hidden because vendors are evil. They're hidden because nobody—not the vendor, not the buyer—wants to talk about the complicated, unglamorous work that makes technology actually function. So budget for it. Build it into your business case from the start. And when the invoice arrives, you won't have to explain to your CFO why you look like you don't know what you're doing.

AI recruiting technology works. I've seen it work. Organizations that implement thoughtfully see returns of 150-400% within 18-24 months. Time-to-hire drops. Quality improves. Recruiters do work that matters instead of work that exhausts them. But "thoughtfully" means understanding what you're actually buying—not what the sales deck shows, but what the invoice will demand. Most organizations are already investing in AI recruiting. The only question is whether you go in knowing what you're actually buying—or whether you find out later, staring at an invoice that makes you wonder how you missed so much.