Building a procurement process that people actually follow: Justin Etkin of Tropic

Most attempts at a procurement process fail in one of two directions, and Justin Etkin can name both. Either it is too complex too quickly and the organization loses faith in it, or leadership never enforces it. The work is landing between them.

Etkin is COO of Tropic, which is rebuilding how businesses buy and manage software. He describes his own entry into procurement as learning by doing, making mistakes, and having bad experiences at companies that did not do it well.

Why procurement changed

Etkin's history is short and worth having.

Procurement began as cost reduction on large commodity inputs. Saving money on jet fuel and the major spend items that go into what a company makes, in service of protecting margin.

What has happened since is a broadening of scope rather than a change of purpose. Procurement professionals became generalists, not only across types of spend but across responsibilities: process compliance, vendor risk management, environmental and governance reporting.

Which turned it into an operationally intensive function rather than a sourcing and cost-containment one. And that, he argues, is why every business should now be thinking about it, because it is among the most important drivers of efficiency at any scale.

When it becomes a real function

The evolution is predictable, and it usually begins with an incident.

Someone in finance or operations is handed a project because the CEO or CFO believes the company is spending too much, or because there was a catalytic event. Etkin's example is getting caught on an auto-renewal and spending hundreds of thousands of dollars nobody wanted to spend.

Michael's version from his own experience is the runaway usage-based bill: somebody commits code, the system makes two million API calls, and a monster invoice arrives. Which is not only a code problem.

In the technology companies Tropic works with, the question of a dedicated function usually arrives around a couple of hundred people. And it resolves the way every headcount question does: whether the cost of a full-time person is justified by the value they generate, which typically means once there are millions of dollars under management.

Michael's own origin story is instructive. His first guest on this podcast, on his move into his first COO role a decade ago, told him to go look at his software, because he had too many licenses and was over-provisioned, and could save hundreds of thousands in his first month through an audit. Back then, that audit meant going into every system by hand.

The three things you need visibility on

Etkin's core argument is that most companies approach procurement from the outcome they want rather than from the foundation.

The quick-win instinct is savings: negotiate the contracts, eliminate duplication. He does not dismiss it, and notes procurement has to show impact quickly to justify the investment in it. But those engagements are discrete by nature, and sustainable practice requires something broader.

Good procurement, in his framing, rests on three sources of visibility, and the value comes from having all three in one place.

What you expect to spend. Contracts centralized in a system, so you know what you have committed to and with whom.

What you are actually spending. Pulled from the accounting system or ERP, on an as-expensed basis.

How the tools are actually used. Who holds licenses and how they interact with them, which matters particularly for technology companies.

His point about why one system beats three: each of these exists as its own product category, contract management, software management, accounting. Having them normalized on a like-for-like basis across supplier relationships is what lets you compare what you expected to do with a vendor against what actually happened.

Then controls, then impact

Step two is process governance. Once you know where the spend is going, everyone needs to understand the expectations.

He is careful to keep this proportionate. For an early-stage company, process can mean something as simple as the CFO approving anything above a threshold. The specific choice is cultural. What matters is that somebody knows what spending decisions are being made, so governance can exist at all.

Step three is where the impact lands, and only then. With visibility and controls in place, you can layer intelligence on top.

That is what Tropic built: a networked database of pricing and benchmarks, so finance and procurement people know what they should be paying and how to get there.

Leveling a genuinely unequal field

The argument Etkin makes for why this data matters is the most persuasive thing in the episode.

The story of procurement is a team that is under-resourced and under-equipped relative to its counterparties. Sales teams arrive armed with enormous investment in data and tools, and every intent signal about a buyer that a good seller can measure. Procurement teams have a fraction of that.

So the goal is levelling the field. Give buyers a networked understanding of what a fair price looks like and they at least have a fighting chance.

But he immediately qualifies it, and the qualification is what separates this from a data pitch. You can have the best data in the world and it is worthless if you cannot deploy it at the right moment with the right strategy. You cannot walk up to a vendor, announce you know what you should be paying, and expect anything but to be told to go away.

The value is in orchestrating that data into a story with credibility and staying power that can actually justify a conversation.

What a renewal looks like with the data

His worked example makes it concrete.

A contract comes up for renewal in four months, executed through a reseller, with a 30-day opt-out. Which means the real decision window is three months, and that is the period in which you have any leverage.

Going in, you know your spend trajectory over the past year. Seats added or removed, upgrades, downgrades, package changes. And you know how your pricing compares to the market.

The half people forget is what to do when the answer is good. If your pricing is already fair, pass it through. Do not spend cycles banging your head against a supplier relationship that is already in an acceptable place.

If the pricing is bad, you run a real sourcing motion. In the reseller case, the relevant knowledge is that many resellers exist, all operating on thin margins, and moving business between them is a powerful lever on price. So you bid it out to resellers with strong relationships in that product, and use the competitive tension either to stay where you are on better terms or to move.

Right-sizing the process to the company

The barbell failure is Etkin's most useful observation for anyone standing this up.

Companies either go too hard and too complex too fast, losing organizational buy-in, or leadership never provides enough top-down enforcement to signal that this matters. The goal is landing between them, sized to what the organization is actually trying to achieve.

His two contrasting cases make the point.

A company introducing procurement for future efficiency, where savings are not today's goal, needs something everyone will actually adopt. That means being unaggressive about approvals, letting stakeholders own their own relationships and tools while arming them with data, and resisting the urge to centralize the process into finance. Keep it simple, get a system in place.

A company that has just been taken private, where cutting cost is the priority, needs the opposite. More visibility from leadership, specific targets, and genuine top-down enforcement.

And the truth underneath both, which he states plainly: nobody in the organization enjoys the procurement process. It is a necessary discipline. If you are focused on growth and closing deals, the last thing you want is hoops to jump through before you can renew a supplier you depend on. Which is exactly why the design has to be deliberate and the expectations communicated.

What changed after the correction

Etkin's read on the market is that 2021 and 2022 produced growth in every direction, which left substantial consolidation opportunity and redundancy, and that 2023 and 2024 were the belt-tightening response.

What he hears now is different again. Companies want efficient spending they can rely on and sustain. Sometimes that means cost cutting and eliminating redundancy. Mostly it means being set up so the earlier mistakes do not repeat.

His phrase for what everyone is carrying: scar tissue from the period when everything was green-lit and costs then went out of control.

The structural consequence is a push from point solutions toward platforms. Companies are willing to sacrifice some capability in a strong point solution because it does not talk to the other point solution their function also depends on, which creates friction and waste.

In procurement specifically, that has meant absorption. The software management category is being folded into procurement. Contract management is being pulled in. So is vendor risk management. Categories that each saw their own proliferation are consolidating into single platforms.

Where AI actually helps

Etkin points at the same opportunity everyone identifies in accounting: large teams doing manual work.

In procurement that means creating purchase orders, matching invoices to them, extracting metadata from contracts, and populating one system with data from another. Pervasive at scale, and exactly the kind of work that can be automated.

His framing of the payoff connects to the theme: remove the overhead required to make procurement work, and redeploy those dollars into growth. Which is how procurement becomes a growth lever rather than a cost centre.

Why contracts are harder than invoices

The technical explanation here is genuinely interesting.

On the scale of structured documents, a contract is surprisingly unstructured, because the format varies enormously between suppliers. Invoices and receipts are highly structured, follow the same pattern, and have been automatable for years.

So Tropic uses language models to extract contract metadata. But Etkin is clear the extraction is not the hard part.

The hard part is normalization. Working out the unit of measurement that is comparable across every Salesforce contract, then across every Snowflake contract, then building a data model consistent enough to compare like for like across thousands of vendors with wildly different pricing structures.

He describes it as years and thousands of hours of investment, and it is what makes the benchmark database possible.

The fat head and the long tail

Michael raises the question everyone wonders about: will a procurement agent eventually negotiate with the seller's agent?

Etkin's answer uses the Pareto distribution, and it is the sharpest framework in the conversation.

Roughly 20% of vendors account for 80% of spend. That fat head is where humans will keep playing a role, because large complex purchases involve genuine customization between what a buyer needs and what a seller can offer, and because people buy from people.

His counter-example, offered before Michael can raise it, is car buying, which used to be entirely human and has been substantially automated. His explanation for why that is different: cars are standardized commodities, and removing the human is far easier there than in complex technology purchases.

The other 80% of vendors, making up 20% of spend, is where automation makes sense. The result of each individual deal matters less, so procurement teams may reasonably sacrifice the best possible outcome for efficiency rather than deploying expensive human attention.

His summary: fat head human-driven, long tail automated. And he expects we will all eventually see clips of AI sales representatives negotiating with AI procurement agents.

The sidekick, not the negotiator

For the near term, Etkin's view of the useful application is a coach riding shotgun rather than an agent acting alone.

An assistant that accounts for everything relevant to a specific supplier engagement: timing, historical context, how long the relationship has run, how the tool has been used, sentiment. All of it feeding a strategy for how to approach that particular negotiation.

Give that to a procurement professional and they become substantially more effective, which matters most precisely where the money is, in the fat head.

The sentiment gap

The sentiment point produces the episode's clearest illustration of the imbalance Etkin keeps returning to.

Call recording software is ancient history on the sales side. Every sales conversation at a technology-enabled company is recorded, and multiple products apply sentiment analysis to judge where a conversation is heading.

Procurement teams get none of that investment. No analysis of what the sentiment in their calls implies about a counterparty's willingness to discount or move.

And the structure of the two organizations is identical: multiple managers covering individual categories, with varying experience, needing varying coaching, under managers doing pattern matching. Tropic's own head of procurement strategy, who came from Qualtrics, ran war games with his team to simulate negotiations, because recorded-call coaching simply did not exist.

Etkin's prediction is that this changes within a couple of years, because the investment community has recognized procurement's importance.

Why the innovation is slower

Michael's observation is the honest commercial one: he will write a check all day for something that increases sales 10%, and it is much harder to justify something that saves 10%.

Etkin agrees, and identifies the constraint as budget rather than technology.

A go-to-market team will test an unproven product for $50,000 a year if there is a chance of a threefold return, because that contributes to growth. Procurement teams have no equivalent experimental budget.

Which means new capability reaches them differently: as a module inside a platform they already run. It is far easier to tell a CFO that your existing procurement platform has released a new capability that integrates with everything you already have than to advocate for an unproven startup. He notes that conversation was not happening in the CFO's office until a year or two ago, and is now starting to.

Orchestration is the real problem

Michael's closing question is about the number of moving pieces: contract management, license management, finance, legal, security.

Etkin's answer names the actual constraint at a growing company. Legal and security often mature faster than procurement, particularly as cyber risk has driven those hires earlier and earlier. So by the time you make your first procurement hire, adjacent functions already have their own tools.

Which means the new procurement person has to fit into an existing web of systems that are not going anywhere. Security will not give up the tool that makes them effective. Neither will legal.

Hence orchestration as the primary capability of a modern procurement platform: serving as a source of truth while pushing information out to and pulling it in from systems that belong to other functions.

At the largest scale, companies move to full source-to-pay suites from the long-established vendors, which bring enterprise credibility and cover more of the requirement from one place, at the cost of the flexibility earlier-stage products offer.

Painting the walls

Etkin's crazy story is from that same day.

Tropic was moving out of its office, and as he looked around wondering what the move-out would involve, he realized it fell to him. So he spent the day sitting in the office painting the walls white before handing the space back to the landlord.

Michael's contribution is a white stripe on a favorite fleece, from an office move-out in 2017. You end up sweeping the floors and painting the walls. It is one of those things.

The 5 things I took away from this conversation

1. Visibility, then controls, then impact, in that order. Everyone starts at impact because savings are what justify the investment. Justin's sequence is the correction: you cannot negotiate well against data you do not have, and you cannot hold a process nobody agreed to.

2. The three data sources have to sit together. What you committed to, what you actually spent, and how the tool is actually used. Each exists as its own product category, and the value is in comparing them on a like-for-like basis. That comparison is what tells you whether the contract matches reality.

3. Right-size the process to what you are actually trying to achieve. A company optimizing for future efficiency needs something light enough to be adopted. A company under a cost mandate needs enforcement. Getting this backwards is how procurement processes get quietly ignored.

4. Good pricing is also an answer. The point of the benchmark is not always to negotiate. It is sometimes to confirm you are in a fair place and move on, which is a genuine saving of the scarcest resource procurement has, which is attention.

5. Automate the long tail, keep humans on the fat head. Twenty percent of vendors, eighty percent of spend. That is where judgment earns its cost. The rest is where you accept a slightly worse outcome in exchange for not spending a person on it.

FAQ

What is a procurement process? The set of steps and controls governing how an organization buys and manages what it buys. Etkin's framework has three layers: centralized visibility over contracts, actual spend and usage; process governance defining who approves what; and only then the sourcing and negotiation work that produces savings.

When should a company hire a dedicated procurement person? Usually around a couple of hundred people at technology companies, and specifically when the cost of managing spend outgrows what someone can do alongside their main role. The test is the same as any headcount decision: whether a full-time person returns more than they cost.

Why do procurement processes fail? Etkin describes a barbell. Either the process is too complex too quickly and the organization stops using it, or leadership never enforces it and it has no authority. The design has to match both the culture and the actual goal, whether that is adoption or cost reduction.

How do you negotiate a software renewal? Start well before the opt-out date, since that is the window where you hold leverage. Know your usage trajectory and how your pricing compares to the market. If your pricing is already fair, pass it through. If it is not, create competitive tension, including between resellers where the product is sold through channel.

Where does AI actually help in procurement? Extracting contract metadata, matching invoices to purchase orders, and the administrative work that consumes large procurement teams. Etkin's near-term view of the strategic use is an assistant that briefs a human on a specific supplier negotiation rather than one that conducts the negotiation.

Also mentioned

• Tropic, its pricing benchmark database and procurement platform
• Resellers and value-added resellers, and using competition between them as a pricing lever
• Coupa and Ariba, the established source-to-pay suites at the largest scale
• Qualtrics, where Tropic's head of procurement strategy ran negotiation war games
• The Pareto distribution, and why it decides what gets automated

Listen to the full episode

Justin Etkin on Between Two COO's

Between Two COO's is hosted by Michael Koenig. Subscribe on Apple Podcasts, Spotify, or wherever you listen.