Podcast Summary 5 min read

Without benchmarks you're naked — Michiel de Bruijn on greenfield SaaS procurement

Building the function from scratch at a remote-first SaaS company, why cloud spend is utilisation not negotiation, and how to procure AI when there's nothing to compare it to.

Michiel de Bruijn is the sole procurement lead at Bloomreach, a 1,000-person remote-first SaaS company personalising e-commerce for 1,400 global brands. He cut his team size from 25 to one to take the role — the pull was greenfield. Most procurement careers train you to operate inside legacy systems and inherited policies; building the function from scratch at a modern SaaS company is its own discipline. The conversation is the third procurement episode on SoF, and it sits at the smallest, fastest, most-SaaS end of the spectrum after the enterprise-procurement masterclasses from Reba Cox at MongoDB and Murali Sundararajan at Wipro.

The spine is the operational reality of procurement at a remote-first SaaS company: a different spend base (negligible real estate, heavy cloud and SaaS, long tail of digital tools), a different methodology (infrastructure → policies → spend visibility → high-impact early negotiations → long tail), and a different set of hard problems (the apples-to-apples reconstruction problem, cloud as utilisation rather than negotiation, AI as the new no-benchmark frontier). The universal line — without benchmarks you’re standing in the street naked — is the operational thread holding all of it together.

Greenfield procurement, in order

The most useful operational frame in the episode is the build sequence.

First you build infrastructure, you get basic policies in place, then you're trying to understand the spend, and then you start looking at negotiations.

The order matters. Skip the infrastructure and you have no system of record; skip the policies and the conditional approval routing collapses; skip the spend mapping and you negotiate the wrong contracts. The non-obvious step is the cherry-pick — the early high-impact negotiations Michiel ran in his first months at Bloomreach were the function’s proof-of-value, and the credibility he banked then is what made the long-tail compliance work possible later. Most greenfield procurement attempts skip this step and go straight to PO hygiene; that’s correct work, but it doesn’t earn the internal voice the function needs. The team structure that followed is deliberately distributed — three people across US, Czech Republic, and Dubai — so every internal stakeholder can find a procurement partner in their time zone and ideally their language. At a 1,000-person remote-first company the team design is a cultural-fit decision, not a coverage decision.

Cloud is utilisation, not negotiation

The reframe that distinguishes modern-SaaS procurement from its enterprise predecessor.

Cloud spend is way more about utilization, and optimizing utilization and infrastructure, versus the actual negotiation.

Cloud contracts negotiate in small percentage points — moving a spend band, hitting a commitment threshold, an extra 1% discount on a renewal. The real money sits one layer underneath: right-sizing instances, killing idle workloads, refactoring architectures over-provisioned for current demand. The lever is 5-30%, not 1%. Procurement teams that obsess over the cloud-vendor terms and ignore the engineering layer underneath leave the bigger number on the table. Michiel’s view is that the procurement lead should take the free vendor training that engineers take — AWS, GCP, Azure all offer it — because the negotiation conversation is meaningless without the utilisation conversation underneath. This is the FinOps end of procurement, and at a SaaS company of this shape it looks more like engineering partnership than traditional category management. The same pattern shows up across the software stack: SaaS-rationalisation, licence-bloat tracking, and the question of whether a $50/month tool genuinely earns its place in a 2,000-supplier ecosystem.

The apples-to-apples reconstruction problem

The universal procurement problem at modern SaaS scale.

Without benchmarks, you're just kind of standing out in the street naked.

SaaS vendors structure their pricing — per-seat with feature tiers, per-usage metered, per-bundle with overages — to make direct comparison hard. Wow, you can’t compare us is the salesperson’s most common line. The procurement lead’s job is to reconstruct comparability: normalise the pricing units, source external benchmarks, put apples-to-apples on the table before negotiation starts. Michiel benchmarks any software contract above $2K, through aggregator platforms, procurement networks, peer conversations, and former colleagues. The structural trend Michiel flags as the future is the independent SaaS marketplace — aggregator platforms publishing standardised pricing, standardised contract terms, monthly billing instead of annual upfront, SOC2/GDPR compliance baked in. If that pattern matures, the apples-to-apples problem gets solved at the marketplace layer rather than the procurement-team layer, and the function moves further upstream toward strategy.

Procuring AI without a reference price

The new frontier where the playbook hasn’t been written yet.

Ironically, every crisis for procurement is a good one — because suddenly procurement gets a voice.

AI is the current crisis. No reference pricing, no industry-standard contract structure, often no SOC2 maturity, and pitches that all converge to ROI from AI without the back-up. Bloomreach’s intake process now explicitly asks if a software purchase has an AI component; a yes triggers privacy and legal review regardless of contract size, because the $50/month spelling tool that suddenly trains on customer data becomes a compliance risk overnight at a data business. The pitches that land with Michiel pre-empt the questions — we don’t train on your data, we keep it on a private instance, here’s a customer of your shape and geography with this measured outcome — and the pitches that get filtered are the generic ROI decks that don’t acknowledge the buyer’s use case. The compliance-vs-window trade-off is real: Michiel will occasionally accept a temporary risk on an early-stage AI tool that doesn’t yet have SOC2, by documenting the business risk acceptance and tracking it for closure, because the alternative is missing the window on a tool that genuinely moves the business. The discipline is procurement at its most cross-functional and least price-focused — exactly the version of the job Michiel argues is undersold.

What to listen for

The full episode is the practitioner’s view of procurement at modern SaaS scale: the $4-7.5M recurring-SaaS spend rule of thumb for a 1,000-person SaaS company, the 65/35 renewal-to-new vendor split, the math of moving from a one-year to a two-year contract, and the closing thesis that procurement is still such an undersold profession. His three-word descriptor is Honest. Straight. Over thinker. Listen at /podcast/ep-032-michiel-de-bruijn; for the other procurement conversations, see Reba Cox and Murali Sundararajan, or /topics/procurement-spend.

Related questions

What is Michiel's methodology for building procurement from scratch at a SaaS company?
A four-step sequence, executed in order, that he ran at Bloomreach when he joined as the sole procurement hire. First, build the infrastructure — the tools, the intake process, the conditional approval routing, the system of record. Second, set the policies — what gets reviewed by legal, by privacy, by finance, at what thresholds. Third, understand the spend — map the categories, the contracts, the renewal dates, the vendors that nobody can name. Fourth, cherry-pick the high-impact early negotiations to prove the function's value to the rest of the business; only then move down the long tail to the smaller contracts. The trap most greenfield procurement leaders fall into is skipping the early-impact step and going straight to compliance hygiene — which is correct work, but doesn't earn the internal credibility the function needs to do everything else. Bring money in first, then build the system around it.
Why does Michiel call cloud spend a utilisation problem and not a negotiation problem?
Cloud contracts at scale (AWS, GCP, Azure) negotiate in small percentage points — moving the spend band, hitting a commitment threshold, picking up an extra 1% discount. Real money is in the engineering layer underneath the contract: right-sizing instances, killing idle workloads, refactoring legacy architectures that are over-provisioned for current demand. Procurement teams that obsess over the cloud-vendor discount terms miss the 5-30% utilisation lever sitting next to it. Michiel's view is that good cloud procurement requires the procurement lead to take the free vendor training that engineers take — to understand the dashboards, the metering, the architecture — because the negotiation conversation is meaningless without the utilisation conversation underneath it. This is the FinOps end of procurement and it looks more like engineering partnership than traditional category management.
What is the 'apples-to-apples reconstruction problem' in software procurement?
SaaS vendors deliberately structure their pricing so direct comparison between two products is hard. Per-seat with feature tiers, per-usage metered, per-bundle with optional add-ons, with quotas and overages — every vendor's price list is shaped to make like-for-like comparison difficult. The procurement lead's job is to reconstruct comparability: normalise the pricing units, benchmark against external data sources, and put real apples-to-apples on the table before negotiation starts. Michiel's universal line is that without benchmarks you're standing in the street naked — you cannot anchor a negotiation, you cannot prove to the internal business that the price is fair, you cannot judge whether the vendor is being transparent or playing you. Community benchmarks (peer sharing, aggregator platforms, network conversations) are the next frontier, and the trend toward independent SaaS marketplaces that publish standardised pricing is the structural response to the comparability problem.
How does Michiel handle procuring AI applications when there are no benchmarks?
Three operational moves. First, the privacy and compliance review gets pulled forward — Bloomreach's intake process now explicitly asks if a software purchase has an AI component, and a yes answer triggers privacy/legal review regardless of contract size, because a $50 spelling tool that suddenly trains on customer data becomes a compliance risk. Second, the ROI pitch has to be specific to the buyer's actual use case — generic 'ROI from AI' decks get filtered out, and vendors who pre-emptively name the data-handling posture (no training on customer data, private instance, etc.) and tailor the ROI to the customer's geography and business model land better. Third, Michiel accepts that in genuinely novel AI categories he may have to take a temporary compliance risk on an early-stage tool that doesn't yet have SOC2 — by documenting the business risk acceptance, vetting the gaps with additional questions, and tracking the risk for closure once the vendor matures. The alternative is missing the window on a tool that genuinely moves the business.

Updates

  1. Editorial pass under the v2 podcast-summary guideline.