Back to Hub

McKinsey buys Orpheus: Valuation Estimates, Product Strengths/Weaknesses and Implications for the Firm & Its Clients

02/13/2020 By


This month, McKinsey announced it was acquiring Orpheus, a regional yet specialized (and highly capable) spend and procurement analysis provider based in Germany. For a primer on Orpheus and the acquisition, see our initial Nexus coverage. Long-time subscribers to Spend Matters PRO and SolutionMap Insider (analytics) know that analytics represents a set of capabilities that we pay very close attention to, owing to its importance in identifying broad-based procurement savings opportunities — and tracking results.

Indeed, when implemented correctly, spend and procurement analytics can drive savings and compliance across a range of areas, including sourcing, category management, procure-to-pay, contract compliance (including commodity and currency clauses). These solutions can also help procurement and finance organizations and consultancies to identify and manage working capital and payment opportunities to create a new balance sheet lever for “spend” that extends beyond how most companies think of spend management — a clever artifice that consultancies have leveraged in cost take-out engagements for decades.

This Spend Matters Nexus brief begins by sharing a back-of-the-napkin valuation and multiple estimate for the Orpheus-McKinsey transaction based on our M&A work in the sector (valuing spend analytics firms is typically not as easy as traditional SaaS). But this analysis focuses primarily on highlighting Orpheus’ strengths and weaknesses and what these will mean as McKinsey takes Orpheus global as part of its “Solutions” arsenal of capabilities outside of just consultancy. Not to mention enhancing the value levers — and speed with which it can pull them — in client studies.

As our analysis concludes in the final brief in this series, we will provide insight into how the transaction may impact the competitive landscape for spend and procurement analytics.

Read our past coverage:

Jason Busch is Managing Partner of Azul Partners’ Investor Advisory Group. He works with sponsors, CEOs and boards on data-driven due diligence, M&A and strategy. Jason is also the lead author of Spend Matters Nexus, a private newsletter and subscription service that publishes 50+ times per year. Spend Matters and Spend Matters Nexus are owned by Azul Partners. His investment disclosures and other activities can be found on LinkedIn.

Back-of-the-Napkin Valuation and Multiple Estimate

  • Let’s assume Orpheus had 5 million to 6 million euro in revenue in 2020 (these numbers fall within the range of those provided to Spend Matters by Orpheus in 2019 prior to the acquisition).
  • Spend analytics provider revenue usually contains a larger component of services revenue than comparative SaaS providers (e.g., sourcing, contract management) owing to the initial and ongoing data loads and refreshes. But AI is helping to bring this component “down” and increasing gross margins in certain cases:
    • For the sake of argument, let’s say a reasonable breakout for a spend analytics provider between SaaS and services is 65/35 or 70/30 compared with an 80/20 or even 90/10 split for other similarly sized firms competing in different procurement technology modular areas.
    • Given this, and given Orpheus’ data science prowess, let’s assume the splits were more favorable to software (70/30).
  • SaaS multiples in procurement are largely dictated by revenue growth, especially among smaller companies (for the component of the valuation that is based on software, not services).
    • Companies that are doubling each year (or realizing greater growth) can achieve a transaction multiple of 20X trailing topline or more (e.g., Scout RFP’s multiple in the Workday acquisition).
    • But for SaaS companies with more modest growth rates in constrained areas like spend analysis (i.e., where there are few independent competitors with strong technology), a top-line multiple of 6-8 on SaaS revenue is possible.
  • European companies usually trade for a discount compared with U.S. firms.
  • Given these considerations, let’s assume Orpheus SaaS revenue was valued, conservatively, at 5-7X and its services revenue was valued at 2X (note, I could have gone higher in my models on the SaaS component given other comparable deals in the market, in cases of vendors with smaller revenue, but we’ll stick at this).
  • This gives us a valuation range of 24-30 million euro, assuming trailing revenue of 5.5 million.

Of course in smaller transactions there are many other considerations as well, including upfront vs. earn-out considerations that can factor into multiples. But to keep things simple, let’s stick to this range, as it seems reasonable after pressure-testing the logic*. And if we eventually learn from an expert source the exact numbers and there’s no NDA, we’ll update the estimate.

Now let’s turn our attention to Orpheus’ strength and weaknesses — and if we were in McKinsey’s shoes, what these would mean to us. We base these on Spend Matters’ latest PRO series brief on Orpheus (see above) with additional implications and takeaways.

* We base the back-of-the-napkin valuation on similar exercises we have done for strategic and financial buyers in recent quarters.

Orpheus Solution Strengths

Truly Multilingual AI Classification

Besides supporting the definition of very complex mapping rules that can use reg-ex and formula across multiple dimensions and related data elements to enrich and classify data, the platform also supports a number of AI techniques based on unsupervised machine learning and supervised learning algorithms that can improve over time, especially as the user confirms classification and creates additional rules. The platform can employ standard clustering, regression, decision trees and artificial neural networks in its classification.

The platform is truly multilingual and can classify not only a host of Romance languages, but also Cyrillic languages. As long as there is enough classified data to identify key suppliers, organizational units, products and groupings to feed the classifiers, the platform can automatically classify multilingual data and improve over time.

McKinsey implication: For McKinsey’s global clients, Orpheus should scale more effectively, especially on a global basis, than more traditional rules-based spend analysis data cleansing and classification capabilities. In addition, for spend analysis, AI also greatly accelerates the rate at which a solution can classify 80-90% of data (or more) accurately, assuming a properly “trained” data set. This should allow McKinsey to rapidly deliver broad-based spend, procurement and finance analytics capabilities in the case of client studies.

In other words, just as McKinsey is expert at having on-site consultants mock-up slides and findings and then having a low-cost follow-the-clock operation turn them into the Kraljic-inspired, MECE data, charts and takeaways in turns measured in hours rather than days, it will eventually be able to guarantee engagement teams similar overnight (or 48-72 hour) turns of data, cutting down the time to identify savings, working capital and other opportunities both in traditional situations.

This approach may even help drive new revenue opportunities for McKinsey (e.g., M&A due diligence to rapidly identify efficiencies that can help PE firms and strategic buyer clients to sharpen their pencils and/or become more comfortable with synergy estimates in diligence). If deployed in this manner, Orpheus could provide a significant competitive advantage to McKinsey over BCG, Bain and other strategy firms with transaction advisory practices, but without the ability to accurately estimate savings opportunities from procurement in due diligence opportunities.

Fingerprints for Supplier Categorization

In addition to the basic AI classification techniques discussed in the last section, Orpheus has its own supplier fingerprinting technique that allows it to identify supplier similarity beyond simple name and location and industry categories. The deep fingerprint technique allows Orpheus to find real similarities based on actual offerings, geographies and even customer ratings, if the data is available.

Since Orpheus also supports vector space models and mappings to high dimensional mathematical spaces, the fingerprints they can create for suppliers can be clustered extremely accurately using the host of mathematical techniques that the platform supports.

McKinsey implication: This particular strength will provide McKinsey and its clients with even more accurate, cleansed and classified data — to drive opportunity identification — than its past capabilities with Sievo and home-grown analytics tools.

Customized Front-ends for Dashboard-Driven Query and Reporting and Data Science

When it comes to analytics, one size does not fit all. Those who want quick reports, insights and drill-downs to relevant data-cause exceptions and outliers want something that is easy to use — because they are usually busy category managers, program managers or business unit managers, not data scientists.

But the data scientists that have to identify the patterns and insights, organize the data in a way that allows others to see the opportunities, and then track them going forward need a different, much more powerful interface.

Orpheus realizes this and provides both types of front-ends. For those familiar with leaders like Tableau, Qlik or Looker (which are the solutions that predominate the S2C, P2P and S2P platforms in our space), the dashboard front-end is everything you would expect from a modern BI tool. It’s clean, interactive, easy to use and widget-based with linked widgets that update across the dashboard regardless of the drill-down point. Out of the box it can be pre-configured with overview dashboards, supplier analysis dashboards, category and supplier trend analysis dashboards, spend profile dashboards by supplier or region, supplier/category/country risk dashboards, and so on.

And for those that need power, the data science, built on their DataCategorizer technology, allows for drilling down to the transaction level, familying of entities, new views (that include derived dimensions) to be created off the one master cube, dynamic rule creation and modification, the application of AI algorithms for clustering and analysis, and the definition of analytical chains that can take raw data and produce new, derived, measurements and KPIs with threaded analysis technology that allows a user to map input to output in a multi-step analysis journey.

McKinsey implication: Imagine a “post-study” leave-behind dashboard (customized) for the CEO, CFO and CPO — not to mention category analysts and other “cogs” in the procurement ROI machine. This is far more actionable than 150+ pages of PowerPoint. Moreover, McKinsey can sell it standalone with or without a specific study preceding it.

Out-of-the-Box Measures, KPIs and Analysis

Orpheus has been around for almost 15 years and has been working hard to incorporate all of the learnings and best practices it has encountered into its platform since day one, and it shows. Few platforms, even among the global giants, have as many canned measures, KPIs and analysis as Orpheus.

Measures include average invoice price, average order price, time analysis, spend flow and over a hundred others. KPIs include payment days MPV, open purchase order value, active suppliers, active materials, single-source quote and dozens of others.

Out-of-the-box analyses include (in the data science front-end) ad hoc supplier spend-over-time analysis, price and quantity trend analysis, portfolio analysis, spend acceleration analysis, multidimensional material price variance (M-MPV) analysis, FX and market effect analysis, payment terms analysis, and currency risk exposure analysis.

McKinsey implication: Canned reports will provide a toolkit for McKinsey as part of client studies to identify different types of savings and risk reduction opportunities. It also makes it easier to sell the solution “out of the box” to clients on a leave-behind or stand-alone basis.

Solution Weaknesses

The Digital Backbone

While the capability is a strength, the fact that it is still a desktop app is a weakness. It’s a powerful tool, but a tool that is limited to admin users with a Microsoft desktop or those able to access a Microsoft desktop via a Citrix Virtual Desktop (with the bandwidth to support it). This is a noticeable weaknesses given that Microsoft does not own all of the desktops, that only a relative handful of organizations have Citrix, and that users should not be publishing to a server from a desktop in a Cloud architecture. The cube builder portion of the digital backbone needs to be rebuilt on the cloud, like the data science tool, and available to the senior data scientists in the organization to build, extend and modify cubes to adapt to not only organizational needs but potential needs with forthcoming projects and organizational changes.

McKinsey implication: Cloud-based apps are important for data analysts and millennials as they rise through the procurement ranks. But most CEOs, CFOs and CPOs are not (yet) as concerned with cloud and mobile-first readiness. Still, McKinsey should invest in modernizing the digital backbone of the solution as quickly as possible.

Limited Number of Out-of-the-Box Connectors and Extractors

Extractor programs are only available for:

  • SAP / SAP BW
  • S/4HANA
  • MS Dynamics
  • IFS

And extract templates are only available for select other major ERP systems via pre-configured ETL tools, which the client may not possess. (For example, in some instances, Orpheus can extract from Oracle, Infor, Baan and JDE — but not all.)

But global organizations employ dozens of systems, and most big analytic/source-to-pay providers have to support dozens of systems, including:

  • Oracle
  • Agresso
  • JDE
  • Lawson
  • Sage
  • Sun

In order to go beyond other best-of-breed analytics solutions, near real-time integration to major enterprise systems is a key factor in a market where most organizations want more up-to-date insight than dumping data to a data lake every month or week.

McKinsey implication: McKinsey should consider deepening out-of-the-box integrations.


The Data Science UX was clearly designed for data scientists and meets their needs to a T, but as a result, it is not an easy or obvious UI for an average procurement practitioner or manager to use, and they will typically be left with the Dashboard Front-end UX (with basic dashboard-driven reporting and drill-down). And while the Data Science UX application was not initially designed to be used by anyone but a data scientist, given the lack of skilled data scientists out there and the need for more and more people to be informed by, and use, self-serve analytics (as out-of-the-box analytics are only useful twice — once to identify the spend opportunity and once to verify it has been captured), it is critical that the categorizer and data science front-end be as straight-forward and self-explanatory as possible to support advanced analysts as well as true data scientists.

This may require the development of a separate “lite” front-end to the data science tool for procurement analysts that only need to define new measures and simple analytics chains that are just strings of customized pre-packaged components (possibly developed or modified by the in-house or consulting data science team) if certain functionality can’t be simplified by its nature, but Orpheus will need to do work here to increase its broad appeal, especially in the North American market.

McKinsey implication: As with other solution weaknesses, McKinsey should invest in making the UX more competitive for typical analytical users, as not all “data wranglers” are “data analysts.”