Category Archives: SaaS

Dear SaaS Provider, Where’s Your Substance? Being SaaSy is No Longer Enough.

As per our January article, Half a Trillion Dollars will be Wasted on SaaS Spend This Year and, as per a recent article over on The CFO, CFO’s are wising up to the hidden bill attached to SaaS and cloud, which might just be growing faster than the US National Debt (on a per capita basis).

As the CFO article notes, per-employee SaaS subscriptions alone are now costing businesses $2,000 (or more) annually on average, and that’s including ALL employees from the Janitor (who shouldn’t be using any SaaS) to the CEO (who likely doesn’t use any SaaS either and just needs a locally installed PowerPoint license).

To put this in perspective, this says a small company of only 1,000 people is spending 2 MILLION on SaaS (and a mid-size company of 10,000 people is spending 20 MILLION), most of it consumer, and likely a good portion of it through B2B Software Marketplaces because it’s easier for AP. If the average salary is 100K with 30K base overhead, that’s costing the organization 15 (or 150) people, or a 1.5% increase in workforce, which is substantial if it’s an organization that needs people to grow.

And the worst part is that a very significant portion of this spend is overspend or unnecessary spend, with many SaaS auditors and SaaS management specialists finding 33% (or more) overspend as a result of duplicate tools, unused licenses, and sometimes outright zombie subscriptions that just need to be cancelled. Plus, poor management and provisioning leads to unnecessary surcharges that is almost as bad as unused licenses.

There’s no excuse for it, and CFOs are not going to put up with it anymore. SaaS Audit and Management tools are going to become a lot more common, and once the zombie subscriptions, unused licenses, and cloud subscriptions are rightsized, when these companies realize they are still spending at least 1,500 per employee on SaaS and cloud, they are going to start grouping tools by function and analyzing value. If there are two tools that do lead management, workforce management, or catalog management, one is going to go. More specifically, the one providing the least value to the organization. It doesn’t support multiple what-if scenario creation yet or true SSDO, but its more than just simple side-by-side comparison and more analysis capability is on the roadmap for later this year.

So, dear SaaS Provider, it’s important to ask:

  • what’s your substance
  • how do you provide more hard dollar value for that substance than your peers
  • how do you measure it and prove it to the customer
  • … and make sure you’re not the vendor that is cancelled during the audit

And, dear organization who hasn’t done a SaaS audit recently, why haven’t you? You’re sitting on 30% overspend in a category which is likely, with most of the spend split between departments and hidden on P-Cards and expense reports, $2,000 per employee and growing daily. You need to do the audit, rightsize your SaaS, and then centralize SaaS management and SaaS acquisition policy. It’s not a minor expense, it’s a major, business altering, outlay.

GEN-AI IS NOT EMERGENT … AND CLAIMS THAT IT WILL “EVOLVE” TO SOLVE YOUR PROBLEMS ARE ALL FALSE!

A recent article in the CACM (Communications of the ACM) referenced a paper by Dan Carter last year that demonstrated that the claims of Wei et.al in their 2022 “Emergent Abilities of Large Language Models” were unsubstantiated and merely wrong interpretations of visual artifacts produced by computing graphs using an inappropriate semi-log scale.

Now, I realize the vast majority of you without advanced degrees in mathematics and theoretical computer science won’t understand the majority of technical details, but that’s okay because the doctor, who has advanced degrees in both, does, can verify the mathematical accuracy of Dan’s paper, and the conclusion:

LLMs — Large Language Models — the “backbone” of Gen-AI DO NOT have any emergent properties. As a result, they are no better than traditional deep learning neural networks, and are, at the present time, ACTUALLY WORSE since our lack of deep research and understanding means that we don’t have the same level of understanding of these models, and, thus, the ability to properly “train” them for repeatable behaviour or the ability to accurately “measure” the outputs with confidence.

And while our understanding of this new technology, like any new technology, will likely improve over time, the realities are thus:

  • no amount of computing power has ever hastened the development of AI technology since research began in the late 60s / early 70s (depending on what you accept as the first paper / first program), it’s always taken improvements in algorithms and the underlying science to make slow, steady progress (with most technologies taking one to two DECADES to mature to the point they are ready for wide-spread industrial use)
  • the technology currently takes 10 times the computing power (or more) to compute “results” that can be readily computed by existing, more narrow, techniques (often with more confidence in the results)
  • the technology is NOT well suited to the majority of problems that the majority of enterprise software companies (blindly jumping on the bandwagon with no steering wheel and no brakes for fear of missing out on the hype cycle that could cause a tech market crash unequally by any except the dot-com bust of the early 2000s) are trying to use it for (and yes, the doctor did use the word “majority” and not “all” because, while he despises it, it does have valid uses … in creative (writing, audio, and video) applications [not business or science applications] where it has almost unequalled potential compared to traditional ML designed for math and science based applications)

And the market realities that no one wants to tell you about are thus:

  • former AI evangelists and some of the original INVENTORS of AI are turning against the technology (out of a realization that it will never do what they hoped it would, that its energy requirements could destroy the planet if we keep trying, and/or that maybe there are some things we should just not be meddling with at our current stage of societal and technological evolution), including Weizenbaum and Hinton
  • Brands are now turning against AI … and even the Rolling Stone is writing about it
  • big tech and companies that depend on big tech (like Pharma) are starting to turn against AI … and CIOs are starting to drop Open AI and Microsoft CoPilot because, even when the cost is as low as $30 a user, the value isn’t there (see this recent article in Business Insider)

Now, the doctor knows there are still hundreds of marketers and sales people in our space who will consistently claim that the doctor is just a naysayer and against progress and innovation and AI and modern tech and blah blah blah because they, like their companies, have gone all in on the hype cycle and don’t want their bubble burst, but the reality is that

the doctor is NOT against “AI” or modern tech. the doctor, whose complete archives are available on Sourcing Innovation back to June 2006 when he started writing about Procurement Tech, has been a major proponent of optimization, analytics, machine learning, and “AI” since the beginning — his PhD is in advanced theoretical computer science, which followed a math degree — and, after actually studying machine learning, expert systems, and AI, he used to build optimization, analytics, and “AI” systems (including the first commercial semantic social search application on the internet)

what the doctor IS against is Gen-AI and all the false claims being made by the providers about its applicability in the enterprise back office (where it has very limited uses)

because the vast majority of the population does not have the math and computer science background to understand

  1. what is real and what is not
  2. what technologies (algorithms) will work for a certain type of problem and will not
  3. whether the provider’s implementation will work for their problem (variation)
  4. whether they have enough data to make it work

and, furthermore, this includes the vast majority of the consultants at the Big X and mid-sized consultancies who graduate from Business Schools with very basic statistics and data analytics training and a crash course in “prompt engineering” who can barely use the tech, couldn’t build the tech, and definitely couldn’t evaluate the efficacy and accuracy of the underlying algorithms.

The reality is that it takes years and years of study to truly understand this tech, and years more of day-in and day-out research to make true advancement.

For those of you who keep saying “but look at how well it works” and produce 20 examples to prove it, the reality is that it’s only random chance that it works.

With just a bit of simplification, we can describe these LLMs as essentially just super sophisticated deep neural networks with layers and layers of nodes that are linked together in new and novel configurations, with more feedback learning, and structured in a manner that gives them an ability to “produce” responses as a collection of “sub-responses” from elements in its data archive vs just returning a fixed response. As a result they can GENerate a reply vs just selecting from a fixed one. (And that’s why their natural language abilities seem far superior to traditional neural network approaches, which need a huge archive of responses to have a natural sounding conversation, because they can use “context” to compute, with high probability, the right parts of speech to string together to create a response that will sound human.)

Moreover, since these models, which are more distributed in nature, can use an order of magnitude more (computational) cores, they can process an order of magnitude more data. Thus, if there is ten to one hundred times the amount of data (and it’s good data), of course they are going to work reasonably well for expected queries at least 95% of the time (whereas a last generation NN without significant training and tweaking might only be 90% out of the box). If you then incorporate dynamic feedback on user validation, that may even get to 99% for a class of problems, which means that it will appear to be working, and learning, 99 times out of 100 instead of 19 out of 20. But it’s NOT! It’s all probabilities. It’s all random. You’re essentially rolling the bones on every request, and doing it with less certainty on what a good, or bad, result should look like. And even if the dice come “loaded” so that they should always roll a come out roll, there are so many variables that there are never any guarantee you won’t get craps.

And for those of you saying “those odds sound good“, let me make it clear. They’re NOT.

  • those odds are only for typical, expected queries, for which the LLM has been repeatedly (and repeatedly) trained on
  • the odds for unexpected, atypical queries could be as low as 9 in 10 … which is very, very, bad when you consider how often these systems are supposed to be used

But the odds aren’t the problem. The problem is what happens when the LLM fails. Because you don’t know!

With traditional AI, you either got no response, an invalid response with low confidence, or a rare (compared to Gen-AI) invalid response with high confidence, where the responses were always from a fixed pool (if non-numeric) or fixed range (if numeric). You knew what the worst case scenario would be if something went wrong, how bad that would be, how likely that was to happen, and could even use this information to set bounds and tweak the confidence calculation on a result to minimize the chance of this ever happening in a real world scenario.

But with LLMs, you have no idea what it will return, how far off the mark the result will be, or how devastating it will be for your business when that (eventually) happens (which, as per Murphy’s law, will be after the vendor convinces you to have confidence in it and you stop watching it closely, and then, out of the blue, it decides you need 1,000 custom configurations of a high end MacBook Pro in inventory [because 10 new sales support professionals need to produce better graphics] in a potentially recoverable case or it decides to change your currency hedge on a new contract to that of a troubled economy (like Greece, Brazil, etc.) because of a one day run on the trading markets in a market heading for a hyperinflation and a crash [and then you will need a wheelbarrow full of money to buy a loaf of bread — and for those who think it can’t happen, STUDY YOUR HISTORY: Germany during WWII, Zimbabwe in 2007, and Venezuela in 2018, etc.]). You just don’t know! Because that’s what happens when you employ technology that randomly makes stuff up based on random inputs from you don’t know who or what (and the situation gets worse when developers [who likely don’t know the first thing about AI] decide the best way to train a new AI is to use the unreliable output of the old AI).

So, if you want to progress, like the monks, leave that Genizah Artificial Idiocy where it belongs — in the genizah (the repository for discarded, damaged, or defective books and papers), and go find real technology built on real optimization, analytics, machine learning, and AI that has been properly researched, developed, tested, and verified for industrial use.

Procurement Organizations Need Automation, But that DOES NOT Necessarily Mean AI!

A number of leaders in our space, including Sarah Scudder in the comments to this post, have been noting to me that they are seeing AI resonate with companies of all sizes.

Sarah notes that:

1. She’s seeing AI agent automations resonate with smaller companies.

Smaller companies need automation desperately, but it’s important we educate smaller companies that doesn’t mean they need AI. We’ve had adaptive rules-based automation and tailored machine learning in this space for almost 20 years and they can get fantastic results without having to risk being pre-alpha testers for unproven AI while getting the solution they really need for a fraction of the cost of this new, relatively unproven, AI tech! (Remember, firms that dumped millions into this bandwagon need to recoup those millions fast before their investors abandon them, which means high prices for unproven tech!)

2. She’s seeing copilot intelligence resonate with bigger companies who understand risk.

Which makes sense for a small segment of the market who are ready for it because augmented intelligence and automated suggestions with yes/no approvals are great for organizations who

  1. understand risk and
  2. understand the categories/markets/domains they are applying the technology in, because a true expert will identify the 95% of the time it’s working just fine; the 3% of the time it’s probably okay (and not worth the effort to double check manually due to the risk threshold); and the 2% of the time they need to slam the breaks and take over.

However, that’s not a very large segment of the market. What most companies still need is better analytics, category intelligence, and guidance from category experts on how to use it and then where and when to integrate automation and co-pilot capabilities.

Furthermore, I’m also being told that:

3. Mid-Markets are looking for technology they can roll out to the organization at large to get tail-spend under control, manage intake, and/or relieve pressure on Procurement to focus on more strategic efforts.

Which resonates, but, again, this is an area where AI is typically not needed. Catalogs, be they hosted, punch-out, hybrid, etc. with the ability to also request/book standard, pre-negotiated, services, easy search, and easy RFQ where there is no standard item but the buyer has budget authority, the vendors are preferred, and the amount doesn’t hit a threshold is often enough. Maybe a natural language search to find the right policy documents or bring up the right products or forms, but that doesn’t require modern AI either — we’ve had that for quite some time as well.

And, as Sarah implies, while organizations of all sizes need help to overcome their excessive workload and limited market insight so that they can prioritize risk management and mitigation in their procurement activities, this doesn’t mean they need AI. Automation yes, advanced technology a definite yes, but AI, rarely! Remember that when building and recommending ACTUAL solutions and not just buzzwords.

Spendata: A True Enterprise Analytics Solution

As we indicated in our last article, while Spendata is the absolute best at spend analysis, it’s not just a spend analysis platform. It’s a general-purpose data analytics platform that can be used for much more than spend analysis.

The current end-state vision for business data analytics is a “data lake” database with a BI front end. The Big X consultancies (aided and abetted by your IT department, which is only too eager to implement another big system) will try to convince you of the data paradise you’ll have if you dump all of your business data into a data lake. Unfortunately, reality doesn’t support the vision, because organizational data is created only to the extent necessary, never verified, riddled with errors from day one, and left to decay over time as it’s never updated. The data lake is ultimately a data cesspool.

Pointing a BI tool at the (dirty) lake will spice up the data with bars, pies, waves, scatters, multi-coloured geometric shapes, and so on, but you won’t find much insight other than the realization that your data is, in fact, dirty. Worse, a published BI dashboard is like a spreadsheet you can’t modify. Try mapping new dimensions, creating new measures, adding new data, or performing even the simplest modification of an existing dimension or hierarchy, and you’ll understand why this author likes to point out that BI should actually stand for Bullsh!t Images, not Business Intelligence.

So how does a spend analysis platform like Spendata end up being a general-purpose data analytics tool? The answer is that the mechanisms and procedures associated with spend analysis and spend analysis databases, specifically data mapping and dimension derivation, can be taken to the next level — extended, generalized, and moved into real time. Once those key architectural steps are taken, the system can be further extended with view-based measures, shared cubes where custom modifications are retained across refreshes, and spreadsheet-like dependencies and recalculation at database scale.

The result is an analysis system that can be adapted not only to any of the common spend analysis problems, such as AP/PO analysis or commodity-specific cubes with item level price X quantity data, but also to savings tracking and sourcing and implementation plans. Extending the system to domains beyond spend analysis is simple: just load different data.
The bottom line is that to do real data analysis, no matter what the domain, you need:

  • the ability to extend the schema at any time
  • the ability to add new derived dimensions at any time
  • the ability to change mappings at any time
  • the ability to build derivations, data views, and mappings that are dependent on other derivations, mappings, views, inputs, linked datasets, and so on, with real-time “recalc”
  • the ability to create new views and reports relevant to the question you have … without dumping the data to Excel
  • … and preserve all of the above on cube data refreshes
  • … in your own copy of the cube so you don’t have to wait for anyone to agree
  • … and get an answer today, not on the next refresh next month when you’ve forgotten why you even had the question in the first place

You don’t get any of that from a spend analysis solution, or a BI solution, or a database pointing at a data lake. You only get that in a modern data analysis solution — which supports all of the above, and more, for any kind of data. A data analysis system works equally well across all types of numeric or set-valued data, including, but not limited to sales data, service data, warranty data, process data, and so on.

As Spendata is a real data analysis solution, it supports all of these analyses with a solution that’s easier and friendlier to use than the spreadsheet you use every day. Let’s walk through some examples so you can understand what a data analysis solution really can do.

SALES ANALYSIS

Spending data consists of numerical amounts that represent the price, tax, duty, shipping, etc. paid for items purchased. Sales data is numerical amounts that represent the price, tax, duty, shipping, etc. paid for items sold.

They are basically the inverse of each other. For every purchase, there is a sale. For every sale, there is a purchase. So, there’s absolutely no reason that you shouldn’t be able to apply the exact the same analysis (possibly in reverse) to sales data as you apply to spend data. That is, IF you have a proper data analysis tool. The latter part is the big IF because if you’re using a custom tool that needs to map all data to a schema with fixed semantics, it won’t understand the data and you’re SOL.

However, since Spendata is a general-purpose data analysis tool that builds and maintains its schema on the fly, it doesn’t care if the dataset is spend data or sales data; it’s still transactional data and it’s happy to analyze away. If you need the handholding of a workflow-oriented UI, that can also be configured out of the box using Spendata‘s new “app” capability.

Here are three types of sales analysis that Spendata supports better than CRM/Sales Forecasting systems, and that can’t be done at all with a data lake and a BI tool.

Sales Discount Variation Analysis Over Time By Salesperson … and Client Type

You run a sales team. Are your different salespeople giving the same mix of discounts by product type to the same types of customers by customer size and average sales size?

Sounds easy right? Can’t you simply plot the product/price ratio by month by salesperson in a bubble chart (where volume size correlates to bubble size) against the average trend line and calculate which salespeople are off the most (in the wrong direction)? Sure, but how do you handle client type? You could add a “color” dimension, but when the bubbles overlap and the bubbles blur, can you see it visually? Not likely. And how do you remember a low sales volume customer which is a strategic partner, so has a special deal? Theoretically you could add another column to the table “Salesperson, Product/Price Ratio, Client Type, Over/Under Average”, and that would work as long as you could pre-compute the average discount by Product/Price Ratio and Client Type.

And then you realize that unless you group by category, you have entirely different products in the same product/price ratio and your multi-stage analysis is worthless, so you have to go back and start again, only to find out that the bubble chart is only pseudo-useful (as you can’t really figure it out visually because what is that shade of pink (from the multiple red and white bubbles overlapping) — Fuchsia, Bright, or Barbie — and what does it mean) and you will have to focus on the fixed table to extract any value at all from the analysis.

But then you’ll realize that you still need to see monthly variations in the chart, meaning you want the ability to drag a slider or change the month and have the bubble chart update. Uh-oh, you forgot to individually compute all the amounts by month or select the slider graph! Back to square one, doing it all over again by month. Then you notice some customers have long-term, fixed prices on some products, which messes up the average discount on these products as the prices for these customers are not changing over time. You redo the work for the third (or is it the fourth? time), and then you realize that your definitions of client type “large, medium, and small” are slightly off as a client that should be in large is in medium and two that should be in small were made medium. Aaarrrggghhh!!!

But with Spendata, you simply create or modify dimensions to the cube to segment the data (customer type, product groups, etc.) You leverage a dynamic view-based measure by customer type to set the average prices per time period (used to calculate the discount). You then use filters to define the time range of interest, another view with filters to click through the months over time, a derived view to see the performance by quarter, another by year. If you change the definition of client type (which customers belong to which client type), which products for customers are fixed prices, which SKU’s that are the same type, time range of interest, etc. you simply map them and the entire analysis auto-updates.

This flexibility and power (with no wasted effort) gives you a very deep analysis capability NOT available in any other data analysis platform. For example, you can find out with a few clicks that your “best” salesperson in terms of giving the lowest average discount is actually costing you the most. Turns out, he’s not serving any large customers (who get good discounts) and has several fixed price contracts (which mess up the average discounts). So, the discounts he’s giving the small clients, while less than what large customers get, are significantly more than what other salespeople provide to other small customers. This is something you’d never know if you didn’t have the power of Spendata as your data consultant would give up on the variance analysis at the global level because the salesman’s overall ratio looked good.

Post-Merger White-Space Analysis

White space sales analysis is looking for spaces in the market where you should be selling but are not. For example, if you sell to restaurants, you could look at your sales by geography, normalized by the number of establishments by type or the sales of the restaurants by type in that geography. In a merger, you could measure your penetration at each customer for each of the original companies. You can find white space by looking at each customer (or customer segment) and measuring revenue per customer employee across the two companies. Where is one more effective than the other?

You might think this is no big deal because this was theoretically done during the due diligence and the opportunity for overlap was deemed to be there, as well as the opportunity for whitespace, and whatever was done was good enough. The reality couldn’t be further from the truth.

If the whitespace analysis was done with a standard analytics tool, it has all the following problems:

  • matching vendors were missed due to different name entries and missing ids
  • vendors were not familied by parent (within industry, geography, etc.)
  • the improperly merged vendors were only compared against a target file built by the consultants and misses vendors
  • i.e. it’s poor, but no worse than you’d do with a traditional analytics tool

But with Spendata, these problems would be at least minimized, if not eliminated because:

  • Spendata comes with auto-matching capability
  • … that can be used to enrich the suppliers with NAICS categorization (for example)
  • Spendata comes with auto-familying capability so parent-child relationships aren’t missed
  • Spendata can load all of the companies from a firmographic database with their NAICS codes in a separate cube …
  • … and then federation can be used to match the suppliers in use with the suppliers in the appropriate NAICS category for the white space analysis

It’s thus trivial to

  1. load up a cube with organization A’s sales by supplier (which can be the output from a view on a transaction database), and run it through a view that embeds a normalization routine so that all records that actually correspond to the same supplier (or parent-child where only the parent is relevant) are grouped into one line
  2. load up a cube with organization B’s sales by supplier and do the same … and now you know you have exact matches between supplier names
  3. load up the NAICS code database – which is a list of possible customers
  4. build a view that pulls in, for each supplier in the NAICS category of interest, Org A spend, Org B Spend, and Total Spend
  5. create a filter to only show zero spend suppliers — and there’s the whitespace … 100% complete. Now send your sales teams after these.
  6. Create a filter to show where your sales are less than expected (eg. from comparable other customers or Org A or Org B). This is additional whitespace where upselling or further customer penetration is appropriate.

Bill Rate Analysis

A smart company doesn’t just analyze their (total) spend by service provider, they analyze by service role and against the service role average when different divisions/locations are contracting for the same service that should be fulfilled by a professional with roughly the same skills and same experience level. Why? Because if you’re paying, on average, 150/hr for an intermediate DBA across 80% of locations and 250/hr across the remaining 20%, you’re paying as much as 66% too much at those remaining locations, with the exception being San Francisco or New York where your service provider has to pay their locals a cost-of-living top-up just so they can afford to live there.

By the same token, a smart service company is analyzing what they are getting by role, location, and customer and trying to identify the customers that are (the most) profitable and those that are the least (or unprofitable when you take contract size or support requirements into account), so they can focus on those customers that are profitable, and, hopefully, keep them happy with their better talent (and not just the newest turkey on the rafter).

However, just like sales discount variation analysis over time by client type, this is tough as it’s essentially a variation of that analysis, except you are looking at services instead of products, roles instead of client types, and customer instead of sales rep … and then, for your problem clients, looking at which service reps are responsible … so after you do the base analysis (using dynamic view based measures), you’re creating new views with new measures and filters to group by service rep and filter to those too far beyond a threshold. In any other tool, it would be nigh impossible for even an expert analyst. In Spendata, it’s a matter of minutes. Literally.

And this is just the tip of the iceberg in terms of what Spendata can do. In a future article, we’ll dive into a few more areas of analysis that require very specialized tools in different domains, but which can be done with ease in Spendata. Stay tuned!

Don’t Zip Through the Zip-sponsored Spend Matters authored Intake and Procurement RFP! [2024] (Collected Links)

Don’t Zip Through the Zip-sponsored Spend Matters authored Intake and Procurement RFP!

Please note this is NOT coverage of Zip. See this post for Zip solution coverage!

BONUS

BONUS 2