Emptoris: Readying the Spend Visibility Armaments for Battle
In the past few weeks, both Emptoris and Ariba have announced some significant enhancements to their respective spend visibility and analytics products. I'll tackle Ariba's enhancements in a separate post later this week, and focus on Emptoris enhancements in this one. But before diving in, with apologies to both readers and vendors, I should say that I've been unusually slow on the uptake covering these enhancements owing to a large backlog of blog fodder of late (I also have two forthcoming feature entries on how vendors aPriori and Akoya are working to bridge the design engineering / direct materials sourcing intelligence and costing gap, so keep your eyes peeled for these as well).
But I digress ... back to Emptoris, and their news. Given that Purchasing did a good job summarizing Emptoris briefing slides already, I'll focus this post on analyzing the news rather than relaying the specific enhancements in detail, which, in their own words, focus on such areas as core data management capabilities, spend classification, data extraction consulting, and the extension of existing BI and DW capabilities.
In my view, Emptoris has realized the core limitations of many earlier generations of spend visibility approaches which treated such efforts as one-off technology "island" approaches. In their current iteration, Emptoris' capabilities are directed at helping companies integrate and extend spend visibility and analysis capabilities on a continuous basis regardless of existing back-end technology, BI, or data warehousing environments. And they're focused on enabling visibility in a relative quick timeframe (I know some like BIQ would disagree that 90 days is quick, but in the traditional enterprise applications world, 3 months can go by in the blink of an eye, and is commendable).
From an enabling technology perspective, Emptoris is now offering different types of auto-classification algorithms including Bayesian, natural language processing and nearest neighbor. These are extended by rules-based learning and machine learning capabilities to constantly improve the accuracy of the cleansing and classfication capabilities. In my view, what's important here from a practitioner perspective is not the specific technique -- leave the sausage making to the pros, that's what I say -- but the ability to increase the percentage of spend which can be accurately auto-classified and analyzed on a continuous basis.
On the pragmatic level, Emptoris has some real advantages in their procure-to-pay contract monitoring capabilities (gained from leveraging diCarta's contract management with the rest of the Emptoris suite). On top of helping tee-up strategic sourcing opportunities, this can help companies spot control violations (e.g., contractual payment term violations) and potential fraud (e.g., rounded payments, phantom vendors, PO changes after approval) before it's too late to take a rapid yet forward-looking corrective approach. This proactive, results-driven compliance angle to spend visibility and analytics is one that should have strong appeal for more advanced organizations which are looking for next level Spend Management savings and cost avoidance opportunities.
- Jason Busch
But I digress ... back to Emptoris, and their news. Given that Purchasing did a good job summarizing Emptoris briefing slides already, I'll focus this post on analyzing the news rather than relaying the specific enhancements in detail, which, in their own words, focus on such areas as core data management capabilities, spend classification, data extraction consulting, and the extension of existing BI and DW capabilities.
In my view, Emptoris has realized the core limitations of many earlier generations of spend visibility approaches which treated such efforts as one-off technology "island" approaches. In their current iteration, Emptoris' capabilities are directed at helping companies integrate and extend spend visibility and analysis capabilities on a continuous basis regardless of existing back-end technology, BI, or data warehousing environments. And they're focused on enabling visibility in a relative quick timeframe (I know some like BIQ would disagree that 90 days is quick, but in the traditional enterprise applications world, 3 months can go by in the blink of an eye, and is commendable).
From an enabling technology perspective, Emptoris is now offering different types of auto-classification algorithms including Bayesian, natural language processing and nearest neighbor. These are extended by rules-based learning and machine learning capabilities to constantly improve the accuracy of the cleansing and classfication capabilities. In my view, what's important here from a practitioner perspective is not the specific technique -- leave the sausage making to the pros, that's what I say -- but the ability to increase the percentage of spend which can be accurately auto-classified and analyzed on a continuous basis.
On the pragmatic level, Emptoris has some real advantages in their procure-to-pay contract monitoring capabilities (gained from leveraging diCarta's contract management with the rest of the Emptoris suite). On top of helping tee-up strategic sourcing opportunities, this can help companies spot control violations (e.g., contractual payment term violations) and potential fraud (e.g., rounded payments, phantom vendors, PO changes after approval) before it's too late to take a rapid yet forward-looking corrective approach. This proactive, results-driven compliance angle to spend visibility and analytics is one that should have strong appeal for more advanced organizations which are looking for next level Spend Management savings and cost avoidance opportunities.
- Jason Busch
At BIQ our primary marketing task has been one of education. We are a very different proposition, in that we're not selling a spend data warehouse. Instead, we take the following (somewhat radical) positions:
1) A spend data warehouse is not an "analysis" platform. Drilling around a cube? Canned reports? Where's the analysis? Analysis requires instant data classification changes, instant hierarchy manipulation, and intimate multidimensional data exchange with existing and familiar analysis tools.
2) Loading and manipulating raw input data should be accomplishable by ordinary business users -- soup to nuts. No "IT guys in white lab coats" should be required; send 'em back to playing "World of Warcraft" in the server room.
3) Creating and hosting new datasets must be as easy as creating a spreadsheet model. Spend analysis users should anticipate building and hosting many spend datasets, for many different purposes.
4) It should be possible to run a massive spend dataset directly on an ordinary desktop or laptop computer, without any network connection or server -- on an airplane, for example. It should also be easy to host datasets for hundreds of users, if desired; but only if desired.
BIQ users routinely build dozens of private datasets for one-off analyses. BIQ customers typically also retain and host multiple datasets -- not only a traditional AP level spend dataset, but also "MicroMap" datasets for analyzing specific commodities (see my post at E-Sourcing Forum, http://www.esourcingforum.com/?cat=12).
Because their primary deliverable is a generic and difficult-to-differentiate OLAP warehouse, spend data warehouse vendors have competed primarily on the quality of their data cleansing and mapping (and associated services). Unfortunately for them, and despite all the marketing noise and bluster they've generated (I remember contributing to the bluster at Zeborg), this particular Emperor has no clothes. Data cleansing and mapping is actually quite easy to accomplish given the proper tools, and requires no automated or outsourced assistance except in quite rare cases around complex direct materials hierarchies. Many BIQ customers do their own mapping, to self-reported 97% accuracy rates.
Spend data warehouse vendors also run into difficulty when trying to differentiate themselves from traditional BI tools like Cognos and Business Objects -- and from ERP warehouse platforms like SAP BW. We saw such a differentiation attempt by Tim Minahan in a recent comment here on Spend Matters; bravo, Tim, well done. But unfortunately, at the end of the day, what is the difference between cleansed spend data loaded into BW, and cleansed spend data loaded into a custom spend data warehouse? Corollary: How long will it be before a decent rules system is grafted onto Business Objects or SAP BW, and spend data warehouse products become irrelevant?
One of my consulting friends would respond at this point, "ERP vendors will never do spend analysis, because admitting that their data are not clean violates their basic marketing message." Which is funny, but probably not axiomatic.
Thanks for raising the noise level on spend analysis, Jason -- certainly a very timely topic. We'll continue to drive BIQ down our own path in 2007, with innovations like federated datasets and meta aggregation. Of course, those ideas will require yet another round of education -- but nobody is confusing us with Cognos or Business Objects now, and we're driving the gap wider with every release.
Cheers,
Eric
Here’s MY radical thoughts on the topic…
You still need an analytic data model and a physical data store as George Carlin said “a place for your stuff”. Let’s call that, in luddite terms I know, a data warehouse. Yes, that warehouse should be quick to buy/rent, have a small footprint, be easily re-configured, and allow both inbound and outbound (warehouses are not black holes). And it’s part of a larger network, so you may want just one, or you might want to have a distributed network (e.g., aggregating pointers rather than just aggregating/replicating data). Data warehouses in principle are not mutually exlusive to inuitive, flexible analyses that analyzes data in the warehouse.
There’s also nothing wrong with drilling around pre-defined cubes (a lot of people would kill to have just a single view of spend by supplier, commodity, org/region, and cost center for a certain time period) that let you pick the pre-defined variables in your favorite graph/table type and then group/filter/sort/etc. Do I need “intimate multidimensional data exchange”? I’ll save intimacy for the bedroom (“With most grievous dispatch I shall open the latch …”).
Also, I do NOT want users loading and manipulating raw input data. Ever watch an errant outer join shut down a business – don’t GO there! You WANT those guys on the firewall, you NEED those guys on the firewall. Let them set up the pipes and let the users dice and slice and slice to their hearts’ content. Larry Ellison said it best: “users should not be hobbyists”. If however the IT guys are wearing lab coats and there’s not IT ‘embeds’ in the business, they should be fired and/or outsourced (whichever comes first). But, speaking of lab-coats… “federated datasets”? “meta aggregation”? I really do need to sign up for that marketing education course again! Up next: Super-elasticbubbleplastic and Polyrazmatazz (yes, the fool watched way too much TV as a child)
Yes, I agree that Creating and hosting new datasets must be as easy as creating a spreadsheet model as I mentioned before. I also do heartily agree that spend analysis users should anticipate building and hosting many spend datasets, for many different purposes…and that users… typically also retain and host multiple datasets. Yes, many people call these datasets…. “cubes” and the latter are often called… “saved analyses”. “Some call me……Tim?”
The thing is though, software functionality for a data warehouse may be a commodity (including the “proper tools” – whether w/ augmented rule bases / inference engines – or parsing/mapping tools alone), but getting cleansed/enriched/etc. data is not. But it’s hard to blame software vendors for that – users own the data. Having a single data model for master data and analytic data is “necessary, but not sufficient” (to quote Goldratt), nor do ERP vendors claim that. Warehouses don’t have to mean dirty warehouses and it’s not fair to blame software vendors who provide a “place for everything” with user responsibility for “everything in its place”. It’s like blaming your gym because you never went and had a heart attack – puh-lease! And do not taunt the power of a good inference engine. There is value for “relevant search” (or “relevant finding” in Requisite, Inc. terms. Ahhh, Requisite. Let’s say a quick ode’ to RUS – which is not Rodents of Unusual Size btw – but I digress). Google’ market cap is $150B. wow.
I also totally agree with “how long will it be before a decent rules system is grafted onto Business Objects or SAP BW, and spend data warehouse products become irrelevant?” But, shall we extend that metaphor to “meta-aggregation” and other feature-functions that BIQ offers? Just add another year or two for the Borg to work through the latest Feature 500 lists that you’ll be helping populate. But, at least you’re not $100M in deep like Emptoris. Stay lean, mean, and profitable like the Zycus guys.
I must however give BIQ a lot of credit for not falling victim to the buzzword du jour of on-demand, SaaS, blah blah. However, I will soon be adapting BIQ’s web page to the lyrics of John Lennon’s Imagine……. “Imagine there’s no warehouse, I wonder if you can….”.
Just kidding. I won’t. I’ll let someone else have the fun.
Fool out.
One only needs to look at the financial performance of ariba, verticalnet, cmgi & others...
As Jason mentioned above, Emptoris provides our Spend Analysis software as either a Service/on-demand or as behind the firewall software. We also support customers moving from one model to the other, which is seen by many as an advantage. We believe that the flexibility inherent in our approach is ideal for both the customers and for Emptoris. Supporting both delivery models credibly and at proven scale (# of customers, size of customers) affords us a unique and significant advantages in the marketplace. For us it is not about the latest buzz, rather the provision of choice and proficiency in delivery of both models. Many folks mistakenly assume that because we’ve been so successful in the high-end, we are only Enterprise software. If you are in this camp, and talk to some of our largest customers, then you may be extremely surprised at the answers you would get.
We’ve been doing Spend Analysis for over 6 years and there is a fair amount of solution detail and examples on our website for anyone wishing to learn more. http://www.emptoris.com/solutions/spend_analysis.a... .
On Spend Analysis, IT, and Warehouses:
The sad reality is that many companies have attempted to solve the Spend Analysis problem with a BW or BI approach and failed -- miserably. Our solutions are very different from what Warehouses provide and customers “get it” within 5 minutes of hands-on, free-form work, with their own data. Lets give Procurement professionals some credit for being able to identify the difference, even if some here can’t see it. Further -- what we’ve found is that giving the CPO the tools to embrace-and-extend the prior [failed] IT centric strategy creates a win-win for both Procurement and IT.
Spend Analysis is an area where IT has more power to kill/delay projects simply because they feel threatened by its existence. And as Jason notes, helping procurement to overcome this hurdle is a key element in the recent news we made, and a key element in driving real-world adoption of Spend Analysis by procurement.
If the Spend Analysis vendor’s design focus is a departmental solution, then a cavalier attitude towards IT, their values, and needs is fine. However, if the vendor’s aim is provisioning of a strategic solution that has a transformative and enduring value within the blueprint of a company, then the vendor won’t get very far with that song and dance. If you are strategic, then you are mission critical, and if you are mission critical then IT will get involved.
The same applies to the standalone laptop analysis comment. Our SEs do this type of standalone slice-and-dice every day with massive datasets and it works great, instant refresh, instant drill-down, etc. But imagine the repercussions when the CPO has to explain to the CFO and CIO that a laptop containing the entire corporate or division’s spend data was ‘misplaced’. The same applies to data security and segmented visibility within the enterprise. The trick here is to be able to do the analysis magic not just on a one-off project basis, but enterprise wide, analyzing tens or hundreds of detailed level projects concurrently. Examples: http://www.emptoris.com/solutions/speed_and_scalab... .
Happy holidays season everyone.
- Ammiel
Is Emptoris' approach to On Demand to HOST its behind the firewall technology on a unique set of servers for each customer and call it On Demand? Or are there two code sets...even though your VP of Technology guy said no?
Which is it??????????????
And in fact, the code-base is a three-trick pony, because we also power 3rd party Exchanges and some large BPO providers. They use the multi-tenant capabilities in various configurations (e.g. some are fully segmented, and some share supplier master across tenants), all using a single HW/SW environment.
You are right, that as a vendor you need to know what you are going after and design appropriately upfront. We’ve been fortunate to have had requirements and customers, from very early on, for all of these deployment scenarios.
- Ammiel
Throwing spend data into a warehouse and drilling around it is exciting, if you've never done it before. But what happens next, after the blush is off the rose, and the low-hanging fruit has been harvested? I'll tell you exactly what happens next, because I've seen it so many times. If you want to do more, you have to hire a smart analyst, and pay him to extract raw transactions to his desktop. Then he will hack at them with the same stone knives and bear skins that you were using BEFORE you bought your data warehouse, with the same limited results.
Why isn't a data warehouse useful for analysis? Because it's a fixed schema with fixed hierarchies, and provides only the views that were baked in a priori. It can't be modified for two reasons: (1) it's hard to modify (you have to take it offline and hack at it with difficult-to-use tools), and (2) if it's shared by hundreds of users, you can't modify it anyway without a committee decision. And what's the output of a committee decision? A compromise schema that's useless for everyone. Now, what if I'm an analyst, and I need to group stuff together that isn't currently grouped, or I need an entirely new cut at a data dimension -- what do I do? Repeat after me: "download raw transactions to your desktop, and hack at them with stone knives and bear skins."
You can use BIQ as a data warehouse -- it's a damn good one. But you have the flexibility to change its structure any time you want, in seconds; and, you have the ability to run it stand-alone on a personal computer, or host private copies of it on the server so that analysts can have private playgrounds.
Now, as far as users loading and manipulating their own data, there are certainly many people who have no desire to do so. However, we have lots of business users who routinely load data and build their own datasets -- really large ones, too. They do so because they are interested in their data, and interested in driving savings through analysis. An argument against allowing users to do this is is akin to arguing that business analysts shouldn't be allowed to create their own Excel models. We see a real hunger out there for self-empowerment around data analysis, and we intend to feed it. Business users with no IT skills at all are building great BIQ datasets.
There is a big difference between thinking of spend cubes as read-only monolothic entities that require great care and expense to construct, versus building them and discarding them as casually as Excel models. Yes, some cubes should be carefully cleansed and distributed and maintained as "truth" for a wide audience. But that cube hardly answers analysis questions that require profound changes to dataset structure before they can be answered. BIQ makes those profound changes easy, and allows private copies of datasets to exist; thus the deep difference in perspective.
As far as worrying about being "embraced, extended, and extinguished" is concerned, I haven't felt any pressure from ERP or BI vendors so far, and I don't think that we'll feel any in future. When you control your own custom OLAP technology, as we do at BIQ, you can accomplish things that products based on generic technology just can't consider.
Customers need accuracy and coverage- "I need accuracy and coverage of all my spend data, regardless of source system, data quality, languages."
Customers need speed - " I want to get my reports whenever I want them, today, tomorrow or end of the month just by a push of the button."
Customers need Flexibility - " I need this solution to work on my existing IT stack (or some times out side it) and my existing ERP based or custom data warehouse."
Zycus has successfully helped more than 45 Global 500 customers achieve these objectives in last three years. It is good to see others catching up and trying to emulate what we have been talking about so long.
SF- we continue to be mean and profitable but definitely not lean. Let me know whenever you want me to update you on our new developments.
-Anurag
Debate is healthy - not like the wooden software CEOs who get up on podiums at conferences with their competitors and end up singing Kumbaya. 'you're great, no YOU'RE great!' puh-lease.
Actually, good to see Zycus out there talking a little trash. It's like Andy Kaufman's Latka Gravas turning into Elvis or the wrestler guy.
This guy did not even have the brains to write something on his own. Just copied what his website says ! Ask him a couple of questions more bout his own solution and he will faint ;) Hahahaha.....
BTW any offers spend analysis and analytics for a company with very poor IT infrastructure?? and the million dollarquestion, " How much does it cost ? "
Cheers.......