Getting the right answer to these questions will greatly improve your odds of a successful spend analysis initiative.
- 1. Can I build as many datasets as I want? At no extra charge?
The answer should be a resounding "yes." Most spend analysis vendors behave as though there's only going to be one spend dataset, built from high-level A/P data. The problem is, after the first few months of discovery, there's not much value left to discover in a high-level A/P dataset. In order to cash in on the continuous ROI potential in your data, you have to apply the spend analysis system over and over again to commodity-specific invoice-level data. It's through this repetition that the keys to spending behavior and accurate invoicing are revealed. When you understand spending at a granular level, real changes can be made and your returns can be extended indefinitely.
Many spend analysis vendors will argue that downstream value comes from "compliance" and "monitoring". But, it's obvious that only the most cursory compliance can be done at the A/P level. For example, you can't see whether you're receiving the contracted price, or for that matter whether the contract even applies to the A/P line items in question. In fact, compliance monitoring requires invoice-level datasets, by commodity, in order to ascertain whether the price is the right price, whether SKU substitution is occurring (and if it should be), or whether more subtle terms of the contract ("industry best price", for example) are being met - or, as is common, completely ignored.
- Can I build datasets all by myself? From raw data? Without help from the vendor or from IT personnel?
The answer had better be "yes." There's no way you can afford to pay a vendor to build (or host) individual datasets for individual analysts — for example, so your commodity managers can find the gold that could be hiding in overcharges at the commodity level.
Which means that the spend analysis system should be operable by business people. It should be possible for them to transform and load raw data on their own, derive their own data dimensions, map transactions and spending, and create custom hierarchies to support their analyses. If you have to rely on a third party, you'll just end up standing in line, frustrated.
Because raw data can be awkward to work with, you need translation tools and data import tools that are able to handle inconsistencies and mismatches. You should be able to build a spend dataset in an imperfect world, and the dataset building process should account for missing joins and missing data in a reasonable and intelligent way, without destroying control totals and without ever — EVER — dropping a single transaction. Finally, you need simple tools that transform one format of data into another. And, all of these tools need to be operable by business people, not just by IT data experts.
- Can I cleanse my own spending data?
The answer – you guessed it. Any vendor who argues that you cannot is ignoring twenty years of sourcing experience during which hundreds of consultants and sourcing experts have built thousands of spending datasets. The
process of classifying and cleansing spend is well understood, and can be followed by any business professional with minimal training.
The key, as always, is selecting a tool that will enable business users to perform data cleansing and mapping on their own, without being "data experts." A spend analysis vendor should be able to show you how to map spending, or alter a commodity hierarchy, with a few simple mouse-clicks.
- Can I change/alter/modify datasets once they're built? By myself? Without help from the vendor or from IT personnel?
Again, the answer had better be "yes." Many vendors with legacy spend analysis products would have you believe that information hierarchies are static, that changes can wait until "refresh time," that canned reports based on those static hierarchies can provide all the value necessary, and so on. Those are false
hopes. In fact, to perform the ad hoc analysis necessary for spend analysis, and especially for commodity-specific spend analysis, data
hierarchies and rules classifications must change continually. Unless those changes are supported by the product, and unless they can be performed quickly and easily by business users, the analysis cannot take place. Analysts must instead take the "back to the future " approach of downloading raw transactions to Microsoft Access or Excel, and hacking at them manually to produce meaningful results — which is exactly where they were prior to purchasing the spend analysis system.
- Can I build my own analysis models, and leverage the spend analysis system to populate them with relevant data?
Again, the answer had better be "yes." A set of canned reports, or a manual to a third-party reporting system won't do the trick. You should be able to build custom analysis models in an environment with which you're already familiar, and have those models populated with data directly from the spend analysis system — all under control of the slice/dice capability of that system — so that you can revisit the same analysis from many different perspectives and cuts.
This is where it's important to distinguish between "ad hoc" reporting and reporting in general. Does the spend analysis system have the ability to produce complex and custom reports, guided by you? Or are its reports written in some programming language like Java or C++, the source code for which is inaccessible to you and unmodifiable by anyone but the vendor?
Canned reports have limited use in the real world. Let's face it:
if one is to believe that a static report from a spend analysis system can
really perform a useful opportunity assessment, as several spend analysis
vendors are now claiming, then one should ask how sourcing consultants (and for
that matter, commodity managers) could possibly stay employed. The fact that
sourcing consultants and commodity managers are doing very well at the moment should
settle the point.
Analysis power is precisely the power that you wield as a business
user, independent of canned reports supplied by a vendor. Complex, multi-page
reports such as the original MMG Commodity Spending Report (below), variants of
which are now commonplace across the e-sourcing space, should be within your
reach to create quickly and easily -- without any programming, database
queries, or other IT magic, and yet with full flexibility to build whatever it
is that you need.
Click here to expand the above.
- Can I deploy the spend analysis system so that individual analysts have the freedom and flexibility to build and modify datasets independently of other users?
You should be able to, absolutely. And, it should cost no more on a per-user basis to do so than if those users were sharing a common dataset.
Consider this reality: if hundreds of users are sharing a common dataset, then nobody can change the dataset, because those changes would affect everyone. If nobody can change the dataset, ad hoc reporting and analysis outside the limited confines of the pre-defined structure of the dataset is impossible. It's as simple as that.
Not only should analysts be able to build multiple datasets on their own PC's and laptops, those analysts should also be completely independent of any central server or Internet connection. And, if the dataset is so enormous that it won't fit on an ordinary PC (true of some insurance claims datasets with 50M or more transactions), then the server should be able to host multiple independent copies of the dataset for individual analysts, each dataset being that analyst's private playground.
There is little value in a centralized spend dataset compared to
the value that comes from allowing individual commodity analysts to attack the
data, and to build their own datasets based on more detailed information.
Detailed RFP Questions
1. Can I build my own datasets? Without vendor assistance?
1.1 How long does it take?
1.2 How many datasets can I have? Does this cost extra?
1.3 Can I share my datasets with other users? Can I send them to others?
1.4 Can other users connect to my computer and view my datasets?
2. Can I run the spend analysis application without Internet
2.1 Can I build datasets on my laptop without Internet connectivity?
2.2 Can I alter datasets on my laptop without Internet connectivity?
3. Can I build my own reports and analyses?
3.1 Can I extract multidimensional data directly into my analysis models?
3.2 Can I easily and quickly modify reports that the vendor provides?
4. Can I map my spend to any classification, or just to UNSPSC or
to the vendor's own custom classification system?
4.1 Why do I need to worry about UNSPSC in the first place, when UNSPSC is not
a useful classification system for many types of sourcing?
4.2 Why would I want to classify my spending by vendor alone? What about
vendors who sell multiple commodities?
5. Can I map spending myself? How long does it take?
5.1 If not, who pays for the changes? Who performs them?
5.2 Who republishes the dataset? How much does this cost?
5.3 Do I need to map the whole dataset, or can I just map the commodities that
I need, as I need them?
5.4 Is mapping an online or offline process? Can I use the context provided by
the dataset viewer to assist in mapping, or must I use an offline tool?
6. Can I change dimensional hierarchies myself? How long does it take?
6.1 If not, who pays for the changes? Who performs them?
6.2 Who republishes the dataset? How much does this cost?
7. If spending is mapped by an automated tool, who performs QA on the automaton to ensure that mistakes are corrected?
7.1 How much does this cost?
7.2 Who corrects this error on refresh, the next time the automaton makes the
8. If I don't want to do any of the work myself, how many
alternative suppliers are there to build datasets, refresh datasets, and so on?
Is the vendor my only choice?