Data
Warehousing and Data Mining
|
| Data warehousing and data
mining are popular terms you will hear more and more about. Here are some
common questions and answers related to data warehousing. |
What is data warehousing?
|
|
Data warehousing is the collection of all
your business critical information into a single store or warehouse. This
data warehouse is then organized and optimized to provide answers to
common business questions. Another feature of well-designed data
warehouses is flexibility through their ability to answer future questions
that may not have existed during the original design of the warehouse.
|
What is data mining?
|
|
Data mining is the process of extracting
useful information from a data warehouse. There are a multitude of
commercial data mining tools available. The primary function of data mining
is to uncover hidden trends and focus points and then to perform what-if
scenarios gauging overall impact on your bottom line.
|
What can you do with data warehousing?
|
|
Well-designed
and well-stocked data warehouses have no limit to
the business questions they can answer. Everyone's primary objective is to
increase profits, so we will concentrate on this area.
|
|
First, take a look at this chart from a prior
business alert from OPSoftware. It shows the results of an
across-the-board price cut of 1% on the bottom line of several industry
segments:
|
|
|
Clearly, across-the-board price cuts can have
a devastating effect on your bottom line. Closer examination of pricing policies
utilizing data mining tools can significantly reduce the negative effects
of price cuts while maintaining the increasingly important low price
perception.
|
|
Individual customers are the 'micro' profit
centers of your business. Precise knowledge of these profit centers and
the impact price and purchasing changes will have on these profit centers
is critical to maintaining overall profitability.
|
|
Good data warehousing tools not only provide
uncovered insights but allow 'what-if' scenarios so you can test the
impact of policies before implementation.
|
Know your customers, the micro view:
|
|
The first steps are to understand your
customers. Below is a chart detailing the number of unique items ordered
by customer. This data was taken from an actual dealer database over a
six-month period using OPSoftware's data mining tool. Your situation may
be different.
|
|
|
64% of this dealer's customers ordered 15
unique items or less. 79% ordered 50 unique items or less. A full 92% of
this dealer's customers ordered 100 unique items or less. On average,
customers were only ordering 34 unique items. This information proved to
be startling in many ways.
|
|
First off, this dealer had over 80,000 items
in his database. How is it possible that 92% of his customers only order
100 unique items? The reason is that even though 92% only order 100 unique
items, each customer may be ordering a different group of 100 unique
items.
|
|
Likewise, why were so many customers (64%)
only ordering 15 unique items? Part of the reason is that customers may
not know the variety of products you offer. Looking closer at this group,
again using data mining tools, you can uncover those customers ordering
from one product group such as toner, but not ordering from another
product group such as paper. Also, these low unique item count customers
are prime targets for increased emphasis on your broader product line advertisements.
|
Pricing at the micro level:
|
|
Just like every customer may order a different
group of 100 items, every customer may have certain items they consider to
be price sensitive. Pricing that relies solely on company-wide sales data
can lead to reduced margins when you take into account individual
customers. For example, let's say that you assume toner cartridges are
generally perceived to be price sensitive and you have instituted low
pricing for these products. Then when you encounter the customer ordering
few toner cartridges, but many specialized ink jet cartridges, you may be
forced to offer additional profit reducing discounts on these ink jet
cartridges while you continue to receive low profits from the few toner
cartridges this customer orders. In this scenario, the ink jet cartridges
are price sensitive and the toner cartridges should be 'catalog' priced
items.
|
|
The process of pricing based on usage and
perceptions is called matrix pricing and is a great tool for increasing
your profits. The power channel utilizes this exact same process except
that their matrixes are fine tuned to actual sales by region and even by
store. Utilizing data mining tools you can monitor the performance of your
pricing policies and fine tune your policies right down to the micro
level.
|
Macro Views:
|
|
You can also use data mining to provide
informative macro views of your company. Orders per day, lines per day,
lines entered by order taker, and sales by day are just a few examples.
Below is an interesting chart showing the number of lines processed by day
of the week. Again, this comes from an actual dealer database covering the
six month period of November 2000 through April 2001. Your situation may
be different.
|
|
|
Staffing can be the highest expense for some
companies. A complete understanding of your process flows and knowing when
to commit staff and when to schedule off days is another way to
'fine-tune' your business with data mining.
|
|
Everyone knows January is the busy season, but
knowing exactly when the slow-down in ordering occurs could be critical
both for staffing and inventory levels. These ups and downs certainly vary
from business to business and again pin-pointing and predicting the exact
timing of a pick up or slow down is a valuable function that can be
provided through data mining. The chart below tracks the number of orders
received by month from our dealer between November 2000 and April 2001.
|
|
|
Looking at this chart, the dealer experienced a
large increase in orders between December and January and a large decrease
between March and April. But what day did these trends begin? Do these
trends repeat year after year, or were they unique to this year? If you are
a stocking dealer, you can benefit greatly from more detailed analysis to
know exactly when stock is needed and exactly when your warehouse should be
empty. This leads to another bottom line boosting benefit of data
warehousing and data mining.
|
Cutting Costs on the Buy Side:
|
|
Since well-designed and well-stocked data
warehouses contain all informational pieces from your business arranged into
efficient and flexible formats, predicting stocking levels and warehouse
delivery times becomes a much easier task. Also, the what-if scenarios of
data mining tools allow for trends to be included in your calculations.
Trends can also be accounted for by the tools themselves such as an average
increase in sales over the last so many years.
|
|
Taking this further, knowing what to stock and
when to stock it can have a significant impact on your bottom line. Take for
example the following chart of Hewlett Packard lines ordered over the same
six month period. Notice that although there is a spike at the predicted
busy season, the line is much flatter indicating less seasonality for this
product line.
|
|
|
Compare this to the number of Smead lines
ordered for the same period. Of course Smead sells filing supplies which are
expected to peak when companies change out files for the previous year. As
predicted, filing supplies have a much more defined peak in January and drop
off in April.
|
|
|
Compare the same time period to the At-A-Glance
product line with many dated goods. Sales are strong in November and drop
off to near nothing by April.
|
|
|
All of these examples show the number of lines
ordered across the entire manufacturer product line. This is only one
example of what you can do with data mining. You could of course plot
individual items or product categories of items in the same way.
|
Putting the Pieces Together:
|
|
Using data warehousing and data mining tools,
you can confidently plan your purchasing well in advance, take advantage of
mid-season specials, reduce or entirely eliminate over-stocks, and reduce
warehousing costs significantly. It is certainly possible to fine tune your
purchasing extracts from a data warehouse to such a level that you could
place all purchase orders and schedule pin-point delivery for an entire year
all within a single day.
|
|
The well tuned stocking dealer should be out of
stock on seasonal items on the exact day the cut, according to your
individual situation, is reached where it becomes cheaper to fill in locally
rather than carry the inventory expense. Likewise, the well tuned stocking
dealer will be fully stocked, in the correct lines, the correct items, and
the correct quantities, the day prior to the requirement.
|
|
Both knowing what your customer orders, and what
they don't order, are major opportunities for increased sales.
|
|
Using data mining, you can begin to anticipate
orders and even go as far as to pre-prepare orders for customers based on
historical usage. Turning usage into orders is a proven technique for
managing and maintaining valuable customers. This managed approach to
customer needs also provides increased value to the customer and helps to further
tie the customer to your business. With data mining, you can provide the
same pin-point, just-in-time fulfillment for your customers.
|
|
Another advantage of many data mining tools is
their ability to simplify the categorization process. As we have repeatedly
said, well-stocked data warehouses contain all data relevant to your
business. This can include data from other sources not commonly used, such as
wholesaler eContent with their excellent product classifications. Taking
advantage of these additional information sources can make the
categorization of customers and items much less of a task that previously
required.
|
|
Using categorization, you can group like
customers and like items to evaluate missed opportunities. If most legal
offices are ordering filing supplies, then you could assume that all legal
offices should be ordering filing supplies. If customers are ordering
printer supplies, you can assume they must be printing, and therefore need
paper. Are they buying it from you? There are an unlimited number of
associations of this type that you can make if you have the correct
information.
|
Bottom Line:
|
Using data warehousing and data mining you can
make great strides in:
- Understanding, identifying and categorizing
your customers.
- Planning for resource usage by identifying
process flows.
- Monitoring, experimentation with what-ifs,
and fine-tuning pricing policies down to the micro level.
- Turning seat-of-the-pants purchasing into
pin-point accuracy.
- Taking advantage of missed sales
opportunities through cross-over product line sales.
|
OPSoftware's Data Warehouse and Mining Tools:
|
|
OPSoftware currently provides data warehousing
and mining tools for dealers using the DDMS computer system. Our extensive
knowledge of this platform and its data structures was the natural starting
point. But our warehouse structure is in no way dependent upon the DDMS
system. It is a standardized and normalized data warehouse fine tuned to the
specifics of the office products industry. As we continue to incorporate
data from many outside and relevant sources, we will continue to grow the
product and extend the capabilities well beyond the current focus.
|
|
|
