cu») United States
c2) Patent Application Publica
High et al.
(54) PREDICTIVE PATTERN PROFILE PROCESS
oy
Applicant: Wal-Mart Stores, Ine, Bentonville, AR
ws)
Inveators: Donald High, Noel, MO (US); Michael
Atehley Springdale, AR (US) Jennifer
Stegemoller, Pea Ridge, AR (US)
(21) Appl. Nos 14707628
(22) Filed: May 8, 2015
Related U.S. Application Data
(60) Provisional application No, 61/991,143, filed on May
9.2014.
US 2015032470281
iON (10) Pub. No.: US 2015/0324702 Al
(43) Pub, Date Nov. 12, 2015
ublication Classification
(1) neck
GON 5704 2006.01)
(2) US.CL
fd GOBN $/047 (201301)
on ABSTRACT
A system and method for providing a prediction is disclosed
A series of historical data or profile ae collected and stored
‘ona database, Using sofware to provide ft set of predic-
tive patiers from the historical profiles and a second set of
predictive patters from the fis set of predictive pattems froma
‘which the predietion fora particular subjet such as sales ean
be provided
[PREDICTIVE PATTERN
16
12
PREDICTIVE
< PATTERNS
[| repository
8
SI
FORECASTING) PREDICTIVE PATTERN
eee "ANALYTIC PROFILES ENGINE,
SOFTWARE sie PREDICTIVE PATTER
L106 a
FORECARTING TEAM.
PREDICTINE
PATTERNS
REPOSITORY
118
i FORECASTING]
7 ANALYTIC.
i SOFTWARE
06
128
EXECUTIVE
128
PREDICTIVE PATTERN
DISTRIBUTION ENGINE
120
‘GRANNOLAR LEVEL
PATTERNUS 2015/0324702 AI
Nov. 12, 2015 Sheet 1 of 3
Patent Application Publication
wh
L-Old 2NLNORXS
9 ca
NWALIVd
TBAT UV INNES
307
wr BYYNLIOS
aNioNa NONERULSIO ‘ULATHNY aa
NwaLivd sU9IORd | fnusvoauog Reuse
ar
AYOLISOdSY
SNuaLIvd
BAULOIGRYS
zy
wr Or
NYALIVd BALLOIORYe FT SRVNLIOS a
BNIONA STIIOUd OMLATNY
NUBLIVd SAUOIGRYd)—[ONILSVORNIOS ee
RYOUSOGTY
SNYALLWd
BNULOIGRYd
or
NYALLVd 3AUSIOBdPatent Application Publication Nov. 12, 2015. Sheet 2 of 3 US 2015/0324702 AI
Percentage Data for Hourly Profile Pattern by Day of the Week
037% 0.34% 0.24% 1.495% 055% 0.36% 0.72%
0.25% 0.15% 0.20% 0.53% 0.13% 059% 0.07%
011% 0.07% 0.43% 0.46% 011% 019% 0.15%
0.22% 1.06% 0.06% 0.69% 0.12% 0.35% 0.16%
O10% 306% G.o7% 0.13% 017% 0.49% 0.18%
0.13% 0.07% 0.24% 0.09% 139% 0.14% 0.10%
033% 0.18% 0.15% 045m 036% 0.22% 0159%
100% 0.71% 0.57% — 2.36% 0.08% 0.96% 0.52%
136%" 4.04% 2.42% 4.44% 1.09% 187% 188%
3.80% 0.12% 5.07% 7.48% 550% 3.22% 3.60%
565% 4.23% 7.58% 5.30% 2108% 4.81% 6.65%
8.69% 12.73% 5.94% 11.19% 977% = ATT — 4.42%
BE7% 940% 9.77% 4.1% aign% 928% A4.26%
7.03% 6.99% 6.08% 14.43% 11.40% 4.88% 6.00%
11.02%" 11.91% 453% 0.37% 5.37% 11.18%) 12.35%
9.29% 7.74% 10.07% = 4.76% 775% 11.49% 4.42%
638% 11.85% 4.66% 071% 10.60% 5.77% 4.26%
12.01% 5.96% 12.71% 3.88% «12.67% «15.35% 3.42%
434% 12.03% 12.70% 15.17% 6.28% 5.59% 8.57%
9.12% 1.51% 6.31% 5.49% 5.44% 11.17% 6.24%
227% 750% 2.88% 4.91% 719% 152%" 20.55%
5.11% 0.01% 5.14% 1.36% 118% 1.08% 1.24%
1.63%" 2.92% 1.58% 2.20% 335% | 3.35% 9.46%
0.83% 0.42% 1.49% _ 0.77% 0.78% 0.67% _ 3.65%
200
FIG. 2Patent Application Publication Nov. 12, 2015. Sheet 3 of 3 US 2015/0324702 AI
an
302
HISTORICAL,
FACTS.
TAKE HISTORY
AND MAKE
PREDICTIONS
306
FORECAST |~
PREDICTIONS
PREDICTIONS AND
MAKE PATTERNS
116
PREDICTIVE |[~“
>) PATTERN
310
INPUT SELECTION
CRITERIA FOR
FUTURE
‘CREATE FUTURE
PREDICTIONS
FROM PATTERNS
AND HISTORICAL
| FACTS BASED
UPON SELECTION
CRITERIA,
PREDICTIONS |»
34
GRANULAR [~/
EVEL PATTERN
FIG. 3
‘VIEW FUTURE
PREDICTIONSUS 2015/0324702 Al
PREDICTIVE PATTERN PROFILE PROCESS
RELATED APPLICATION
10001] This application claims priority to U.S. Provisional
Pateat Application No, 61/991, 143, eaitled “Predictive Pat-
tern Profile Process”, filed May 9, 3014. The contents of the
above-relerenced application are herein incorporated by ref
‘erence in their entirely
FIELD OF THE INVENTION
10002] Thepresent invention relates generally to predictive
pattem profile. More particularly, the present invention
felates to predictive pattem profiles that are more accurate
‘ad takes ess storage spoce,
BACKGROUND OF THE INVENTION
10003] Predictive Pater Profile isthe process of breaking
‘down forecast estimate into any dimensional pattern. The
‘dimensional pattem could be by date, hou, item, losation,
‘departaent, son, division, and the lke, Forecasting allows,
‘operation management to match supply with potential eus-
tomer demands, Forecast modeling has been used in acoual-
ing t0 provide cash management, finance to predict equip-
‘ment replacement needs, and in operations to predict work
assignments and workloads,
10003) "Forecasting is very important to supply chains and
listibution inorder to prevent shortages or excesses, which
‘can cause miss deliveries, poor customer service, and work
dliseuptions. Accordingly, itis dosirable to provide aceurate
and timely predictive pattem profiles so that operations man-
fagement can propery allocate the resourees to mest any
potential spike in demands or adjust to decrease in demands
based on certain factors such as seasons or other market
forces.
SUMMARY OP THE INVENTION
10005] The foregoing needs aremet, toa great extent, by the
present invent, wherein in one aspect a apparatus is pro-
Vided that in some embodiments includes & non-trnsient
‘computer readable medium containing program instructions
Tor causing computer perform the metho of collecting @
firsthistoricl profile fora desired subjoct, creating forecast
forthe desired subject based on the fst historical prole
‘creating first set of predictive pattems based on the created
Torecast, discarding the Fist historical profile once the Fist set
‘of predictive pattems for the desired subject are created,
receiving request fora predictionof the desired subject, and
‘creating a second set of predictive patterns forthe desired
subject asedon the frst set of predictivepatters to erate the
prediction.
10006] In accordance with another embodiment of the
present invention, amethod of forecasting ofadesired subject
Js provided and includes the steps of eollecting a frst histoi-
cal profile for the desired subject from a database stored ona
‘computing device, reating, wih a processor ofthe compute
Ing device, a forecast for the desired subjoct based onthe fst
historical profil, creating, withthe processor ofthe comput-
ing dovie, a first set of predictive patterns based on the
‘create forecast, discanting the fisthistorical profile from the
database once the first set of predictivepattems forthe desired
subject are created, receiving a request fora prediction ofthe
desired subject, and erating, withthe processor ofthe m=
puting device, a second set of predictive pattems for the
Nov. 12, 2015
desired subject based on the fist set of predictive patterns to
‘reat the prediction and an updated fist historical profile
[0007] In accordance with another embodiment of the
present invention, computing device that provides aprec-
tion that inckades a processor in communication with a
memory, a database having a fist historical profile for a
sired subject and being stored on the memory wherein the
processor performs the following steps of ereating a forecast
for the desired subject based on the first historical profile,
creating fist set of predictive patterns based on the created
Torecas, discarding the first historieal profile Irom the data-
‘ase once the frst set of preitive pattems for tho desired
subject are created, receiving a request fra prediction of the
desired subject, and creating, with the processor ofthe com-
pring device, a sccond set of predictive pattems for the
Sesired subject based on the fist st of predictive patterns 0
‘reat the prediction and an updated fist historical profile
[0008] There has thus becn outlined rather broadly, ceiain
‘embodiments of the invention in onder that the detailed
description thereof herein may be better understood, and in
‘order that the present contribution to the art may be beter
Appreciated. There are, of course, atonal embodiments of
the invention that will be deseribed below and which will
{orm the subject matter of the claims appended hereto,
[0009] In this respect, before explaining at least one
‘embodiment of the invention in detail, iti to be understood
‘thatthe invention snot imited in its application to the details
‘ofeonstmicton and to the arrangements the components et
{orth in the following description or illustrated inthe draw
‘ngs. The invention is capable of embodiments in addition to
those described and of being practiced and carried out in
various ways. Aso, ts tobe understood that the phraseology
and terminology employed herein, as wellas the absiret, are
{or the purpose of deserption and should not be regarded as
limiting.
[0010] As such, those skilled inthe art will appreciate that
the conception upon which this diselosure is based may
eu be ullized asa basis forthe designing of othe stuc-
tres, methods and systems for carrying out the several pur
poset ofthe present invention. [is important, daeefore, that
‘he claims e rogardeas inching sich oquivalent constrne-
tions insofar as they do not depart from thespiritand seope of
the present invention.
BRIEF DESCRIPTION OF THE DRAWINGS,
[0011] FIG. 1 is a forecasting system according w an
embodiment of the iavention,
[0012] FIG. 2 illustrates an exemplary percentage data
hourly profile according to embodiment ofthe invention,
[0013] FIG. 3 illustrates a method that provides proietive
pattems according to an embodiment ofthe invention,
DETAILED DESCRIPTION
[0014] The invention will now be desribed with reference
‘othe drawing figures, in which like reference numerals refer
to like parts throughont. An embodiment in accontance with
the present invention provides a system that ean dynamically
szencrite predictive patterns for varity of uses such as pre-
‘ictng sales, cash low, potential customer claims, fees 36s0-
ciated with third-party services, warranty, etums, energy.
supplies, et.
[015] "Anembodiment ofa forseasting system 100s illas-
‘rated i FIG. 1 Forecasting team 102 ean include one personUS 2015/0324702 Al
‘ora group of persons that specializes in forecasting various
subjects, such as sales, inventory, behavior, weather, stocks,
purchases, deliveries, elections, spread of diseases and other
Subjects. A computing device 104, sueh as a compute, note-
book or tablet can be used to acess a remote computer oF
server 112. Aliematively, forecasting team 102 can directly
access server 112 witha user input that isciretly orindiretly
‘connected tothe server 112. Computing device 1M inclides
{information that has been transfered from the various data-
bases 108, 110. Altematively, the databases 108, 110 are
‘connectedto server 12. The information from databases 108,
110can be accessed remotely viaa wired or wireless connec”
tion othe databases ean be stored on computing device 104
‘and/or server 112.
10016] The databases 108, 110 may con
sales, cashflow, customer claims, foes as
party serviees, warranty, returns, energy usage, supplies, ten-
‘dor types, cash on hand, shrink (difference berween actual
pliysical inventory and computer inventor), computer usage,
‘tc. Ry mining this data, operations manager can make better
‘decisions on a variety of subjects for the company. The data
‘ean range from previously collected data (bstricl) to cur
rently collected data or realtime data,
[0017] As stores may accept different types of tender
Including creditcard, debitearis, checks, cash, coupons, gift
‘cards, food stamps, employee pay ears, electronic cash
Bitcoin, PryPal) and the ike. By having accurate tender type
{oreeas information, the company can better predict the foes
associated with various tender types, how sell the dilfeent
tender types are performing, and have only the necessary
amount of eas on hand for & given time period (eek, day,
hour, ot). Fach tonder type could be assigned a cenain
percent and categorizedby day, by site, or by hour that ould
then be used to ealctlate or erete the forecast at any level
‘within the company, such as store, region, division, of even,
total company. Purlier, depending on when the fords from
the various tender ae predicted be eveived, the company
‘ean device when to invest the funds or when pay down debt
[0018] Often times, a store is changed for ert cant pro-
‘cessing foes or debit procesting fees based on the number oF
‘amount of transactions that were processed during a cert
time period, such a daily, weekly, monthly or yearly. By
predicting the amount including volume of tansoetions fora
‘ziven period, the company can potentially negotiate better
Fates on the fees based on either total amounts oF total volume
‘of various transactions.
10019] The siore change fund isthe amount of eash that &
sore needs to have on hand to perform their business for @
given day. Too much cash inthe store means the cash is not
being used to provide the most benefit tothe company since it
‘could be used to invest oF pay down debt. Too litle cash
tans the store canot perform their business because they
‘don't have the funds available for payroll, check cashing,
petty eas, or providing cash back on other transactions to
their customers. The store's change fund varies by day and by
_denoiination needed inthe store. Each cash register routes
‘certain amount of eash of specifie denominations tobe able
to provide change and cash back to customers. Fach denomi-
nation could be assigned a percentage ofthe total number of
transactions that are required for that particular denomination
to complete the transaction by repster, by store, of by hour.
0020] Shrink s the difference heween the physical inven-
tory on hand and the merchandise that sreportel a available
jn the system, This may result from thet, damages! goods,
Nov. 12, 2015
wise. Bach store noes
1 eran periods of ine, sch
as month, quarter, season, or year, Fach prouet could be
‘signed a percentage By store oF by dy thl wouldallo for
better forecasting of shrink
{0021} Customer claims are filed when acustomerhas been
injured within the premises ofa store, such as sipping or
provided incomest medication. Hach type of claims ean be
{bsigned a pereeataye by store orby department tha coukl be
‘edt prot costomerclaims. This allows forproper insur
!ncocaverageto be pease haseon the foeeeast aims
forthe soe
{0022} Mainframe used is measured in MIPS (Millions of
Instructions per Second, which has ests associate with it
MIPS ean be purchased your or more in advance based ona
4 our average of peak usage. New projets or upgrades to
cemputee systems that wil consume MIPS are estinsted and
aukedothecurent usage andused to calelatetheamount of
‘MIDS datncads tobe purchased IFMIPS usage was assigned
4 percentage for each hour of the year, loreasis could be
enerted hat woul show theeureat peaks and where teans
ould apply efoto reduce MIPS usage and thus, decrease
the amt of MIPS that aces to be purchased
{0023} "The data from databases 108,110 can be fed into
{orecasting analytic software 106, which converts the data to
foyecst results by computing device 104. The Forecasting
analytic software 106 cin be stored on a memory ofthe
computing device or remotely. Then the forecast results can
befidintoaprodictive pattems profiles engine 11 thatcanse
‘aris software modules t create pater profiles, such as. 9
Tistsetorpredivtvepaters 16-Stepsusel io eee test
set of proictve pattems 116 can inlode detemining the
purpose of the forecast, defining time period using the
pprpriate data or dat points, using the appropriate snd
relevant foreasting ecnigues, creating the patterns, 0
toring the scl pattems for accuse, and adjusting aceon
inal:
(0024) Forecasting can inchde qualitative andor quanitar
Sve data, Qualitative dat allows for human factor sch as
Inuehes, and feetings while quantitative data is objective or
‘ard data. Other tecinigues tht can be sed With Bistrial
data inte inear regression analysis and various averaging
tecinigues suchas moving avenge, weighted moving aver
age and exponential smoothing. The averoging techniques
allo foestocthingof peaks ad valleysthat acer nthe ta
an tend to remove random vadations that coeur. Movin
verge ube a umber of recat actual values, and can be
Updated as new valves of data points become availble
Weighted moving averages are similar w moving averages
but assign more weight fo the most recent vale na series.
Exponential smoothing uses previous forecasts pls a per
eatage of te forecast emo.
(0025) Sil other techniques tht could be sed to forscst
Jnchie trend forwastng using tend equation (linear rend
quation) andor endadjustedexponcatal smoathing or ed
torecurting events sich 3 seasonal variations and eyes. The
cxtmples of techniques described rein are ot meant to Be
Timitng and the techniques canbe sed hy self on come
bination with eachother These examples of forecssting tech.
niques can be stored on aa intemal memory of computing
device 104 or server 112, o nan external memory (attached
orremot)
(0026) Once the st set of predictive pattems 116 re gene
erated, they eam be stored on predictive patterns repositoryUS 2015/0324702 Al
118 on server 112, Like the forecasting teclaniques, the pre-
dictive pattems repository 118 can be stored on an intemal
memory of the server 112, computing device 104 or in an
‘extemal memory (attached or remote).
10027] The frst set of predictive patterns 116 can inelude
dimensional pattern such as date, hour item, fine line, depaet-
‘ment, store, and division. Thus, each day of the week, each
hour ofthe day can includea profile or pattern of sles, visits,
scans, umber of cashiers, andl her setvities. The fist et oF
predictive pattems 116 ean be eapred to any granlar evel
‘desired by the user and can represent the percentage of the
amount for that time period of dimension. The first sct of
predictive pattems 116 ot Kmited to time and dae but msty
be built based upon any dimensions such as who, what,
where, when, how aad the lke including vendors, items oF
‘can even be based on hierarchy. The fist set of predictive
patterns 116s stored and reused 2s needed andl can be stored
‘in realy identifiable manner.
10028] "In one embodiment, the data stored in the database
108, 110 is continuously and dynamically updated with the
existe datapoints, However, once the fist set of predictive
patlems 116 is generated, the data that was use to create the
fint set of predictive pattems 116 may be discarded or
deleted. Thisallows for even moredatato beused to ereatethe
first set of predictive patterns 116 but also decrease the
amount of storage needed fr the database 108, 110 as vast
‘amounts of data eat be needed to create the first set of pre=
dictve pattems 116, Further, as new pattems are generated
they will dynamically replace old ones thereby improving the
‘orecasting.
10029] In other embodiments, the frst set of predictive
pattems 116 ay'be created if they ar 5% different fom each
biher, Aditioglly if to oF more patterns ate within 95% of
‘each other, then one would be discarded. That is, if two
pallems are 98% similar to cach other in dat, den data from
the first pattem wll beassimilated into thesscand pater and
the first patter is deleted. reference ean be made thatthe
‘information of the first patter is now inthe second pattern.
Further it may not be necessary to reat all possible patterns
but ereate patterns when a now pattern wold be more than
atstially 58% different from ‘an existing patter These
‘embovdiments can save storage space and costs every pat=
teen does not have to be generated
10030] | Once te frst set of predictive patterns 116 as been
‘created and stored, ational manipulation of the forecast
‘can oocur. At this point, the previous data that was used to
make the frst set of predictive pattems 116 can be deleted 10
savestorage space. Theft set of predictive pattems 116 cua
be dynamically updated as addtional data becomes available
‘and stored in databases 108, 110. The forecasting team 102
‘can at tis point create addtional pattems ora second set of
predictive pattems 122 based on the frst set of predictive
pattems 116, The second set of predictive patterns 116 can be
‘generated and stored on prictve pattems repository 118 or
‘can be generated when requested. be ereation ofthe fist set
‘of predictive pattems 116 required heavy processing povver
and memory, potentially using multiple servers and signifi
‘cant investments of ime from hours daysto weeks depend
Jing onthe mumber of datapoints utilize.
10031] The creation ofthe second set of predictive pattems
122 should take considerably less time and processing power
‘8 the data points (the fist set of pattems) should be signifi-
‘cantly less. The second set of protictve pattems 122 can be
tenerated using the same or similar forecasting techniques
Nov. 12, 2015
that generated the fist set of predictive pattems 116. That
the fist staf predictive patterns 116 ar fd into the vaio
Torecasting techniques 1 generate a higher level of forcast
resulls, which is then fed into the predictive pattern distibu-
‘ion engine 120 to generate the second set of predictive pat-
tems 122. The second set of predictive paterns 122 ca be
stored in the predictive pattems repository 118 for later
retrieval. It shold be noted thatthe forecasting technique(s)
don'thaveto be the same ones utilized tcteate the frst set of
predictive patters 116 in order to ereste the second set of
predictive pattems 122. The creation of the second set of
predictive pattems 122 can be done on the ly or as requested.
Additionally, the second set of prodictive pattems 122 can be
‘genenited using Use predictive patter profiles eagine 114
loos) Aa
patterns repository 118, The
‘granular level pattems 124 ean include information abot
purchased (male, female, parent, ee), what item (candy.
paper towels, etc), where store, division region, ec), when
ute, ec.) how (cash, credit, check, et)
[0033
‘and second predictive patterns 116, 122 can be positioned in
the predictive patters repository 118 using various position-
ing andor hierarchy schemes depending on the desired infor-
‘mation, The granular level pattems 124 can be stored hierar
The granular level pattems 124 and any of the ist
chically for example, going from top to bottom, by county,
by region, by store, by department, by product, by product
code, te. Tis allows for the desired predictive pate to be
easly identified, etieved and manipulated or easily drilled
down a desired. The granular level patterns 124 may be
positioned horizontally for additional manipulation
[034] Once the granular level patteras 124 are generated
(cither previously or upon request), they are available for use
to create requested forecasts At this point, an executive 126
‘ean request a prediction as to the numberof Parent's Choice
(Giapees), a Wal-Mart product that willbe sold inthe second
{quarter of 2017 as she is negotiating logistics contracts or
Suppliers’ agroement in order to maintain curent pricing
levels. The ganar level pattems 124 canbe use to provide
the requested forecast for Parent's Choice. The requested
{orecast ean be viewed on a display 128 by the executive 126
andor the forecasting team 102. As many granular level
pattems 124 or combinations thereof can be generated as
equested by the executive 126 oF the forecasting team 102
Additionally, the granular level patterns 124 ean be generated
relatively quickly so tat business decisions ean be made i a
timely manner.
[0035] FIG. 2illastates « percentage data hourly profile
200 according to embodiment of the invention. Profile 200
ithisttates an example of pereentage data for hourly pofileby
ay of the Week and is also known as historical profile a it
includes historical data. Profile 200, for example, can be
percentage data forcertain typeof eredittranssetion, sles for
& particular item, number of people entering the store, the
amount of electricity used, the number of bags used, or any
‘other subject matter desired by the user. Profile 200 can also
be chanterized various ways, sue as percentage data
related to seasonal, monthly. yearly holiday, daily. and any
other types of profile desired by the user. Profile 200 can be
stored in database 108, 110. Thus, prile 200, for example
‘an include information sich s electricity usage forthe storeUS 2015/0324702 Al
‘or division by each hour, each day, each week, each month,
‘each quarter, each season, each year, and the like-This allows,
‘oumber of data points in onder (0 create aeeurate
ive patterns 116, 122
10036] FIG. 3 illustrates « method 300 that provides pre-