Confi on figu guratio ration n and Usage Usage of Demand Demand Sensing February, 2016
Objectiv bjective es Slid lide e
At the end of this lesson, you you will be able to: • Expl Explai ain n the the deman demand d sens sensin ing g funct functio iona nalility ty with within in SAP SAP Inte Integr grat ated ed Business Business Planning Planning for demand 6.1 in detail. detail. • List List the the key key feat featur ures es of shor shortt-te term rm dema demand nd fore foreca cast stin ing g • Unde Unders rstan tand d the the prer prereq equi uisi site tess for for shor short-t t-ter erm m deman demand d forec forecas asti ting ng
Agen Ag end da
•
Create Create a forecas forecastt model for demand demand sensing sensing
•
Manage snapshot snapshot configurati configuration on
•
Load Load hist histor oric ic snap snapsho shott data data and and sche schedul dule e snaps snapshot hot jobs jobs
•
Run demand sensing from the SAP IBP add-in for f or Microsoft Excel
Cr eat e a Fo r ec as t Mo d el f o r Dem an d Sen s i n g
Create Crea te a Ne New w Foreca For ecast st Model fo r Demand Demand Se Sens nsin ing g
Set Set up a new new fo recast recast model for the demand demand sensin g algorithm via the Manage Manage Forecast Forecast Models app .
On the next next scree screen, click click „ Crea reate“ te“ in the lowe lower right right corn corne er and sele select ct your your plann planning ing area rea from from the the appe ppearing ring list. list.
Create Crea te a Ne New w Foreca For ecast st Model fo r Demand Demand Se Sens nsin ing g Time Settings
Manda ndatory tory:: Periodi riodicit city y nee needs to be set set to „ Daily“ ily“ for for the dema demand nd sens sensing ing algorit lgorithm hm
The The peri period odss are are ente entere red d in day dayss when when the the peri period odic icit ityy is set set to Dail Dailyy. Historical Periods: As best practice from past customer projects, we recomm end using 52 or more weeks as historic data input. This showed very good go od demand sensing results. Foreca Forecast st Period Periods: s: This This depends depends on the busine business ss case. case. Demand Demand sensi sensing ng is usually usually used used in a forecast forecasting ing rang range e of 4-8 4-8 week weeks. s. Th er er e i s a mi mi ni ni mu mu m o f h is is to to ri ri ca cal p er er io io ds ds t ha hat n ee eed s t o b e u se sed i n or or de der t o ex pe pec t g oo oo d r es es ul ul ts ts . This This de depends pends on your your overa overall ll settings. settings. The UI will give give recomme recommendat ndations ions upon saving saving the model model w he hen t he he n um um be ber yo u en te ter ed ed i s t oo oo s ma mal l. l.
Create Crea te a Ne New w Foreca For ecast st Model fo r Demand Demand Se Sens nsin ing g Prepr Pr eproce ocess ssing ing Ste Step p – Pr Promo omotion tion Sal Sales es Lift Eli Elimin minati ation on (opt (option ional) al) As of SAP Integrated Business Planning Planning for demand 6.1, you you will find a new preprocessin preprocessing g step called called „Promotion „Promotion Sales Lift Eliminatio Elimination n “. This ste step p nee needs to be used used in combin combina ation tion with Demand mand Se Sensing nsing (Full) ull) in case case you you have have promo promotio tiona nall sale sales s lift lift inform informa ation tion in your your sy syste stem. m.
Prep Prepro roces cessi sing ng steps steps othe otherr than than Promo Promoti tion on Sales Sales Lift Lift Elimi Elimina nati tion on do not work work with with Demand Sensing Sensing algorithms. algorithms. The Promo Promotio tion n Sales Sales Lift Lift Elimin Eliminati ation on prep preproc rocessi essing ng step step is not limite limited d to Demand Demand Sens Sensin ing g in its its usag usage. e. It can can for for examp example le also also run run stan standa dalo lone ne..
Create Crea te a Ne New w Foreca For ecast st Model fo r Demand Demand Se Sens nsin ing g Prep Pr epro roce cess ssin ing g St Step ep – Co Confi nfigu gura rati tion on of Pr Prom omoti otion on Sal Sales es Li Lift ft Eli Elimi mina nati tion on
system is to multiply the range of accepted values Outlier utlier Multiplie Multiplier: r: A decimal number by which the system when the variance test is used, thus including additional values in the range or excluding a set of values from it. Key Fig Figur ure e that that cont contai ains ns the Sales Sales Hist Histor oryy including the the sale saless lift lift from from the the prom promot otio ions ns Sales Sales History: History: Key (tota (total) l).. For Dema Demand nd Sensin Sensing, g, this this is usua usuallllyy the Conf Confir irme med d Quan Quantit tityy or the Deli Delive vere red d Quan Quantit tityy. Figure re that that conta contains ins the the planne planned d prom promoti otion on upli uplifts fts or downl downlift iftss in the the histo history ry Pla Planned nned Sale Sales s Lifts: Lifts: Key Figu and future. future. In case case of SAP6, SAP6, this is PROMO PROMOSPLI SPLIT T („Pro („Promotio motion n Uplift Uplift“) “) Figure that contai contains ns the Consens Consensus us Fore Forecas castt excluding the the sales ales lift lift from from Consensus Forecast : Key Figure promotions! promotions! In case case of SAP6, this this is CONSENSU CONSENSUSDEMA SDEMAND ND („Consensus („Consensus Demand w/o Promotions“ Promotions“)) Output ut Key Key Figu Figure re in whic which h the the resu results lts shoul should d be stor stored. ed. This This would would provi provide de the Sales Sales Save Save Results Results In: Outp Hist Histor ory y exlud exludin ing g the the promo promotio tions ns
Excurse: Promotion Promotion Sale les s Lift Elimination Pre Preprocessing processing Step Identify ou tli ers (sale (sales s li fts) associated associated with promot ions, and to remove them them fro m the sales sales hist ory Ste tep p1
Iden Identi tifi fies es outl outlie iers rs with within in the the sale sales s hist histor ory y with with Outl Outlie ierr Dete Detect ctio ion n Logi Logic. c. Rang Range e for for outl outlie ierr dete detect ctio ion n is calc calcul ulat ated ed base based d on Vari Varian ance ce Test Test.. It detect detectss both both posit positive ive and and negati negative ve outlier outliers. s. Result: Result: Outl Outlier ier Perio Periods ds
Ste tep p2
Identifies positive sale sales s lift lifts, s, caus caused ed by prom promot otio ions ns in the the sale sales s hist histor ory y wher where e promo promoti tion on quanti quantity ty is great greater er than than zero. zero. Result Result:: Promoti Promotion on Periods Periods
Ste tep p3
Sear Search ches es for for corr correl elat atio ions ns to iden identtify ify the the peri period ods s when when the the outl outlie ierrs were were actual actually ly caused caused by promo promoti tions ons
Ste tep p4
Eli Elimina minate tess the the valu values es from from the the sale sales s hist histor ory y wher where e a corr correl elat atio ion n was was foun found d betw betwee een n outl outlie ierr and and prom promot otiion peri period ods s
Ste tep p5
Look Looks s for for clos closes estt matc match h to adju adjust st valu values es prop propor orti tion onal allly, in case case no exac exactt corr correl elat atio ion n can can be dete detect cted ed
Ste tep p6
Stor Stores es the the resu resullts for for sale sales s hist histor ory y, clea cleans nsed ed by the the sale sales s lift lifts s in the the outp output ut key key figure
Create Crea te a Ne New w Forecast Model for fo r Demand Demand Sensing Sensi ng (Full ) Forecasting Forecastin g Steps (1)
Main Main Inp Input ut for for Fore Foreca cast stin ing g Step Step:: Key Key Figur Figure e that that cont contai ains ns the the data data that that acts acts as the the main ain sale sales s hist histor ory y. For For Dema Demand nd Sens Sensin ing, g, thi this is usua usualllly y the the Conf Confir irme med d Quan Quanti tity ty or the the Deli Delive vere red d Quantity. Note Note tha thatt the the „M „Mai ain n Input Input for for Fore Foreca cast stin ing g Step Steps“ s“ nee needs ds to be the the sam sames as •
Either her one one of the the input nputss for for Histo storic rical Or Orde dere red d Quan uantity tity or Deli Delive vere red d Qua Quanti ntity for the the Demand Sensing (Full) Model.
•
„Inp „Input ut for for Algo Algori rith thm“ m“ on on the the Prep Prepro roce cess ssin ing g Step Step for for Prom Promot otio ion n Sale Sales s Lift Lift Elim Elimin inat atio ion n
Excur xcurse: se: De Demand Se Sensin nsing g – Full and Update Update Runs Runs Dema Demand nd Sens Sensin ing g can be run in full full and and upda update te mode modess. A Ful Fu l l Run – • • •
Is norm normal ally ly sched chedul uled ed at the the begi beginn nniing of a week week to get get the the regr regres esssion weig weight hts. s. In most most case cases, s, a ful fulll run run in dema demand nd sens sensin ing g woul would d foll follow ow an an upda update ted d week weekly ly midmidterm term forec forecas astt and, and, subs subseq equen uentl tlyy, a snap snapsho shott that that has been been taken taken base based d on that that.. Thes These e outpu outputs ts are are then then used used in the the fina finall sense sensed d dema demand nd calc calcul ulat atio ion. n.
An Upd ate at e Run – • •
•
Is sched chedul uled ed when whenev ever er ther there e are are delt delta a chang hanges es in the input nputss (new (new open open orde orders rs/sh /ship ipme ment nts) s) duri during ng the the week week.. Thus Thus,, it take takess into into acco accoun untt any any new new info informa rmati tion on for for order orders/ s/sh ship ipmen ments ts duri during ng the the week week and and then then calc calcul ulat ates es the the sens sensed ed deman demand d base based d on the the regr regres essi sion on weig weight htss obt obtained as part of the full run that was don done at the beginning of the week. An update run cannot be scheduled until and unless a full run is already done.
Create Crea te a Ne New w Forecast Model for f or De Demand mand Sensing Sensi ng (Full ) Forecasting Forecastin g Steps (2) Add the Demand Sensing (Full) algorithm to the model
Plea lease refe referr to the nami naming ng conve conventions ntions for certa certain in key key figure figures s t ha hat ar e u se sed as i np np ut ut f or or d em em an an d sensing. This This is descri describe bed d in deta detail il in the SAP6 SAP6 Planni Planning ng Model Model slide deck.
Create Crea te a Ne New w Forecast Model for f or De Demand mand Sensing Sensi ng (Full ) Fore Fo reca cast stin ing g St Step eps s (3 (3)) – Ke Keyy Fig Figur ures es Consensus Forecast: Need Needs s to be in syn sync c with with the the
same same entry entry on the Prep Preproce rocessi ssing ng Step Step for Promot Promotio ion n Sales Lift Eliminatio Elimination. n. In SAP6, this is CONSENSUSDEMAND (“Consensus (“Consensus Demand w/o Promotions”). This key figure should not contain promotion information!
Future Future Ordered Ordered Quantity: Quantity:
Key Key Fig Figure ure that that contai contains ns the open open orde orders rs for the future future horizon horizons. s. In SAP6, SAP6, this is FUTUREORDE FUTUREORDEREDQT REDQTY Y (“Future Ordered Quantity”)
Figure that that contains contains Historical Historical Ordered Ordered Quantity: Quantity: Key Figure the Sales Sales Histo History. ry. Usually Usually,, this this is the Confi Confirme rmed d Quanti Quantity. ty. In SAP6, SAP6, this this is CONFQTY (“Confirmed (“Confirmed Quantity”)
Delivere Delivered d Quantity: Quantity: Key Key Figure Figure that that contai contains ns the deliv delivere ered d quan quantit titie ies s from from the past past weeks weeks.. In SAP6, SAP6, this this is ADJDELIVQTY (“Delivered Qty Adjusted”).
Create Crea te a Ne New w Forecast Model for f or De Demand mand Sensing Sensi ng (Full ) Fore Fo reca cast stin ing g St Step eps s (4 (4)) – Fu Furt rthe herr Settin Settings gs Consensus Demand Demand Snapshot Suffi x:
Need Needs s to matc match h the the suff suffix ix that that is used used when when creati creating ng the foreca forecast st snapsh snapshot ot (see (see chapte chapterr 2)
Bias Horizon: Bias Bias in SAP´s Demand Demand Sensin Sensing g can
be bette betterr defin defined ed as the consum consumpti ption on or order orderin ing g patte pattern rnss of the custom customers ers.. It deter determin mines es or predi predicts cts what what the future future consu consumpt mptio ion n looks looks like like,, base based d on calcul calculate ated d weig weights hts from from the past past bucke buckets. ts. Bias Bias Horizo Horizon n 1|2|4|6 1|2|4|6|8| |8|9 9 as in the scree screensh nshot ot is an exampl example e and and varie varies s depe depend nding ing on your your indi individ vidua uall data. data. Maxim Maximum um is is 10 as the high highes estt bias bias hori horizon zon entry. entry. The The last 2 entry entry fields fields can also also be be left blank blank in in case case that that is need needed ed for for the the spec specif ific ic data data set. set.
These four values are are used for capping capping the final sensed demand numbers. See next slides for more information. These thresholds are checking the result of the Demand Sensing Run (before putting uplifts back) against the Consensus Forecast without promotions.
Daily Daily Average Calculation Horizon:
Rolling window (number of weeks in the past) over which we average average daily shipments shipments for every Monday, Tuesday Tuesday,…, ,…, Sunday Sunday of the weeks. weeks. In this this case, case, for for the last 4 weeks weeks..
Thre hreshold shold Settings
The maxim maximum um forec forecast ast incr increas ease e / decrea decrease se setti settings ngs are used used to cap the demand demand in case case of excep excepti tions ons in the data data and smooth smooth the result results. s. The user can define define absolut absolute e and percenta percentage ge values values.. •
The threshold settings cap the outcome of DS before the order adjustment.
•
It is a global threshold for all planning o bjects that are planned with the forecast mod el, and thus has to be set se t very carefully. carefully.
•
In case you have items with very ve ry different different volumes, one might think of o f creating different sensing profiles to avoid setting too aggressive or too limiting limiting thresholds. Example : • • • •
Max Forecast Increase Increase (absolute value): 50 PC (MAX_INCREASE) (MAX_INCREASE) Max Forecast Increase Increase (percentage (percentage value): 30% (MAX_PCT_I (MAX_PCT_INCREASE) NCREASE) The consensus demand is 100 PC (CONSENSUSDEM (CONSENSUSDEMAND) AND) The The sensed sensed demand demand that that would ould be calcul calculate ated d withou ithoutt takin taking g the foreca forecast st incr increas ease e factor factorss into into accoun accountt is 300 PC
UPPER_BOUND = Max (100+50 , 100*1.3) => Max (150,130) = 150 PC So, if the final sensed demand numbers calculated during the sensed demand run are higher than 150, then they are capped. Sensed Demand (300 PC) PC) > UPPER_BOUND UPPER_BOUND (150 PC) => Sensed Demand = 150 PC
Examp Exa mple: le: Dail Daily y Average Average Ca Calc lcul ulation ation Horizon
The Daily Average Calculation Horizon Horizon is a rolling window (number (nu mber of weeks in the past) pa st) over which the demand sensing averages daily shipments to determine the daily shipment profile. In this case, for 4 weeks into the past, the shipments for each day of the week are averaged. For last four Mondays, Mond ays, Tuesdays etc. This in turn is used u sed in calculating the shipment sh ipment profile of each day d ay.. Lower numbers numbe rs than 4 weeks will will influence the resu lts in a negative way. way. We recommend recomme nd using a value equal or higher than 4. The default default value is therefore therefore set to 4.
Create Crea te a Ne New w Forecast Model for f or De Demand mand Sensing Sensi ng (Full ) Fore Fo reca cast stin ing g St Step eps s (5 (5)) – Wo Work rkda days ys
Select Select Workd ays: You can control the calculation of the
sensed demand with the workday selection. By this, the system will not determine any quantities from the consensus forecast for the unmarked days. However, in case there are quantities booked for these days as part of sales orders (confirmed qty), these will still show up as sensed demand later on.
Create Crea te a Ne New w Forecast Model for f or De Demand mand Sensing Sensi ng (Full ) Forec Fo recas asting ting Ste Steps ps (6) – Mi Minim nimum um Data Data Poin Points ts and and WMAP WMAPE E Thre Thresh shold old Baseline WMAPE threshold:
Minimum Data Points:
It is the minimum number number of weeks of historical data (forecast, open orders, deliveries) that the demand sensing algorithm needs to recommend reliable short term forecasting results. 15 week weekss are are a recomme recommend nded ed value value base based d on the the experien experience ce from past projects. projects.
The system system calcula calculates tes the WMAPE (Weighted (Weighted Mean Absolute Percentage Error) results results for the Consensus Forecast Forecast (based (based on the the snapshot snapshot input). input). For For certain certain planning planning objects (product/location/customer (product/location/customer), ), the consen consensus sus forec forecast ast might might alre alread ady y compu compute te the best best forecast. This This thresh threshold old value value is there therefor fore e used used to deter determin mine e whether whether or not the Demand Demand Sensing Sensing algorith algorithm m could could impr improve ove the the resu result lts s or not. not. So if the calculated WMAPE for the Consensus Consensus Forecast Forecast is less below this threshold, then the demand sensing algo will not run and the Consensus Consensus Forecast values are are taken. The calcul calculatio ation n is done done per indivi individua duall plannin planning g object. object. The WMAPE calcula calculation tion results results are not stored. stored.
Create Crea te a New New Forecast Model for f or De Demand mand Se Sens nsin ing g (Update) Forecasting Steps
Add the demand sensing (update) algorithm to the model
Create Crea te a Ne New w Foreca For ecast st Model fo r Demand Demand Se Sens nsin ing g Postprocessing Postproces sing Steps
Error Error meas measure ures s are are not calcula calculate ted d during during the dema demand nd sensin sensing g run
Manage Snapshot Configuration
Introduct ion – Wha hatt is a Sna napshot? pshot? A snapshot is the current picture picture of any time key figure figure for any time time range. In case of demand sensing, sensing, it is for the the mid-term forecast forecast (e.g. consensus consensus demand) A forecast snapshot is needed by the demand planners – •
to review how the forecast changed over a period of time
•
to learn the patterns of demand dem and and correlate forecast with demand for different lags
Historical snapshots are needed by the demand sensing algorithm in the calculation of the sensed sensed demand demand regres regression sion metric metrics. s. Note Note that that it is mandator mandatoryy to load load histor historic ic snapsho snapshott data. data. It is a mand manda atory tory inpu inputt fact factor or to the the dema demand nd sens sensin ing g algorithms . Historical forecast snapshots are loaded manually before the first demand sensing run. After the first run, the snapshot snapshot crea creation nee needs to be schedule scheduled d on a regula regular r Microsoft Excel (e.g. weekly or monthly, monthly, basis via the SAP IBP add-in for Microsoft depending how often your mid-term demand plan is generated)
Generate Ge nerate Sna Snapsh pshot ot Key Figure Figur e fo forr demand sensing sensin g
A new n ew snap s nap sh o t key k ey fi f i gu re is i s auto au to mat ic all y gener g ener ated f or you yo u r p lan ni ng area when a snapshot configuration configuration is saved saved (see (see next slides)
This This snap snapsho shott key figur figure e •
stor stores es the the snap snapsh shot ot valu values es when when a snap snapsh shot ot is take taken n on a regu regula larr basi basis. s.
•
is used used to save ave the the his histor torical ical snaps napsho hott value aluess duri during ng the ini initial tial sys system set up. up.
Not Note that that his histor toric snaps napsho hott value aluess are are need needed ed by the dema demand nd sensi ensing ng algo algorrithm ithm to calc calcul ulat ate e high high-q -qua ualility ty resu result lts. s. So a certa certain in snap snapsho shott hist history ory need needss to be load loaded ed duri during ng the the init initia iall sy system stem setu setup p (see (see chap chapte terr 3) and and whil while e prod produc ucti tive vely ly usin using g dema demand nd sensi sensing, ng, the the snap snapsho shott crea creati tion on needs needs to be sc sche hedul duled ed regu regula larl rlyy.
Generate Ge nerate Sna Snapsh pshot ot Key Figure Figur e fo forr demand sensing sensin g
Prod Produc ucti tive ve use use of dema demand nd sens sensin ing g
Time
Load Load hist histor oric ic snap snapsh shot ot data data for for a cert certai ain n his histori oric time timefr fram ame e at this his poin pointt and and up to this this poin pointt
Generate Genera te sna snapsh pshots ots reg regula ularly rly (e.g. weekly weekly)) via the SAP IBP add-in for for Microsoft Excel
Manage Ma nage Snapshot Snapshot Confi Configur gura ati tion on (1 (1))
Create Create a new snapshot sn apshot key fig ure via vi a the Manage Manage Snapsh Snapshot ot Configu Conf igu ratio n UI: UI:
On the next next sc reen, reen, select select yo ur pl anning area via the dropd ow n list .
Manage Ma nage Snapshot Snapshot Confi Configur guration ation (2 (2a) a)
Mandatory: Snapsho Snapshop p Type = Change History REV
Suff Suffix ix:: Make Make sure sure it is the the same same suffix suffix that that is also also entered entered in the Forecast Forecast Model
Mandator Mandatory: y: Use the same plan planni ning ng level level on which which the inpu inputt key figur figure e is defin defined ed.. In this case: LOCPRODCUSTWEEKLY (Location | Product Product | Customer | Weekly level)
Forecast Model: Manda Mandator tory:s y:sele elect ct the key figur figure e that that is defin defined ed as the Conse Consensu nsuss Forecast Forecast in your your Forecast Forecast Model Model (see (see slide slide 13) e.g. consensus consensus demand demand without without promotio promotions ns or statistica statisticall forecast forecast qty
Manage Ma nage Snapshot Snapshot Confi Configur guration ation (2 (2b) b) How to calcu calcula late te the „ To“ To“ time timeframe frame::
The The amo amoun untt of data data that that need needs s to be capt captur ured ed with within in the the snapsh snapshots ots for for deman demand d sensin sensing g depe depend nds s on certai certain n facto factors. rs. The The follow followin ing g formul formula a should should be used used to calcu calculat late e the snapsh snapshot ot futur future e hori horizo zon n that that is need needed ed for deman demand d sensin sensing: g: To = [SensingHorizon + Snapshot Snapshot Frequency -1] -1] * 7
--------------------------weeks s shoul should d repre represen sentt the the numbe numberr that that Sensing Sensing Horizon Horizon in week REV
you entered entered in the Forecast Forecast Model Model at „Forecast „Forecast Periods“ Periods“ for your your respe respecti ctive ve deman demand d sensin sensing g model: model:
In this this case, case, 42 days days woul would d transl translate ate to 6 weeks. weeks. --------------------------This This is is the the rang range e of peri period odss that that the the snap snapsh shot ot shou should ld be take taken n for. for. These numbers numbers represent represent days. days. From From = 0 mean meanss that that the the snap snapsh shot ot will will be take taken n alwa always ys from from the the current current day, the snapsho snapshott run is executed executed,, onwards. onwards. E.g. if the snapsh snapshot ot is sched schedul uled ed every every Monda Monday, y, the data data inclu include ded d in the the snapsh snapshot ot starts starts from from that that respe respecti ctive ve Monda Mondayy onwa onwards rds.. To = 63 mean meanss that that the the snap snapsh shot ot will will be be take taken n for for the the next next 63 day dayss (9 weeks) weeks) in the future. future.
Snapshot Frequency in weeks weeks depends depends on whether whether you want want to take take the snapsh snapshots ots week weekly, ly, monthl monthly, y, etc. for week weekly ly snapsh snapshots ots use 1, bi-w bi-wee eekl kly y use 2, for for monthl monthly y snapsh snapshots ots use 4 as input. input. The The frequ frequen ency cy depe depend nds s on how how often often your your midmidterm foreca forecast st (e.g. (e.g. consensus consensus demand) demand) is being being updated updated.. At least one one snapshot should be taken every four weeks!
Example: Sensing Sensing Horizon Horizon = 6 weeks weeks Snapshot Frequency Frequency = monthly To = [6 + 4 - 1] * 7 = 63 days days
Manage Ma nage Snapshot Snapshot Confi Configur gura ati tion on (3 (3))
Once you save save the snapsho snapshott configur configura ation, tion, a new new key key figure figure will be gene genera rate ted d for the the respe respecti ctive ve planni planning ng area rea.
Note tha that the the plann plannin ing g area rea is set set to „ Inact nactiv ive e“ now, now, so you nee need to rea reactiv ctiva ate it. it.
Loa Lo ad Histo is tori rica call Snaps napsho hott Dat a an d Sche ch ed u l e Sn ap s h o t Jobs
Hist istori orical cal Snapshot Snapshot Upload Prerequisites It is mandatory mandatory to h ave a Snapshot Snapshot Key Figure as as defined i n c hapter 2 and and at least least on e data data point loaded for thi s snapshot key figur e to be able able to run demand demand s ensing at all. Please Please not e that that hig h qualitativ e results i n demand demand sensin g can only b e achieved achieved wit h enough histo ric snapsho t inf ormation lo aded aded (we recommend 52 snapshot snapshot revisions - see chapter chapter 2 for more information) Procedure: •
Create a new snapshot key figure for fo r the planning area
•
Check and adjust mandatory settings in the planning area
•
Pay attention to prerequisites during data upload of master data and key figures
•
Load Stati Statist stica icall Forecast Forecast qty qty/ Consensus Consensus Demand Demand Plan Plan qty for the past past period periodss that that you want ant to use use as his historic orical al per periods iods in your our forec orecas astt mode modell.
•
Eith Either er load load Stat Statis isti tica call Forec Forecast ast qty qty / Consen Consensu sus s Demand Demand Plan Plan qty qty for for the the futu future re (at (at lea least st one one data data poin pointt is need needed ed)) or gene genera rate te it via via a sta stati tist stic ical al fore foreca cast st run. run.
•
Load the historical forecast snapshots before the first demand sensing run
Mandatory Se Setti ttings ngs in Planning Area Change History
Change History: must be “Planning Area Enabled” • Prerequisite for demand sensing as the latter requires a snapshot key figure • Snapshot key figures can only be defined for planning areas with change history enabled
Prerequis Pre requis it ites es for Ma Mast ste er Data Upl Upload oad •
Make sure to load the master data in the correct order.
•
First, the data for the master m aster data of type ”simple”, then ”compound”, then ”reference” (if available), e.g., in SAP6:
MDT Si m p l e
MDT Co m p o u n d
1
Time Profile
2
Location, Product, Customer
3
LocationProduct (o (optional)
4
Currency (optional)
5
Exchange Rate (optional), Currency Curre ncy To To
6
UoM To
7
UoM Conversion
8
Sales Order
9
Delivery
10
Promotions
11
Key Figur Figure e Data Data (e.g (e.g., ., Cons Consens ensus us Deman Demand, d, Deliv Delivere ered d Quant Quantit ityy, etc) etc) Important: many key figures are now stored in technical weeks. If you upload data for such key figures you should either upload the data also in technical weeks or you upload it aggregated ag gregated by the time, e.g. on calendar calenda r week or monthly level, by using the new “Time Profile Level” within uploading (new with IBP 6.1):
12
Hist Histor oriical cal Snap Snapsh shot ot Data Data for for Cons Consen ensu suss Dema Demand nd Qua Quant ntiity
Upload the th e His Histor toric ical al Snapsho Snapshott via v ia the WebU WebUII
Crea reate a new new impor importt job of data data type type „ Snapsh napshots ots““ to uploa upload your your histor historica icall snapsh snapshot ot file. file.
Sampl Sa mple e Snapshot Snapshot Fil ile e
Forecasted periods within revision (14 weeks weeks in this this examp example) le)
Foreca Forecaste sted d QTY QTY
Revision of April 20, 2015
Revision of April 27, 2015
How Ma Many Histori Histo ri cal Snapshot Snapshot s Should Shoul d Be Loaded? Loaded? The The amo amount of weekly snapsh pshot history to be loaded shou hould be equa qual to the „Hist „Historic orical al Period Periods“ s“ value value (in weeks!) weeks!) that that you entere entered d in the time sett setting ings s of your forecast model. E.g. E.g. If the the „His „Histo tori rica call Per Perio iods ds““ val value ue is 280 280 in in your our fore foreca cast st mode model, l, that that woul would d be 40 week weeks s and and in this this case case,, weekl eeklyy snap snapsh shot otss for for the the last last 40 week weekss shou should ld be load loaded ed.. Note Note that that the the qual qualiity of the the resul esults ts will will decr decrea ease se in case case you load load less less snap snapsh shot ot data data (in (in this this case case less less than than 40 week weeks) s).. Load Loadin ing g more more data data will will not not prod produc uce e bett better er the the resu result lts! s!
Troubleshooting
Error: Cannot import data because period & planning object data doesn’t exist for the selected revision, Invalid Revision for Planning Object or Period. Solution: This means the UoM (Unit of Measure) templates are not imported impo rted or imported in the wrong order. So, reload all master ma ster data types types that have ‘Attributes as key figures’ and then reimport the snapshots (see the the “Prerequisites during during Data Upload“ slide). slide). In the SAP6 template, there are two “Attributes as key figures” used for UoM and currency conversion:
UOMCONVFACTOR EXCHANGERATE
Set up Ba Set B atc tch h Job Jo b to Automaticall Autom aticall y Create Create New New Snapsho Snapshots ts for Fut uture ure Period Periods s (1) (1) Schedule Schedule th e snapshot op eration eration to r un automatically v ia the SAP SAP IBP add-in add-in for Micro soft Excel Excel
Set up Ba Set B atc tch h Job Jo b to Automaticall Autom aticall y Create Create New New Snapsho Snapshots ts for the Future Future Pe Periods (2)
Run Demand mand Sensin ns ing g from fr om t h e SA SA P IB P ad d -i n f o r Micro ic roso soft ft Exc xce el
Demand Sensing Run
Demand Sensing can be run in two modes: •
Batch mode (as a background background job)
•
Simulation mode (directly in MS Excel)
Run Demand Demand Sensing Sensi ng i n Batch B atch Mode(1a) Mode(1a)
Log on to yo ur respective planning area in the SAP SAP IBP add-in add-in for Micr osof t Excel xcel and crea create a “ daily” daily” or a “ weekly” weekly” view.
Note: Note: Dema Demand nd sensing is alwa lways run run on prod produc uctt – location location – custome customerr leve level. l. So in orde orderr to see see the the dema demand nd sensin sensing g foreca forecast st model, model, you nee need to sele select ct the the data data on this this level.
Run Demand Demand Sensing Sensi ng i n Batch B atch Mode(1a) Mode(1a)
Optional st ep: Filter based on any attribut e (e. (e.g. g. Produ Produ ct ID, ID, Produ Produ ct Group, etc). Note: ote: The The dema demand nd sensi sensing ng algorit lgorithm hm will will then then only only run on the the filte filtere red d subse subsett of data data..
Run Demand Demand Sensing Sensi ng i n Batch B atch Mode(1a) Mode(1a)
Pick an an appropr iate “ Unit of Measure” Measure” for th e conversion of th ese key figur es
Run Demand Demand Sensing Sensi ng i n Batch B atch Mode(2a) Mode(2a)
Start Start th e full demand sensing job after after pic king the forecast model that incl udes your demand demand sensin sensin g algor algor ithm
Monitor Job Status
Run Demand Demand Sensi Sensi ng in Sim imul ulation ation Mode Mode(1 (1b) b)
Log on to yo ur respective planning area in the SAP SAP IBP add-in add-in for Micr osof t Excel Excel and create create a “ daily” view. Note: ote: The The dema demand nd sensi sensing ng algorit lgorithm hm will will only only run on the the filte filtere red d subse subsett of data data in case case filte filters were were applie pplied. d. Simulation of DS can only happen on a “Daily” View since the DS algorithms algorithms are always always set to a periodicity of ‘Daily’ in the forecast forecast model model (Slide (Slide 12)
Run Demand Demand Sensi Sensing ng in i n Simul ati on Mode (2 (2b)
Run demand demand sensing fr om the ‘Simulate’ drop dow n after after pic king the forecast model that includes your demand demand sensing algorith m.
Demand De mand Sensi ng Run (3) (3)
In comparison to th e mid-to lon g-term statisti cal algorithm s, there there are some diff erences erences for demand sensing (full and update) algorith ms.
Demand Sensing Models…
are run run on on Loca Locati tion on – Produc Productt - Custom Customer er level level only à only visible in the forecast model list if the corresponding IDs are selected attributes are run for the baseline version version only à the other versions are disabled and baseline is preselected automatically Time Period is set se t to daily da ily..
Demand De mand Sensing Sensin g (Full) Outpu ts (1 (1))
Demand emand sensing output s show up on ce the job is completed completed successfully and the data is refreshed refreshed fr om th e databa database se (Click (Click “ Refresh” Refresh” ).
Run Demand Demand Se Sens nsin ing g (Update) 1
Situation: Future ordered ordered quantiti es (Current (Current Open Open Order Order QTY QTY)) were modif ied in the middle middle of a ce certain rtain wee week. In this exampl xample e, ne new orde orders came came in during during the middl middle e of the wee week and the the curre current nt open open orde order qua quantity ntity was was incre increa ased sed from 340 to 600 pie pieces. ces.
Run Demand Demand Se Sens nsin ing g (Update) 2
Re-run Re-run demand sensing and observe the new new output s. In such a case, case, we woul d choos e the dema demand nd sensing sensing (update update)) algorithm algorithm
Pick Pick the demand demand sensin sensing g update update forec forecast ast model model that that you you creat created ed earli earlier er
Lesson Summary
You should now be able to • Expl Explai ain n the the deman demand d sens sensin ing g funct functio ional nalit ityy with withiin SAP SAP Inte Integr grat ated ed Busine Business ss Planni Planning ng for demand demand 6.1 in detail detail.. • List List the the key key feat featur ures es of shor shortt term term dema demand nd fore foreca cast stin ing g • Under Underst stand and the the prer prereq equi uisi site tess for for shor shortt term term deman demand d forec forecas astin ting g
© 20 2015 15 SAP SE SE or an SAP aff affilil iat iate e comp co mpany any.. Alll r i g h t s r es Al eser erved ved.. No part of this publication may be reproduced or transmitted in any form or for any purpose without the express permission of SAP SE or an SAP affiliate company. SAP and other SAP products and services mentioned herein as well as their respective logos are trademarks or registered trademarks of SAP SE (or an SAP affiliate company) in Germany and other countries. Please see http://global12.sa http://global12.sap.com/corpor p.com/corporate-en/legal/co ate-en/legal/copyright/index.ep pyright/index.epx x for additional trademark information and notices. Some software products marketed by SAP SE and its distributors contain proprietary software components of other software vendors. National product specifications may vary. These materials are provided by SAP SE or an SAP aff iliate company for informational purposes only, without representation or warranty of any kind, and SAP SE or its aff iliated companies shall not be liable for errors or omissions with respect to the materials. The only warranties for SAP SE or SAP affiliate company products and services are those that are set forth in the express warranty statements accompanying such products and services, if any. Nothing he rein should be construed as constituting an add itional warranty. In particular, SAP SE or its affi liated companies have no obligation to pursue any course of business outlined in this d ocument or any related presentation, or to develop or release any functionality mentioned therein. This document, or any related presentation, and SAP SE’s or its affiliated companies’ strategy and possible future developments, products, and/or platform directions and functionality are all sub ject to change and may be changed by SAP SE or its affiliated companies at any time for any reason without notice. The information in this document is n ot a commitment, promise, or legal obligation to deliver any material, code, or f unctionality. All forward-looking statements statements are subject to various risks and uncertainties that could cause actual results to differ materially from expectations. Readers are cautioned not to p lace undue reliance on th ese forward-looking statements, which speak only as of their dates, and they should not be relied upon in making purchasing decisions.
