Data sharing in industrial ecosystems (continuation)

by Prof. Dr. Boris Otto, Prof. Dr. Niko­laus Mohr, Mat­thi­as Rog­gen­dorf, PhD and Tobi­as Gug­gen­ber­ger

Key chal­len­ges in exch­an­ging data in indus­tri­al eco­sys­tems

Data eco­sys­tems obvious­ly have clear bene­fits, so what is hol­ding com­pa­nies back from taking full advan­ta­ge of them? The rea­sons are mani­fold. Some invol­ve orga­niz­a­tio­nal, tech­ni­cal, or legal bar­ri­ers or restric­tions, while others are fue­led by typi­cal ways of thin­king and expe­ri­ence-based con­cerns. They can be divi­ded into four major cate­go­ries:

Chal­len­ges rela­ted to cul­tu­re and mind­set:

  1. New value oppor­tu­nities depend on sharing data in lar­ger indus­tri­al eco­sys­tems that may invol­ve par­ties with limi­ted trust in one ano­t­her. Par­ti­ci­pants might be able to ana­ly­ze shared data to deri­ve con­fi­den­ti­al busi­ness infor­ma­ti­on. Con­si­de­ring every even­tua­li­ty is dif­fi­cult. For examp­le, it seems rea­son­ab­le to share wel­ding posi­ti­on data so that ano­t­her par­ty can opti­mi­ze pro­duc­tion pathways, but a com­pe­ti­tor could poten­ti­al­ly use that data to ana­ly­ze the resul­ting shapes – and deter­mi­ne in advan­ce what new pro­ducts the first com­pa­ny is plan­ning. Typi­cal­ly, howe­ver, data in iso­la­ti­on rare­ly crea­tes a com­pe­ti­ti­ve advan­ta­ge.
  2. Ano­t­her major rea­son that com­pa­nies do not enga­ge in data sharing models is the per­cei­ved lack of con­trol when data lea­ves their pre­mi­ses. For examp­le, soft­ware licen­sing has reached a high level of matu­ri­ty (in terms of both con­trac­ting and tech­ni­cal mea­su­res) but data sov­er­eig­n­ty is still hard to gua­ran­tee and imple­ment.
  3. The need to iden­ti­fy enough value to jus­ti­fy invest­ment can also be an obsta­cle. Some orga­niz­a­ti­ons have dif­fi­cul­ty brea­king out of exis­ting thin­king pat­terns, while others lack the capa­bi­li­ties to run the kind of struc­tu­red assess­ment nee­ded to dri­ve enga­ge­ment in data eco­sys­tems. Often, this requi­res a col­la­bo­ra­ti­ve give-and-take gover­nan­ce model at odds with tra­di­tio­nal buy­er-sup­plier-rela­ti­ons­hips.

Chal­len­ges rela­ted to the need for onto­lo­gies:

  1. Data can only be shared if it can be easi­ly inter­pre­ted and quick­ly inte­gra­ted with other data sources. It must be unam­bi­guous­ly unders­tood by all eco­sys­tem mem­bers. Thus, a com­mon lan­guage or onto­lo­gy is requi­red all par­ties agree on. In some indus­tries, such as ban­king or health­ca­re, such data lan­guages were estab­lis­hed deca­des ago to enab­le inter­ope­ra­bi­li­ty and the lever­aging of stan­dard soft­ware. In the indus­tri­al sec­tor, howe­ver, indi­vi­du­al onto­lo­gies and stan­dards exist – such as eCl@ss and the Asset Admi­nis­tra­ti­on Shell, to name just two examp­les. What is nee­ded though is a com­pre­hen­si­ve archi­tec­tu­re of stan­dards which inclu­des dif­fe­rent types of data such as mas­ter data, refe­rence data, manu­fac­tu­ring and sup­ply chain event data etc. This archi­tec­tu­re stack can­not be deve­lo­ped by an indi­vi­du­al eco­sys­tem mem­ber, but requi­res con­sen­sus wit­hin a com­mu­ni­ty of prac­ti­ce.
  2. An indus­tri­al data eco­sys­tem pro­mi­sing a signi­fi­cant com­pe­ti­ti­ve advan­ta­ge could be a strong impe­tus for com­pa­nies to prag­ma­ti­cal­ly defi­ne stan­dards for sharing data bey­ond sin­gle-pur­po­se cases and allow other par­ties to inter­pret it cor­rect­ly. Gene­ric onto­lo­gies for exch­an­ging sen­sor and robo­tics data that can be easi­ly exten­ded are on the rise due to advan­ces in con­nec­ted devices and the need for inter­ope­ra­bi­li­ty in hete­ro­ge­ne­ous envi­ron­ments. Howe­ver, indus­tri­al envi­ron­ments are often a mix of older sys­tems and new IoT-based sen­sors, making it even har­der to reach agree­ment on com­mon stan­dards.

Tech­no­lo­gi­cal chal­len­ges to designing and mas­te­ring the requi­red plat­form and ser­vices:

  1. Even when com­pa­nies under­stand the value of sharing data in indus­tri­al eco­sys­tems, they often lack the skills to make their data avail­ab­le in an effec­ti­ve and effi­ci­ent way. The chal­len­ge starts with inter­nal tech­ni­cal inter­faces for sup­ply­ing data and con­ti­nues with deve­lo­ping an envi­ron­ment to pro­cess and inte­gra­te data from dif­fe­rent sources and run models using advan­ced ana­ly­tics methods. If com­pa­nies deci­de to build new pro­ducts based on such data, they also need the capa­bi­li­ties to pro­duc­tio­n­i­ze the resul­ting models and ensu­re effec­ti­ve com­mer­cial and tech­ni­cal ope­ra­ti­ons, inclu­ding gua­ran­te­eing ser­vice levels and sca­ling to poten­ti­al­ly mil­li­ons of cus­to­mers. The­se capa­bi­li­ties typi­cal­ly do not exist in engi­nee­ring-focu­sed indus­tri­al cli­ents, so they need to be sys­te­ma­ti­cal­ly built eit­her as new enti­ties or with exter­nal help.

Chal­len­ges in effec­tively mana­ging data to make time­ly, auto­ma­ted access to it pos­si­ble:

  1. The last chal­len­ge rela­tes to the inter­nal capa­bi­li­ties requi­red to pro­vi­de data from inter­nal app­li­ca­ti­ons that fits its inten­ded pur­po­se, is of suf­fi­ci­ent qua­li­ty, and is avail­ab­le in a time­ly fashion. A lot of data sits in siloes and is not even acces­si­ble. Other data may be used for its pri­ma­ry pur­po­se (for examp­le, on the shop floor) but not for any others. Also, many orga­niz­a­ti­ons, espe­cial­ly in the indus­tri­al space, have limi­ted capa­bi­li­ties for mana­ging their data assets stra­te­gi­cal­ly. As a result, they have trou­ble dis­tin­guis­hing bet­ween data that can be easi­ly shared and stra­te­gic data that poses signi­fi­cant oppor­tu­nities but ent­ails the risk of expo­sing intel­lec­tu­al pro­per­ty to untrus­ted enti­ties. Befo­re the­se com­pa­nies par­ti­ci­pa­te in a data eco­sys­tem, the need to estab­lish basic data gover­nan­ce princi­ples and sys­te­ma­ti­cal­ly deve­lop their employees’ data liter­acy.

So far, only niche solu­ti­ons in high­ly selec­ti­ve seg­ments have been deve­lo­ped (see boxes for examp­les). In many cases the­se data sharing eco­sys­tems were crea­ted by a domi­nant play­er with the power to man­da­te that other mar­ket par­ti­ci­pants join the net­work.

Examp­le 1: Luft­han­sa AVIATAR

Lufthansa’s AVIATAR is a plat­form for the avia­ti­on indus­try. It inte­gra­tes data from sources on air­line ope­ra­ti­ons, air­craft, main­ten­an­ce sys­tems, and more to crea­te a com­pre­hen­si­ve tool for fleet manage­ment. As it works inde­pendent­ly of orga­niz­a­tio­nal or tech­ni­cal bounda­ries, it can be used to opti­mi­ze all ope­ra­ti­ons of any ima­gin­ab­le con­fi­gu­ra­ti­on of fleets.

AVIATAR focu­ses on digi­ti­zing the value chain, from pre­dic­ti­ve main­ten­an­ce to auto­ma­ted ful­fill­ment and on to com­ple­te digi­tal main­ten­an­ce, repairs, and ope­ra­ti­ons. Add-ons can pro­vi­de com­ple­men­ta­ry fea­tures such as lay­o­ver moni­to­ring, lay­o­ver sourcing, and staf­fing. Final­ly, mar­ket­pla­ces for loa­ning and exch­an­ging com­pon­ents and tools sup­port eco­sys­tem-wide opti­miz­a­ti­on.

Case stu­dy 2: 365FarmNet

365FarmNet is an inno­va­ti­ve soft­ware plat­form for manu­fac­tu­rer-inde­pen­dent farm manage­ment. The ser­vice inte­gra­tes mul­ti­ple part­ners and data sources to crea­te a com­pre­hen­si­ve manage­ment tool. Its modu­lar archi­tec­tu­re com­bi­nes free-to-use func­tions (such as data manage­ment, wea­ther ana­ly­tics, docu­men­ta­ti­on, and basic plan­ning tools) with com­ple­men­ta­ry value-added ser­vices (for tele­ma­tics, rou­te opti­miz­a­ti­on, soil sam­pling, and more). The open inter­face makes it pos­si­ble to inte­gra­te intel­li­gent apps from a varie­ty of agri­cul­tu­ral sup­pliers, such as machine­ry manu­fac­tu­rers or pesti­ci­de and fer­ti­li­zer pro­du­cers, crea­ting a data eco­sys­tem around the plat­form. The approach demons­tra­tes that brin­ging an inte­gra­ted value pro­po­si­ti­on to cus­to­mers requi­res sharing data across sec­tors.

To the final part.