Integrert planlegging i mediebransjen - Respons

Our vision in the Respons project is fully integrated and optimized planning and control of paper based value chains in the media business. Every year, some 500.000 tons of paper are used for printed matter in Norway. Large volumes are transported in connection with distribution, and reverse logistics for control and recycling of unsold products, approximately two million ton-kilometers per day. Today, there is a clear separation between distribution for subscription and non-subscription products. Subscription and non-subscription products are transported separately, in a non-coordinated way. Moreover, the first parts of the supply chain are not optimized. We expect large effects from optimization of the supply chain, coordination of the daily transport of subscription and non-subscription products, and reduction of physical return of unsold products. Good tools for prognosis and management of circulation are vital. We realize our vision by developing user-friendly, web-based tools for automated, optimized planning. The tools are based on novel, branch specific models and solution methods. We utilize cash register data from the retailers and advance statistical methods to provide high quality sales prognoses that will also reduce the transported volumes. The user partners in the Respons consortium are Aftenposten, VG, Amedia Trykk & Distribusjon, and Coop. SINTEF is the research partner, with the Norwegian Computing Center (NCC) as subcontractor. Distribution Innovation (DI) is the project owner, developer of user interfaces, web functions and prototypes. DI is the main channel for exploitation and commercialization. We have prioritized development of the statistical methods for prognosis of product sales, as this is a basic prerequisite for circulation management and better logistics. This work has been much more demanding than anticipated. It has been performed in close cooperation between DI, SINTEF, and NCC. DI and the user partners have established a system for utilization of cash register data, in collaboration with several supermarket and retailer chains. We have developed and assessed two statistical methods: Reduced Rank Regression (RRR) and Exponential Smoothing. The latter is based on the tool that is used by large parts of the industry today, in connection with physical return of unsold products. A comprehensive assessment of the two methods based on historical data shows that RRR gives considerably better prognoses. The assessment also shows that the use of fresh sales data through cash register data improved the prognosis to a large degree. Based on these results, we selected RRR as the method for the prototype. The first version of the prototype was finalized in the autumn of 2014, and testing started in December. As only preliminary testing was done by project finish, we applied for an extension until August 31 2015 that was granted. The statistical prediction model developed in Respons that proposes circulation numbers is based on the use of updated cash register data, but it works also on historical data based on physical return. The model considers annual variations and a number of special days in connection with holidays. From January 2015, DI, VG, and Aftenposten have collaborated closely on testing with real data. The tests revealed several shortcomings that have been remedied, with assistance from NCC and SINTEF. In May, there was a breakthrough, with excellent results for VG. There was more work ahead, as the test results for Aftenposten and small circulation titles were not yet good enough. The model and the cash register methodology were presented at a seminar for Norwegian, Swedish, and Danish logistics managers in spring 2015. According to VG, a 1% improvement on return and sold out percentage will yield yearly savings of NOK 4 million. At project finish in August 2015, DI has implemented a version of the prototype, but is still tuning the model in order to improve accuracy for small volumes. Declining sales is a strong driver for automated solutions. Four MSc candidates at NTNU have written their theses on main topics in Respons. The work has led to 6 published scientific journal papers, 24 scientific talks/reports, and 3 popular science articles. Two book chapters on transportation optimization have recently been published by the Society for Industrial and Applied Mathematics. One is to a large degree based on the work in Respons and its predecessor projects, including information from Aftenposten and DI on the development of transportation optimization tools. In 2015, we have given three talks in well-reputed international scientific conferences, and two papers have been published in prestigious (Norwegian level 2) journals.

I det nylig avsluttede Effekt-prosjektet har DI, brukerbedrifter og SINTEF utviklet en web-basert planleggingsløsning for automatisk konstruksjon og revisjon av optimaliserte budruter. Den dekker i dag vel 80% av det norske markedet, er tatt i bruk også i utlandet, og fører til store gevinster. Løsningen dekker kun planlegging av selve budrutene for abonnementsaviser. Vi ser stort uutnyttet potensial ved mer overgripende planlegging, både langs verdikjeden og ved samvirke mellom distribusjonsselskap på tvers av løssalg og abonnement. Bedre prognoseverktøy, tettere opplagsregulering og optimalisert returlogistikk for løssalgsaviser vil også gi store gevinster. I Respons vil vi utnytte potensialet i viktige planleggingsoppgaver som hittil er uten verktøy støtte. Målet er å realisere gevinster ved integrert planlegging langs verdikjeder fra trykkeri til leser, og på tvers av verdikjedene. Viktige hensyn er produksjons- og lagerstyring på trykkeri, lastekapasitet, sammensetning av flåte for utkjøring, optim al plassering av hentesteder, optimalt valg av trykkeri for produktene. For løssalgsaviser er bedre prognoser, mer dynamisk opplagsregulering og mer optimal returlogistikk viktig for økonomi og miljø. Vi vil lage verktøy som optimaliserer logistikk mellom to eller flere distribusjonsselskap, også ved integrert transport av løssalgs- og abonnementsaviser. De viktigste FoU-utfordringene er modeller og metoder for: - bedre prognoser, opplagsstyring og returlogistikk - integrert planlegging i mediebransjen - produksjons- og lagerstyrt transportplanlegging - optimal sammensetting av kjøretøyflåte Gjennom Effekt har vi et meget godt grunnlag for ytterligere innovasjon. Det består av en web-basert systemarkitektur med gode brukergrensesnitt der en avansert mo tor for transportoptimalisering er integrert i tjeneren. Videre har konstellasjonen av brukere, DI og SINTEF vist evne til god FoU-basert innovasjon med kommersiell spredning og resultatutnyttelse.