The SpendInsight Project

Mark Bishop and Sebastian Danicic

Research by Professor Mark Bishop and Dr Sebastian Danicic carried out at the Department of Computing at Goldsmiths, University of London, together with colleagues at the University of Reading led to the development of a novel spend analysis system: SpendInsight and a spin-off product GreenInsight.

SpendInsight analyses an organisation's purchasing data and then uses artificial intelligence to recommend alternative equivalent cheaper products. GreenInSight uses a similar technology to enable organisations to find `greener' equivalents, thereby reducing their carbon footprint.

Launched in 2011 as a commercial service by KTP industrial partners @UK PLC , SpendInsight has already had a significant impact both on the economy and society.

Impact on the Economy

SpendInsight has so far been used to perform spend analysis for 382 different organisations and although normally confidential data, a number have publicly reported savings; these include the Basingstoke and North Hampshire NHS Foundation Trust (300,000) and the Royal Berkshire NHS Foundation Trust (100,000) .

SpendInsight is listed on the Government's 'Procurement Service'; an agency of the Cabinet Office supporting procurement across the Public Sector, aiming to increase operational and financial efficiency. Thus SpendInsight is available on a 'national framework contract', making it possible for Public Sector organisations to use the service very easily .

The market capitalisation of @UK PLC. increased substantially after it announced that, together with Visa Inc., it was rolling out its CloudBuy platform across Asia Pacific SpendInsight technology to create 'the region's first business-to-business integrated e-marketplace solution'.

In the first weeks of September 2013. Tungsten Corporation sign a five-year agreement to license SpendInsight software, to be branded as 'TungstenAnalytics'. The investment in SpendInsight technology from Tungsten and Visa led to an immediate 500% increase in market capitalisation of @UK PLC. (from 7million to 35million).

Impact on Society

SpendInsight was used to produce the evidence for a report by the National Audit Office (February 2011) which concluded:
"If hospital trusts were to amalgamate small, ad-hoc orders into larger, less frequent ones, rationalise and standardise product choices and strike committed volume deals across multiple trusts, they could make overall savings of at least 500 million, around 10 per cent of the total NHS consumables expenditure of 4.6 billion".

This report was discussed by the House of Commons Committee of Public Accounts in May 2011 which concluded that the Department of Health should require all NHS purchasers and suppliers to make use of a standard, comprehensive product bar-coding system so that price comparisons can easily be made and savings opportunities identified; purchasers must ensure that product bar-coding is in place by April 2014.

The NAO report featured strongly in BBC news and subsequently led to an investigation into health procurement by the BBC radio programme File On 4, an investigation which focussed on the political impact of the report and concluded that - in contrast to conservative manifesto policy decentralising budget-holding and commissioning responsibilities to GPs - money could be saved "if only trusts could get their act together".

(b) GREENINSIGHT In September 2011 the NHS announced that it would work with @UK PLC. on a project using @UK PLC.'s core SpendInsight technology, to develop GreenInsight - a system to measure NHS carbon footprint, reveal which NHS suppliers and goods contribute most to this footprint, and to create a NHS carbon reduction plan.

GreenInsight's significance has been recognised among leading organisations promoting sustainability. Its official launch (on 7th October 2010) was attended by Richard Benyon, Minister for the Natural Environment and Fisheries, and included a presentation from A4S. When the GreenInsight data-format, Green-XML, was subsequently launched on 15th November 2011 there was a video message from the HRH Prince of Wales, an introduction from the RSA's Chief Executive, and a presentation from the CEO of the worldwide benchmark emissions reporting programme, the Carbon Disclosure Project.

The Research

The research which led to the development of the SpendInsight spend analysis system is described in three published papers:
The use of Artificial Intelligence technology in e-commerce spend analysis,
Classification in E-Procurement,
SpendInsight: some remarks on deploying an intelligent spend-analysis system(AISB/IACAP World Congress 2012 - Alan Turing 2012 Symposium on Nature Inspired Computing and Applications)
The main areas of focus of our research were, data re-structuring, de-duplication and product classification.

Data re-structuring:
the problem of attempting to take incomplete differently structured data from many different sources and automatically convert them all into objects of the same well-defined structured schema.

De-duplication:
the problem of deciding when apparently different entities in the structure are all really differently identified representations of the equivalent entity. Central to de-duplication is pattern matching. Equivalent entities although syntactically different often match the same pattern. Our approach to de-duplication resulted in a `cleansed view' produced built by a novel rule-engine. All of the rules are applied iteratively, until the system stabilises. Each rule may use information in both of the staging databases and in the partially built cleansed view (effecting a feedback loop), and the rule may make use of additional custom indices built on these data.

Product Classsification:
from a large set of purely textual descriptions of products and the structured data above, we want to find which products could be used as 'alternatives', i.e. those which are 'semantically' equivalent. Again, semantic equivalence of syntactically distinct objects is the underlying abstraction of the problem at hand.

A novel combination of the decision tree and Bayesian classifier algorithms was introduced to solve our particular problem. The descriptions for each product are reduced to a bag of words and a set of manually classified training data is used to calculate the conditional probability of a word belonging to (a product in) each class. An application of Bayes' Theorem gives the conditional probability of a class given a word; thus words and, in turn, products can be classified.

While, in principle, the Naive Bayes classifier can be applied to the whole data set, in practice the applicability is limited by the training data set. So also we employed a decision tree approach for product classification. In our work the decision trees are taught from the training data by the symbolic rule induction system. Finally, for each product the classification system must decide between the various candidate classes suggested by each of the algorithms. Essentially, there is a trade-off between the accuracy of a method and the proportion of the dataset to which it can be applied.

Links

Goldsmiths' Research involvement: SpendInsight was produced as a result of a KTP project involving @UK PLC Goldsmiths and University of Reading http://www.spendinsight.com/development-team.html
@UK PLC used ground-breaking artificial intelligence developed at Goldsmiths to develop SpendInsight. http://www.spendinsight.com/artificial-intelligence.html
@UK PLC offers a spend analysis service to organisations. The analysis produces a data pack that shows areas for potential savings, and allows organisations to drill down by commodity or service right back to the original purchase order for checking. The analysis is performed using SpendInsight http://www.uk-plc.net/spendanalysis.html
To date (September 2013) @UK have used SpendInsight to analyse spend data for 382 different organisations. Email from Richard Barraclough (Head of Research @UK PLC)
SpendInsight is used to gather data for a National Audit Office report: procurement of consumables by NHS acute and Foundation trusts http://websites.uk-plc.net/UKPLC_SpendInsight/SpendInsight-powers-National-Audit-Office-report-478.htm
The National Audit Office report (produced using Spendinsight) concluded that at least 10 per cent of hospitals' spending on consumables, amounting to some 500 million a year, could be saved if trusts got together to buy products in a more collaborative way. http://www.nao.org.uk/report/the-procurement-of-consumables-by-nhs-acute-and-foundation-trusts/
The House of Commons Committee of Public Accounts debated the National Audit Office report in May 2011 http://www.publications.parliament.uk/pa/cm201012/cmselect/cmpubacc/875/875.pdf
Impact on Health Policy: The House of Commons Committee of Public Accounts concluded that the Department of Health should require all NHS purchasers and suppliers to make use of a standard, comprehensive product bar-coding system so that price comparisons can easily be made and savings opportunities identified. It should ensure that product bar-coding is in place by April 2014. http://www.publications.parliament.uk/pa/cm201012/cmselect/cmpubacc/875/875.pdf (page 5)
Further Impact on Health Policy: The chair of the Commons' Public Accounts Committee, Margaret Hodge MP, reacting to the report, said that - given the NHS in England must find up to £20bn of annual savings by 2015 - the report showed efficiencies could be made. http://www.bbc.co.uk/news/health-12338984
National Audit Office Report discussed on File on 4 on BBC Radio Four
The Programme
Transcript of Programme
Example of Impact on Economy: The Spend Analysis service provided by @UK PLC using spendinsight has already saved the Royal Berkshire NHS Foundation Trust £100,000 a after a single pass. http://assets-production.govstore.service.gov.uk/Giii Attachments/@UK PLC/Bids/SpendInsight Royal Berkshire NHS Foundation Trust Case Study.pdf
Another Example of Impact on Economy: SpendInsight saves money for North Hampshire NHS Foundation Trust. http://www.zoominfo.com/p/Tammie-Purdue/1204748492
Basingstoke and North Hampshire NHS Foundation Trust cut their procurement spend in early 2010, and have already implemented the 300,000 worth of savings they identified using the same technology (SpendInsight) that frames the National Audit Office's report
http://www.uk-plc.net/news/newsItem.asp?lngID=479&lngCategory=1
http://www.publictechnology.net/sector/nhs-health/procurement-nhs-trusts-poor-value-money-says-nao-report
http://alumni.reading.ac.uk/page.aspx?pid=782
@UK PLC launch GreenInsight, a system which classifies items purchased by organisations to reduce their carbon footprint. http://www.uk-plc.net/green-marketplace-launch-7th-october.html
GreenInsight uses the same technology as SpendInsight developed on the KTP project by Goldsmiths. Presentation SpendInsight and GreenInsight by Dr. Richard Barraclough, Head of Spend Analysis,
Impact on the Environment GreenInsight used by Basingstoke and North Hampshire NHS Foundation Trust for reducing their carbon footprint.It highlights surprising results: 27% of their carbon footprint arises from the purchase of medical precision instruments, almost as much as electricity at 30%. Presentation GreenInsight in Practice by Brian Pinnell, Strategic Procurement Leader, Basingstoke and North Hampshire NHS Foundation Trust
Richard Benyon Parliamentary Under-Secretary of State at the Department for the Environment made a speech at the Business and Accounting Software Developers Association) in which he stated ``@UK PLC have utilised ground breaking research by University of Reading, Goldsmiths University of London and the Stockholm Environment Institute at York to create @UK PLC's e2class product database.'' http://www.uk-plc.net/green-marketplace-launch-7th-october.html
Impact on the Economy In August 2013 market capitalisation of firm increased 500% from £7million to £35million in little over a week ahead of news that (a) Tungsten Corporation plc and (b) Visa are going to invest into @UK plc to obtain access to SpendInsight technology. http://www.londonstockexchange.com/exchange/news/market-news/market-news-detail.html?announcementId=11698873 http://www.londonstockexchange.com/exchange/news/market-news/market-news-detail.html?announcementId=11700864
@UK seals five-year licensing deal with Tungsten. Under the five year rolling licence agreement, @UK's Spend Analysis software will enable `TungstenAnalytics' to be delivered across this global network. The total maximum value to @UK over the initial five year period is 3.4m. http://www.stockmarketwire.com/article/4662713/UK-inks-five-year-licensing-deal-with-Tungsten.html
@UK's share price rockets on rollout with Visa to Asia Pacific http://www.proactiveinvestors.co.uk/companies/market_reports/60807/market-movers-uk-north-river-resources-stratmin-global-resources-concha-blavod-wines-spirits-tullow-0000.html

http://www.ifamagazine.com/news/uk-shares-soar-on-rollout-with-visa-to-asia-pacific-282397

@UK PLC ATUK Contract Win - NHS Expenditure Carbon Footprint http://www.bloomberg.com/apps/news?pid=newsarchive&sid=aIyD2F7jYTw4
Unlocking the SME funding market and the power of spend analytics https://ctmfile.com/story/unlocking-the-sme-funding-market-and-the-power-of-spend-analytics#.UmFHO5Sidbk
Conditional Licence Agreement for Spend Analysis http://investor.uk-plc.net/news/conditional-licence-agreement-for-spend-analysis.html
Tungsten moves higher as institutions see appeal of e-invoices http://www.proactiveinvestors.co.uk/companies/news/62204/tungsten-moves-higher-as-institutions-see-appeal-of-e-invoices-62204.html
Tungsten AIM float a success(The Independent) http://www.independent.co.uk/i/matrix/the-business-matrix-thursday-17-october-2013-8884870.html
Picture TFSC_Tungsten_OB10.png
Picture at Bank Station IMG_0030.JPG
Papers
The use of Artificial Intelligence technology in e-commerce spend analysis,
Classification in E-Procurement,
SpendInsight: some remarks on deploying an intelligent spend-analysis system(AISB/IACAP World Congress 2012 - Alan Turing 2012 Symposium on Nature Inspired Computing and Applications)



s.danicic@gold.ac.uk
Sebastian Danicic BSc MSc PhD (Reader in Computer Science)
Dept of Computing, Goldsmiths, University of London, London SE14 6NW
Last updated 2013-10-20