Management Decision

OPTICAL CHARACTER RECOGNITION: A Study of Success and Failure in Innovation

Andrew Robertson,
Article information:
To cite this document:
Andrew Robertson, (1971) "OPTICAL CHARACTER RECOGNITION: A Study of Success and Failure
in Innovation", Management Decision, Vol. 9 Issue: 3, pp.213-223, https://doi.org/10.1108/eb000971
Permanent link to this document:
https://doi.org/10.1108/eb000971
Downloaded on: 06 December 2017, At: 08:22 (PT)
Downloaded by Universiti Teknologi MARA At 08:22 06 December 2017 (PT)

References: this document contains references to 0 other documents.

To copy this document: permissions@emeraldinsight.com
The fulltext of this document has been downloaded 229 times since 2006*
Users who downloaded this article also downloaded:
(2014),"Undergraduate research support with optical character recognition apps", Reference Services
Review, Vol. 42 Iss 2 pp. 336-350 <a href="https://doi.org/10.1108/RSR-09-2013-0045">https://
doi.org/10.1108/RSR-09-2013-0045</a>
(2012),"Open source optical character recognition for historical research", Journal of Documentation,
Vol. 68 Iss 5 pp. 659-683 <a href="https://doi.org/10.1108/00220411211256021">https://
doi.org/10.1108/00220411211256021</a>

Access to this document was granted through an Emerald subscription provided by emerald-
srm:434496 []
For Authors
If you would like to write for this, or any other Emerald publication, then please use our Emerald
for Authors service information about how to choose which publication to write for and submission
guidelines are available for all. Please visit www.emeraldinsight.com/authors for more information.
About Emerald www.emeraldinsight.com
Emerald is a global publisher linking research and practice to the benefit of society. The company
manages a portfolio of more than 290 journals and over 2,350 books and book series volumes, as well
as providing an extensive range of online products and additional customer resources and services.
Emerald is both COUNTER 4 and TRANSFER compliant. The organization is a partner of the
Committee on Publication Ethics (COPE) and also works with Portico and the LOCKSS initiative for
digital archive preservation.

*Related content and download information correct at time of download.

 OPTICAL CHARACTER
RECOGNITION
A Study of Success and Failure in Innovation
by Andrew Robertson

The idea of optical character recognition (OCR), in other words

the "reading" of documents by other than human means, arose as
a practical proposition during the Second World War. Wartime
experience of using computers in the United States had revealed
the contrasts in speeds between the transcription of documents to
Downloaded by Universiti Teknologi MARA At 08:22 06 December 2017 (PT)

be processed (at that time the punching of cards or tape by

operatives working from original documents) and the central pro-
cessing within the computer itself. Visual output was also slower
than central processing but was much speeded up by the intro-
duction of line printers and later of xerography.
This "paired" case study, part of a project sponsored by the
Science Research Council to examine patterns of success and
failure in industrial innovation, is confined to two attempts to
innovate in the field of OCR. There were others, one or two of
which were contemporary, most of which have followed, have a
much more recent history and may be thought to have overtaken,
in terms of market penetration, the innovation here designated a
commercial success. The point of this study when it was under-
taken was to extract data about the two innovations that would be
suitable for general analysis by a computer programme designed to
search out significant groups of explanatory factors so that the
characteristics associated with innovative success might be
recognised as typical within an industry, or perhaps generally. This
study belongs to one of two groups, the instrument industry, the
other group investigated being chemical manufacturing.

The pair of innovations chosen for comparison comprised the Solartron

Electronic Reading Automaton (ERA), design of which was begun in
1954 and which was withdrawn from production in 1963, and the
Farrington Scandex, first manufactured in 1956 and since replaced by
more sophisticated equipment. Up to the withdrawal of ERA, therefore,
the chronologies are more or less within the same time span, although
the former innovation took place in Britain and the latter in the United
States, with the Farrington device benefiting from earlier work at
Intelligent Machines Research. Technically the two devices make use of
different systems, although both were directed at the same kind of
marketdocument processing for computer input in large companies
handling vast quantities of standard paperwork.

Vol. 9 Winter 1971 213

 Optical Character Recognition

There are three basic groups of character recognition methods: mark

reading and sensing relies simply on the presence or absence of a mark
or marks; magnetic ink character recognition (MICR) relies on the
presence of a magnetic field caused by particles in the ink; and OCR.
The subdivisions of OCR all rely on some kind of light sensitive scan-
ning of the character to be read. The simplest forms are slit and divided
slit scanning, in which either the narrow slit or the document are
traversed to give a vertical projection of the horizontal character.
Mechnical scanning uses a perforated disc similar to that used in early
television transmission. Flying spot scanning replaces the disc with a
cathode ray tube which also acts as the light source. The photo-cell
matrix may be regarded as an attempt to reproduce the action of the
retina of the eye, and is not a scanning system in the same sense as
these three.
The Scandex instrument was a simple reader of the mechanical type,
Downloaded by Universiti Teknologi MARA At 08:22 06 December 2017 (PT)

designed to perform the function of reading the symbols 0 to 9 in a

specially designed font. It was, thus, single font numeric only, in
contrast to the much more ambitious multifont alpha-numeric ERA
which used the flying spot technique. Even today, with much more
experience to assist them, manufacturers are still confined to a limited
range of special fonts, although not only letters and numbers can be
read but also special symbols such as plus and minus signs.
To provide a framework of understanding of the difficulties involved
in innovation in this specialised field of instrument design a few words
should be added about the unavoidable constraints of the method.
Apart from character design (font) these include paper quality (flat
finish, low gloss, dead white and not too absorbent), print quality
(sharpness, accuracy, evenness), ink (black, light fast) and handling
(distortions, unwanted marks, scuffing and smearing). These condi-
tions refer only to the input to the instrument, of course, but there is a
relationship between input quality and instrument performance that
finally determines the reject and error rate of the reader. To this may be
added the factor of complexity of input.

The IMR/Farrington Scandex

The contrast between the Solartron ERA and the IMR/Farrington
Scandex lies not only in the difference in technique but also in the
objectives. The designer of Scandex was David Shepard, the founder of
Intelligent Machines Research. To him this was a bread-and-butter
instrument produced at the request of the Farrington Manufacturing
Company and he chose to use, mechanical scanning because of its
relative simplicity and robustness. The objective put to him by Farring-
ton was the reading of petrol station credit dockets upon which an
embossed plate (the Charga-Plate) had imprinted a numerical code. In
addition to designing the Scandex Shepard designed a font (Selfchek),
which he calculated would suffer least from rejects because of misuse
of either Charga-Plate or imprinter.
It is relevant to mention that the Farrington Manufacturing Company
had developed a large business supplying embossed plates or credit

214 Management Decision

 Optical Character Recognition

cards made of metal, together with imprinters, to some of the large oil
companies at a time when the use of credit cards in place of cash was
increasing rapidly at petrol stations, and was being encouraged. The
company had then been asked to solve the data processing problem
that had arisen with the greatly increased volume of paperwork.
Farrington had no facilities with which to provide a solution (their
R & D department was 15 strong and there was no optical work done)
but the work of Shepard and IMR had become known even though,
at that stage, his first reading machine had not been installed. He had
launched IMR in 1950 with a partner, Harvey Cook, having left the
National Security Agency to do so. It was at the NSA that he had
conceived the idea of replacing card and tape punching with some form
of direct transcription, having become familiar with the problem of
input to computers. He did not, however, deliver his first machine till
1956, to Reader's Digest (this was a large, specially designed three-line
Downloaded by Universiti Teknologi MARA At 08:22 06 December 2017 (PT)

reader, using the scanning disc system) and had accepted the Farring-
ton commission alongside this more complex work.
The significant aspects of this outline are that Farrington was
"pushed" into providing a scanner by customers (the first Scandex
installation was for Standard Oil of California in Salt Lake City) and
relayed this request to Shepard with clearly defined specifications. Their
approach was made in 1954 and by 1956 the first delivery had been
made. Three years later the two companies were merged and Shepard
became the largest shareholder in Farrington and a member of the
Board of Directors.

The Solartron ERA Project

The contrast with Solartron could not very well be stronger. Farrington
was a much older company, founded in 1904 to exploit its founder's
patents for snap-lid spectacle cases and jewel boxes and developing
into a diversified light engineering company between the wars.
Solartron was founded in 1947 by two former apprentices, E. R.
Ponsford and L. B. Copestick, to make electronic testing instruments,
registered in the following year and delivered its first products in 1950,
laboratory amplifiers. Most of the work of the young company was done
under government contract. All profits were ploughed back for develop-
ment, but new capital was supplied in 1951 by John Bolton, who
became a director and for a brief time the majority shareholder.
It was on the initiative of John Bolton that the ERA project was
undertaken. The company had experience of making oscilloscopes and
was therefore familiar with cathode ray tube technology. In 1952 a
research worker at University College, F. Roberts, published a paper in
the Proceedings of the Institution of Electrical Engineers (see refer-
ences) describing the work he and Professor J. Z. Young, F.R.S., had
done with the flying-spot microscope on counting, sizing and sorting
of microscopic particles. Using a combination of microscope and
cathode ray tube it was possible, said the article "to translate a two-
dimensional density distribution, such as a photographic plate, into a
varying voltage-time relation suitable for electrical transmission" and

Vol. 9 Winter 1971 215

 Optical Character Recognition

therefore a flying-spot microscope could convert the density distribu-

tion of microscopic objects into voltage-time series. In paragraph 1.6.
of their article Roberts and Young suggested the replacement of card
indexes (punched cards) by microfilm as memory stores for computers,
the film being read into the memory drum or tape by flying spot.
This paper was given in the Radio Section of the I.E.E. Proceedings
and was seen by I.E.E. members of the staff of Solartron. Mr. Frank
Roberts was consulted about the future of computer input equipment
of this type, reported favourably and was asked to design a prototype.
The prime mover in this development was John Bolton, who was then
vice-chairman of the company. An accountant and a graduate of the
Harvard Business School, with some wartime experience in Naval
electronics, Bolton could see the potential in ideas such as this. Also it
was part of a system of "sector analysis" that he had introduced into
the company's forward thinking as a means of planning new products
Downloaded by Universiti Teknologi MARA At 08:22 06 December 2017 (PT)

for growth markets.

In 1954 the Solartron Electronic Group of companies consisted of
Solartron Laboratory Instruments Ltd., the original manufacturing unit,
with a turnover of 342,000 a year, Solartron Engineering Ltd., with
about a tenth of this business and Solartron R & D Ltd., earning no
more than 5,000 a year. The total employment was 400, so that
output a head of personnel was less than 1,000 a year. There was
strong internal opposition to the ERA project, particularly on the part of
the research staff headed by the then research director, R. Catherall,
but John Bolton persisted. Indeed, in the recollection of his fellow
director, John Crosse, the ERA project was mainly his initiative.
In Bolton's own description of the events at that time the opposition
to his project led to the isolation of Roberts and his assistant George
Norris, recruited from University College. Indeed, at the start of the
project Roberts tended to work at home, otherwise it was feared that
the R & D people would have restricted his use of workshop facilities.
In 1956 a new company, Solarton Electronic Business Machines
Ltd. (SEBM) was formed, under the managing directorship of
Bowman Scott, a graduate electrical engineer with a Harvard M.B.A.
and a background of Post Office and Army telecommunications work
and management consultancy. By this time an ERA prototype was in
existence and was demonstrated to potential customers and to the
technical press in the summer of that year.
Roberts had been sent to the United States in 1954, ostensibly to
carry out "sector analysis" in the country of the future, but according to
Crosse had already made up his mind about flying spot OCR, which
was one of his main recommendations on his return, having met and
talked with David Shepard, then working on his first commercial design.
In the first two years of the ERA project there was some exchange of
ideas if not of actual know-how between Shepard's IMR and Solartron,
but it ceased after the Farrington merger if not earlier.
Some of the characteristics of the Solartron Electronic Group are
worth noting as relevant to the ERA innovation. John Bolton and his
wife owned 60% of the ordinary shares at the beginning of the project,

216 Management Decision

 Optical Character Recognition

the company was growing very fast (too fast, some of the former
directors now consider), at an average annual rate of 50%, new
ventures were generally pursued by setting up new subsidiary com-
panies, such as SEBM, and the group was scattered geographically
between Kingston-on-Thames, Thames Ditton and Dorking, which
meant that central control was not very tight. The "new venture"
approach also meant that the managing directors of subsidiaries had a
high degree of autonomy and could not have their resources "milked"
by the parent company after annual allocation. This constitution of the
group helped to protect the ERA project from its opponents even when
its prospects had become doubtful. Bowman Scott was Bolton's
recruit, originally in charge of personnel for the group and personal
assistant to Bolton until SEBM was added to his responsibilities. Both
of them were enthusiastic about the project's potentialities, but
looking back they think that they miscalculated the time factor. "We
Downloaded by Universiti Teknologi MARA At 08:22 06 December 2017 (PT)

were thinking in terms of years when we should have been thinking in

decades." In fact the cost of ERA was calculated on a five-year DCF
present value formula, which worked out at about 50,000 a year. The
final cost of the project was nearer to 500,000 over the years 1956-
63 with a gross revenue of less than 100,000 for the sale of three
machines.

The Market for ERA

The first potential customer for ERA was Boots Pure Drug Company, as
it was then called, of Nottingham. The financial manager of Boots,
Denis Greensmith, had seen ERA demonstrated and it had occurred to
him that such an instrument could be designed to read cash register till
rolls. Boots had some 2,500 retail shops and stores in the United
Kingdom at that time all serviced for central accounting from Notting-
ham. Greensmith's idea was to use ERA to scan the till rolls and feed
the data into a computer (the Emidec 1100 was on order for Boots)
programmed to update central ledgers, reorder, invoice and analyse
sales and stocks. Accordingly the SEBM team began work on a reading
machine which could transcribe the print-out of standard cash
registers.
This work was begun in 1956, two years after Roberts had started his
experimental designs. The prototype shown to the public had been
discarded by that time and Roberts himself had left Solartron leaving
George Norrie in charge of the project. It is worth noting that up to this
time the design work on ERA had not been directed to a specific
object (unlike either the "one off" reader for Readers' Digest or the
Scandex) but to the more general aim of providing a versatile docu-
ment reader that might be acceptable to any large user of standard
documentation. One market that Solartron (or rather the ERA pro-
tagonists) had in mind was the clearing banks, but shortly after the
ERA project got under way the London banks decided in favour of
MICR. Their only hesitation was about the font, the eventual favourite
being E13B, the Burroughs design, which is still in use on British
cheque forms. David Shepard was more fortunate in the sense that

Vol. 9 Winter 1971 217

 Optical Character Recognition

some of the large banks in the United States opted for OCR without
proceeding by way of MICR, but for reading travellers' cheques not
ordinary cheques.
Also in the middle to late 1950s mark sensing became a practical
possibility for certain kinds of order form (J. Lyons & Co. Ltd. had
developed an instrument called the Lector, for use within the company
but also for sale outside) and similar types of simple documentation. It
would have been easy for SEBM to manufacture a mark senser as a
"bread and butter" line, but the idea was barely tolerated, everyone
connected with ERA being convinced that success was not far away.
Peter Curry, now managing director of Unitech Ltd., was appointed
sales manager for SEBM shortly after the order had been received from
Boots, and he recalls his surprise at the apparent youth and inexperi-
ence of Norrie, the newly appointed project leader. One of Curry's early
Downloaded by Universiti Teknologi MARA At 08:22 06 December 2017 (PT)

suggestions to Norrie was that the character font in use at Solartron

should be offered to other companies as a standard, thus making it
possible to use ERA, the circuitry of which was designed to read that
particular font. Nome's reply was that if this were to happen he would
not want to keep the existing font. Asked for the characteristics of an
ideal font for OCR he could not give a definition, having, like his
colleagues, too little experience of OCR.
Eventually Curry obtained orders for three more instruments, from
Littlewoods (football pools and mail order stores), Montague Burton
(mass production clothing) and Domestic Electronic Rentals (tele-
vision hire and service). The R & D teams on ERA were now growing
quite large, Norrie concentrating on the Boots machine, with Peter Hall
leading the Littlewoods design team and John Carroll the team making
the other two.
Of these four the Boots machine was the most complex and the most
expensive (priced at 40,000, its cost was probably more than five
times this figure). An early problem was print quality on the till rolls.
Greensmith had consulted Regna, Norwegian cash register manufac-
turers, to see if they could improve on the quality of the printing by the
existing NCR machines with a different design of register. NCR were
also consulted as were Bell Punch. Eventually the Boots machine was
returned, and SEBM was allowed to keep the money for further
development work, because Boots shelved its plans for an integrated
data processing system in Nottingham.
A further instance of acceptance difficulties arose with the Little-
woods machine. An electronics engineer, Neville Groome, was sent to
Liverpool to service the installation, and stayed so long that he
ultimately joined the Littlewoods staff. This machine was restricted in
use to scanning pools coupons, which meant that only three symbols,
1 for a home win, 2 for an away win and 0 for a drawn match, had to
be recognised. These symbols were hand lettered by pools "investors"
whose circles were irregular and not centred in the squares provided.
The reject rate was so high that 0 was replaced by X on the coupons.
This was a much simpler task than that demanded of Scandex, but
here there was no standard font.

218 Management Decision

 Optical Character Recognition

When the Boots installation was under discussion a reject rate of

one character per million was requested but Solartron suggested 10 per
million. Even this was much too low a figure. Shepard's reject factor,
which the oil companies accepted as realistic, was 10-15 per cent of
documents, the best achievement being of the order of 10 per cent of
which half the rejects were caused by hand lettering in place of
imprinting when a customer had forgotten his credit card and the
attendant remembered his serial number. Where an oil company did not
follow Shepard's guidance, rejects could rise as high as 70 per cent
(this was always a matter of adjustment of the four factors mentioned
aboveprint, paper, ink and handling).

Comparison w i t h Farrington
Another contrast between the Solartron and the Farrington businesses
was that the former sold their machines outright (the Littlewoods
Downloaded by Universiti Teknologi MARA At 08:22 06 December 2017 (PT)

Table I
SOLARTRONFINANCIAL PERFORMANCE 1950-1965

Net profit Development

Deliveries Exports Assets after tax Write-off
Year Personnel 000 000 000 000 000

1950 18 13 8
1951 22 20 12 1
1952 66 34 34 1
1953 110 90 74 3
1954 240 152 10 226 5 10

1955 400 399 20 420 4 49

1956 550 758 80 654 12 88

1957 600 1,005 186 902 3 151

1958 830 1,434 335 1,344 31 192

1959 1.330 2,107 560 2,155 5 206

1960 1,550 2,633 800 2,811 8 180

1961 1,953 4,891 1,100 3,143 (984) 974

1962 1,735 3,765 1,027 3,473 (632) 381

1963 1,329 3,398 813 3,078 4 238

1964 1,486 3,634 1,686 3,175 (159) 320

1965 1,526 4,050 2,236 3,746 (136) 335

Figures are taken to June for 1950-1960 and to 31 st December thereafter. Figures for 1961 relate
to the 18 month period June 1960-December 1961. All previously capitalised development costs
were written off in this 18 month period.
Source: Company records. ( ) = loss.

Vol. 9 Winter 1971 219

 Optical Character Recognition

version was priced at 25,000) while the latter rented them ($1,700 a
month, which included a monthly service and running check). Before
the merger Farrington had been paying IMR $20,000 for each instru-
ment. The advantage of rental is, obviously, the cash flow, and
Solartron found itself constantly short of cash for financing the ERA
project. (See Table I). Besides this, Farrington had the additional
advantage of the card/imprinter business, while SEBM relied solely on
"one off" sales of a complex device which Curry found difficult to sell.
It was an expensive and complicated product, and its adoption by a
company would mean the scrapping of their existing investment in
card or tape punches, verifiers, reader/sorters and other hardware, such
as tabulating machines or accounting machines keyed to a punch
mechanism. Companies were also unconvinced of the accuracy of
OCR compared with key punching.
Both in the U.S. and in Britain there was marked reluctance on the
Downloaded by Universiti Teknologi MARA At 08:22 06 December 2017 (PT)

part of the computer "main frame" makers to countenance support for a

device that would do away with the punched card from which an
important part of their profits came. International Computers and
Tabulators (later ICL) had been approached by Solartron for support
but came to no arrangement with them. In the U.S., IBM were inter-
ested, as were a number of other firms (an OCR pioneer not involved in
this case study was Jack Rabinow, whose company is now merged into
Control Data Corporation). Shepard had registered patents from the
beginning of his work, the first being filed in 1953, issued in 1959 but
contested in the courts for 10 years, the lawsuit being settled finally in
his favour in 1970. This was a key patent and strong enough to compel
IBM to pay a lump sum to Shepard for a licence to use the know-how
in their 1400 series of scanners, since replaced with designs not
requiring the use of Shepard's original patent. In fact, two senior IBM
executives joined the Farrington Board, P. F. McCloskey (president) and
J. F. Benton (vice-president), but have since left.
Apart from the Scandex in all its variants (there was a series of im-
provements in design and performance, but the basic mechanical
system was adhered to) Shepard continued to make larger mechanical
devices for large users like American Telephone and Telegraph, First City
National Bank, Nabisco, Ohio Bell Telephone and Atlantic City Electric,
a two pronged approach to the market for readers that had been
explicitly rejected by SEBM. The lack of revenue and the mounting
costs of the R & D (the total investment in ERA has been estimated to
have been of the order of 500 000) forced Solartron to seek outside
financial help, and John Bolton, who had raised capital for the group in
the past, turned to the Firth Cleveland group for help. This company
was a holding company, mainly involved in television retail and
accustomed to a rapid and high rate of return on its investments. In
December, 1959 (the same year that Farrington acquired IMR) Firth
Cleveland by means of a share exchange and cash settlement obtained
53 per cent of the Solartron equity (not SEBM, but the whole group).
Bolton explains the group's predicament in these terms: "It was already
apparent in March, 1959, that our original forecasts or 'targets' for the

220 Management Decision

 Vol. 9 Winter 1971 221
Table II

Schedule of Deliveries 1954/1955 to 1957/1958 And Targets Through 1962/1963

( I n 000's)
Downloaded by Universiti Teknologi MARA At 08:22 06 December 2017 (PT)

Product Group Actual Possible Targets for Next 5 Years

1954/ 1955/ 1956/ 1957/ 1958/ 1959/ 1960/ 1961/ 1962/

1955 1956 1957 1958 1959 1960 1961 1962 1963

Solartron Laboratory Instruments Ltd.

Standard instruments . ........... 300 589 741 945 1,400 2,000 2,750 3,500 4,500
Government and outside contracts........ 42 63 37 57 60 100 150 250 250
Solartron Engineering Ltd.
Government and outside contracts........ 35 87 115 96 60 100 125 150 200
Solartron Research and Development Ltd.
Government and outside contracts........ 5 10 44 75 100 100 125 125 150
Data processing............ 32 130 250 300 400 500
Radar Simulators Ltd............. 6 111 250 400 500 600 750
Solartron Industrial Controls Ltd........... 6 12 60 150 250 300 350
Solartron Electronic Business Machines Ltd....... 2 10 50 200 300 400 500
Solartron Electronic Group Ltd.
Merchanting and sundries.......... 16 9 54 96 140 200 250 275 300

Total deliveries............ 398 758 1,005 1,434 2,250 3,500 4,750 6,000 7,500

Export content included in above figures...... 20 80 186 335 600 1,000 1,500 2,250 3,500
Total orders.............. 400 800 1,250 1,900 2,750 4,000 5,500 7,000 8,500
Total personnel at year end.......... 400 550 600 830 1,250 1,750 2,250 2,750 3,500

It was apparent in March, 1959, that it would probably be necessary to extend the 1962/1963 targets to 1963/1964, that is, to spread the five-year programme
over six years. Figures for 1959/1960 through 1962/1963 are "maximum" targets. Source: Company records. (Reproduced from Learned, Aguilar & Valtz, op. cit.) Optical Char
 Optical Character Recognition

next five years would have to be spread over six years. Since then we
have found ourselves falling behind our schedule of forecast deliveries
to an even greater extent." Development periods had been under-
estimated "on almost every occasion" and market build-up had been
slower than calculated. This applies to a number of projects, but is
especially applicable to ERA.
Two years later Firth Cleveland decided that they could not afford
to go on supporting long-term R & D and sold its interest to Schlum-
berger, the Franco-American oilfield instrumentation company. They
also found that Solartron was an unprofitable property, and were
obliged to write off 974,000 worth of "capitalised R & D". (see
Table II) in 1961. B. Schneerson, the Schlumberger director responsible
for Solartron said the company had grown too fast, was low in pro-
ductivity, weak in control, too diversified and unclear about its targets.
In 1963 the new Board ordered that the work on ERA should stop. The
Downloaded by Universiti Teknologi MARA At 08:22 06 December 2017 (PT)

Burton installation was working but the DER instrument was never
delivered. Some of the know-how, such as the till roll transport and
rewinding mechanism, which had been patented, was sold to Farrington
who also bought the DER circuit for study. The ERA teams were dis-
banded and several went off to found their own "spin-off" companies,
as Curry and Carroll did. They had been a very loyal, dedicated and
enthusiastic working group, putting in unlimited overtime and working
weekends, at the expense of family strain (there were nine divorces
during the intensive eight months of the Boots machine). Some of
these men still meet and discuss OCR. It cannot be questioned that this
R & D-based company, SEBM, was hardworking and knowledgeable.
If anything was lacking, apart from funds, it was experience, but they
were all breaking entirely new ground.
It was in 1961, when Solartron passed into Schlumberger control,
that Shepard resigned from Farrington, mainly on the grounds
that, because Scandex was successful he was not free to work on the
more advanced instruments which interested him (formally unqualified,
he had been brought up by an uncle, Laurens Hammond, inventor of
the synchronous electric clock and the electric organ, so that he feels
he grew up "in an inventive atmosphere"). He subsequently founded
Cognitronics, New York, a remote data processing bureau using OCR.
Under him IMR had grown to 80 staff, twice the size of SEBM, before
being taken over. Farrington, then 400 strong (1959) now numbers
4,000 in three countries, but has been overtaken in the OCR market by
IBM, CDC and Recognition Equipment Inc. The company has left New
York and Boston and is based at Springfield, Virginia, IMR's original
factory site. It is still one of the "big five" in OCR, with a market share
of 13 per cent world wide by volume, 9 per cent by value (1967).
Whether, had Shepard stayed, Farrington would have kept its techno-
logical leadership is conjectural, once IBM had rid itself of dependence
on his patents. At the time of the first Scandex (8SP1 was its serial
number) Farrington had a lead of five years in OCR, according to
Shepard, who knew most if not all of the other research workers in the
field. He left behind a group of 18 well trained and experienced

222 Management Decision

 Optical Character Recognition

technologists in OCR, including Eugene Milbradt, physicist and engin-

eer, whom he had recruited to 1MB, and Howard Silsby, a Farrington
engineer, with whom he took out a joint patent for a regulatory system
for photo-sensitive optical scanners. From 1955 to 1966 Farrington lost
money on OCR (the figures are disputed still) but made a $501,495
profit in 1967 on sales of $17-5m.

Conclusion
An outstanding difference between the two innovations lies in the
"need push" aspect of the Farrington attempt (although the IMR work
had not originally had a specific customer in mind, Shepard was
acutely aware of the need for OCR), while with Solartron it was the
"technology pull" of flying spot scanning that seems to have been the
main motivating force. Even Shepard's three line reader for Readers'
Digest was designed to do a specific jobscanning addresses printed
Downloaded by Universiti Teknologi MARA At 08:22 06 December 2017 (PT)

from plates. The first ERA had no specific target, compared to the
second for Boots and the third for Littlewoods. It might be added that
Shepard and his colleagues were also much more aware of the physical
constraints on the method than Norrie and his team, of the importance
of font design (which is still a bone of contention in the OCR business)
and of the necessity to accept a high initial reject rate and not aim too
high at first. As if to set the seal on Farrington's success, in 1967 they
sold an OCR machine to Vernon's Pools, Liverpool, Littlewood's chief
competitor.

Acknowledgements
The following have provided indispensable information, advice and
assistance but are in no way responsible for any errors or omissions.
Mr. John Bolton and Mr. Bowman Scott, formerly Chairman of
Solartron and managing director of SEBM, respectively; Mr. Oswald
Boxall, former executive vice-president, Farrington Manufacturing
Company; Mr. David Shepard, president, Cognitronics Ltd., New York;
Dr. John Parkes, National Physical Laboratory; Mr. Tom Ward and Mr.
Neville Groome of Littlewoods; Mr. John Carroll, Data Recognition
Ltd.; Mr. Peter Curry, managing director, Plasmec Ltd. The last two
were interviewed by my colleague Paul Jervis, to whom also go my
thanks.

References
European Problems in General Management, Learned, Aguilar and Valtz, Irwin Book Company,
1963. The Solartron case.
Character Recognition, The Computer Society, 1967.
"The Flying-Spot Microscope", F. Roberts and J . Z. Young, F.R.S., Proceedings of the I.E.E.,
1952,99 (20), Part III A.
"The Reality of Optical Scanning", A. E. Keller, Management and Business Automation, (U.S.A.),
September 1960. In this article Farrington's leadership is established.
"Farrington Scans the Future". Dun's Review, August 1968.

Vol. 9 Winter 1971 223