1. Make use of the findings from the historical data to determine the workings of good trends, such as the animal quality improvements in certain regions feeding Arusha and Moshi, and to try to extend those good practices to other regions and to other markets.

2. Change from a sampling system which tries to be population representative to one which is sex-weight-grade combination representative; this will give more accurate pricing for smaller sample sizes, i.e. better results for less work. The resulting potential loss of data for the analysis of certain trends in the cattle business should be countered if felt necessary by a selective program or programs, building on the historical data and the historical analysis now available. Note that some markets do keep daily summaries (e.g. on numbers and total values by sex; maximum, minimum and average prices by sex, etc..) which may be useful to researchers and which may compensate for the loss of representative sampling.

3. Ensure that all scales used are calibrated regularly, adjusted when required, and repaired if required; calibration weights should be present full-time at all 3 major markets; the weight correction facilities of the computer program should be utilised; consider the replacement at some time of these conventional spring scales by load-cell scales for faster weighing with digital readout which will reduce operator or recorder error. If necessary, discuss the concept with market staff to inculcate the need for proper calibration and correction (instead of the dubious practices I observed).

4. Although the grading data appears remarkably good (even between markets), do organise periodic grading tests and courses if then required whose function will be to set absolute standards but also (probably more important) to ensure homogenous standards between the 3 major markets and among all personnel at these 3 major markets.

5. Adopt practices to ensure that price information is accurate, as discussed above (including possibly more observed pricing); try to research the practice of mob pricing and effect sampling so as to try to avoid being landed with mob price information; give disincentive to data collectors to pass off mob prices as real prices; investigate any and all other factors which produce a ratio of weighings-plus-pricings to total-weighings of less than 100%, since each weighing without a (useable and accurate) pricing is a loss of man-hours and useable sample size.

6. Adopt the log-log best-fit (LLBF) data processing technique to process the data for all 3 major markets (and for other markets if and when required); this will require the reinput or the data translation of data input into the 2 old systems during the period October 1994 through February 1995; use the standard deviation measure for each market day from the LLBF system as an indicator of (price variability + data inaccuracy) - keep an eye on this number - if it gets too high there is something wrong.

7. Change the presentation formats to ones which use benchmark sex-weight-grade combinations, with common values between the 3 major markets to permit easy comparison; also discontinue the presentation of much spurious information on the makeup of the animal population (i.e. average weight, average price and percentage of total population for each grade-sex combination); note that this discontinuation will anyway be necessary if my recommendation on sampling is adopted, since the sampling will therafter no longer be representative.

8. Integrate the presentation of information from all 3 markets, on radio, in the newspaper, and for the export reports; eliminate the present practice of ignoring Arusha and Moshi markets. Especially revamp the export report with the presentation of price data in US Dollars and/or in the currency of the target country; keep a file of press cuttings, radio transcripts, dates and channels, etc..

9. Reactivate the presentation of information at the markets themselves; ensure that there is zero time lag between data collection and information reporting, by the use of notebook computers onsite at the markets; this will anyhow be useful in the case of the possible privatisation of the organisation of these markets.

10. Encourage self-weighing by buyers and/or sellers as a means of reducing the cost and logistical load of the MIS program, and again as an aid in the possible privatisation of the organisation of these markets.

11. Consider whether to effect a check between the data-translated Moshi and Arusha data and the paper raw data for accuracy of the date of the market-day (as discussed above in the section Data Translation).

12. Investigate the raw data from the Tanzanian Texas A&M study (if it can be located) to better determine the effect on edible weight of varying liveweight within each grade-sex combination; apply this refinement to the computer system if such variation exists and can be quantified; with the aim of making the data processing slightly more accurate. Examine also whether the physical characteristics of the different breeds of cattle can or should be taken into account.

13. Institute a yardage count at Pugu (as is done at Moshi and Arusha) in addition to the present railage numbers collection (discussed in (15) immediately below).

14. If the TLMP Project Funding of Cattle Wagon Bodies for Tanzania Railways goes ahead, then consider producing a Management Information System to report on demand and supply of cattle movement by rail, and particularly on variation in supply of cattle to terminal markets and its effect on unit price; regardless of this, modify the present railage data collection system at Pugu Station to note date and railage, instead of totalling by week.

15. Consider strongly to eliminate the collection of data on age, buyer ID and origin, since these are additional work, and since no real analysis of these parameters appears to have been conducted for years; in this respect consider whether the analysis conducted in this report may be sufficient for some time to come, and/or may pinpoint some aspect of one or more of the above 3 fields which should be investigated in some detail. Note incidentally that the Dar railage information (as in (15) and (17) can and could give origin information on 100% of the population (at least from railhead).

16. Consider quickly recapturing the Dar weekly railage numbers for graphical analysis over the 5 year period, from the weekly reports from the Symphony System. This is one aspect which is missing from this report, in which quite detailed yardage and sold information is given for Arusha and Moshi markets.

17. As a very minor item, consider improving the present apparently cumbersome and labour-intensive cattle identification (TRN'ing) process, as discussed above.

18. For computer hardware, I strongly recommend that the TLMP Project procures 3 or more notebook IBM-compatible computers with 2 sets of spare batteries and external battery re-chargers. Also with cigar-lighter adaptors. These notebooks should NOT ever be plugged directly into mains supply for use or even for charging.

One Hewlett-Packard Laserjet 4 Plus and 3 or more Hewlett Packard Deskjet 320 portable monochrome/colour printers should be procured (together with rechargeable batteries and spare cartridges). A high quality surge protector should protect the Laserjet. The bubblejet printers should if possible run from battery only, not from mains.

Software for Statistical Analysis, Backup, and AntiVirus should be procured. Windows 3.1, WFWG 3.11, or Windows 95 should be the operating system.