Admissions Forecaster
- Former user (Deleted)
- Lucidchart Diagrams Connector
- Former user (Deleted)
- Former user (Deleted)
The admissions forecaster attempts to give an indication of what student numbers will be per department based on current application figures, and the likelihood that those applications will become Unconditional Firm (UF). A conversion to UF rate is calculated for each route owner / student mix / application status / intake type (DL, HEP, IPC, Standard) combination based on conversion of applications to Unconditional Firm from that combination last year. For example, a single rate is calculated for All Archaeology, UG standard applicants who on this day last year had an application status of CF. If 50% of them became UF then it is assumed that any Archaeology UG applicants with a status of CF today also have a 50% chance of becoming UF. These rates are recalculated on a daily basis so that we are always considering what applicants did at exactly the same point in the cycle last year
Although, having said a single rate is calculated, 4 rates are actually calculated, but in each case the same logic applies:
- Conversion to UF at end of cycle (last year)
- Conversion to UF at end of cycle (3 year average)
- Conversion to registered on 1st December (last year)
- Conversion to registered on 1st December (3 year average)
We also estimate future applications based on what percentage of applications for a route owner / student mix / intake type had not yet been received this time last year and again of those not yet received, what percentage became UF. This has a large impact on PG estimates
Using these rates we then add together all the UF estimates from each application status estimate to give an estimated total UF for that route owner.
Limitations and assumptions
The logic used for the admissions forecaster has a number of limitations that may affect the validity of the results it predicts:
- It assumes that the past is a good predictor of the future and indeed that students this year will behave as students in previous years have behaved. As external factors change this may not be the case
- There is no adjustment for initiatives that may affect these rates, e.g. a successful marketing initiative last year affected the numbers because students were applying earlier than in previous years, but the algorithm assumed that the same rate of future applications applied. Likewise the early UF offer scheme may have an effect that won't be covered
- The level of granularity does not take account of the differences that may appear between programmes. For example, a department may offer some programmes that are easy to recruit to and others that are not, but a single rate is calculated for them all which may over-inflate the estimate for some programmes at the same time as under-estimating for others.
As of 01/04/2020 extensive testing has been carried out on the accuracy of the the model utilised by the admissions forecaster. It has reaffirmed that the accuracy of the model is hugely dependent on the similarity of the current application year and the year which is used as the basis for the conversions rates. Attempts have been made to mitigate this, by allowing the user to choose the year they wish to use as the basis for the conversion rates, however change to the applications landscape will still have a significant affect on the accuracy of the model and this should always be kept in mind when quoting the figures produced by this report. If the the current application year and the year which is used as the basis for the conversions rates are similar then the model should produce a forecast of final numbers with a 5% margin of error from early May. The forecaster works best for UG and PGT cohorts and is less reliable for PGR cohorts.
The workbook has the following tabs:
UG This tab shows targets and predictions by route owner for UG admissions. The first column shows the prediction against the target that has been planned in RAMP for the upcoming academic year. The second column shows the surplus or shortfall that would occur against the target if the prediction turns out to be correct. The third column shows, where a route owner took students from clearing in the previous academic year, the FTE of those students. The fourth, and final, column adds the number achieved in clearing last year to the number of UF's predicted by the model together and compares that to the target to see whether there will be a surplus or a deficit if the prediction turns out to be correct and the same number of students are taken in clearing. The sheet also shows a total for all route owners at the top. But, it should be noted that a shortfall in one department is not offset by a surplus in another, and where a department reaches target they may not take as many in clearing this year, so the final overall figure is unlikely to be correct Hovering over any bar in the chart will bring up a table that shows the detail of the prediction by current applications status |
UG Table This is a tabular version of the data on the UG sheet which can be downloaded by crosstab to Excel |
PG This tab shows targets and predictions by route owner for PG admissions. The first column shows the prediction against the target that has been planned in RAMP for the upcoming academic year. The second column shows the surplus or shortfall that would occur against the target if the prediction turns out to be correct. The next two columns are a repeat of the first two columns but with the inclusion of an estimate of how many future applications are likely to be received. Again, the sheet also shows a total for all route owners at the top. But, it should be noted that a shortfall in one department is not offset by a surplus in another, so the final overall figure is unlikely to be correct. Hovering over any bar in the chart will bring up a table that shows the detail of the prediction by current applications status |
PG Table This is a tabular version of the data on the PG sheet which can be downloaded by crosstab to Excel |
Department Summary The department summary tab shows all PG and UG predictions for a single route owner on a single page. That is it consolidates what is shown on the UG and PG tabs into a by department view. Clicking on any listed student mix will display a breakdown by programme at the bottom of the page |
SQL Files
apps_conversion_predictions_v.sql
apps_conv_rates_1_year_reg_v.sql
apps_conv_rates_3_years_reg_v.sql
Tableau File
Q:\DISCOVER\Student\Workbooks\Published\Applicant Tracking\Tableau\Admissions Forecaster v0.7 PLANDISC.twbx
Snapshot
a daily snapshot is taken of apps_conversion_predictions_v and inserted into apps_conv_predict_snap by the stored procedure mv_refresh_pkg.apps_predictions_snap which is called by the scheduled job append_apps_predictions that runs daily at 5am
The workbook makes use of the following filters:
| Filter Name | Filter Options | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| DL Flag | This filter allows for the inclusion / exclusion of applicants and targets based on whether or not they are York distance learning applicants. | ||||||||
| HEP Flag | This filter allows for the inclusion / exclusion of applicants and targets based on whether or not they are Higher Education Partners (HEP) applicants. | ||||||||
| IPC Flag | This filter allows for the inclusion / exclusion of applicants and targets based on the IPC flag the records are tagged with | ||||||||
| Student Mix | This filter allows the selection of a single student mix to display results for. The UG tabs are restricted to UGH and UGO, whilst the PG tabs are restricted to PGTH, PGTO, PGRH and PGRO | ||||||||
| Conversion rate to use (PG) | This filter can be used to select one of the following options of conversion rate from current application status to UF:
| ||||||||
| Conversion rate to use (UG) |
** Indicates the default setting for this parameter as specified by colleagues in SRA | ||||||||
Conversion Rate Year to use (Only Applicable to 1 year Conversion Rates) | This filter can be used to select which of the previous three years should be used as the basis for the single year conversion rates. This has no effect if one of the 3 year average options is selected for the "Conversion Rate to Use" Filter. This affects both the UG and PG "Conversion Rate to Use" Filters. | ||||||||
| Route Owner Selection | This filter allows the display of different combinations of route owners, excluding, including or only showing HYMS and Health Sciences | ||||||||
| Route Owner | This filter allows the selection of a single route owner to display |
The following work may be done in the future
| Item | Description |
|---|---|
| Reliability testing | Use apps_conversion_predictions snapshots to assess the accuracy of the predictions against the actual admissions figures at the end of the current cycle. Ideally multiple cycles will be tested to gain a consistent picture |
| Include Accommodation Flag | Information about how soon an application is made for accommodation once the accommodation system is open may be a strong indicator of intention to come to the University. This could be built in as a predicting factor |
| CAS Status | CAS Status is another predictor of whether or not an application is intending to take up their offer at the university. This could also be built into the predictions |
| Remove hard coding of academic year in conversion rate views | The conversion rate views that use previous year average and last three years average is hard coded so need to be manually moved on. |
| Remove schema references in conversion rate views | The views need to use the same code to run on either planpre or plandisc. |