WEBVTT NOTE IDER in-house training materials. Captions follow measured phrase audio timings. Hypothetical worked cases; figures are instructional examples. 00:00:00.000 --> 00:00:06.505 Imagine a service business comparing two urban locations. One has stronger apparent demand; the other 00:00:06.505 --> 00:00:13.011 offers easier access and lower operating costs. The practical question is not which location looks 00:00:13.011 --> 00:00:13.444 busiest. 00:00:13.564 --> 00:00:19.808 It is which location can support the intended service, at an acceptable risk, under an explicit set of 00:00:19.808 --> 00:00:24.663 priorities. In this IDER in-house clinic, we will build that decision from observable evidence. 00:00:24.783 --> 00:00:30.658 All locations and figures are hypothetical teaching examples. Keep a spreadsheet nearby. 00:00:30.778 --> 00:00:37.197 You will calculate a ranking, discover when that ranking reverses, and test whether the preferred 00:00:37.197 --> 00:00:39.764 option actually covers its operating costs. 00:00:39.884 --> 00:00:45.794 The final product is a recommendation another analyst can reproduce and a manager can challenge. 00:00:46.464 --> 00:00:52.222 Start with conditions that a location must satisfy before you compare its attractiveness. 00:00:52.342 --> 00:00:58.885 If the service requires reliable power and a site cannot supply it at a feasible cost, an attractive 00:00:58.885 --> 00:01:01.793 demand score should not compensate for that failure. 00:01:01.913 --> 00:01:08.116 Write each gate as a test, with an evidence source and an accountable reviewer. Then identify genuine 00:01:08.116 --> 00:01:09.575 trade-offs among eligible sites. 00:01:09.695 --> 00:01:16.674 A slightly longer journey may be acceptable if customers obtain better service or lower prices. This 00:01:16.674 --> 00:01:22.344 separation prevents an impressive weighted average from hiding a fatal constraint. Record exceptions 00:01:22.344 --> 00:01:22.780 explicitly. 00:01:22.900 --> 00:01:29.174 An unresolved permission issue belongs in a conditional decision, not inside a quietly reduced 00:01:29.174 --> 00:01:31.863 score that executives might never notice. 00:01:32.533 --> 00:01:38.420 A circular radius on a map is convenient, but it rarely describes the journey customers actually make. 00:01:38.420 --> 00:01:42.575 Build catchments using the transport modes and hours relevant to your service. 00:01:42.695 --> 00:01:49.830 A road crossing, transfer or flooded route can change effective access without changing physical distance. 00:01:49.830 --> 00:01:56.489 Sample customer origins and ask which alternative destination they would use. This reveals overlapping 00:01:56.489 --> 00:01:57.916 catchments and substitution. 00:01:58.036 --> 00:02:04.175 Also define your unit carefully. The same commuter passing twice is two observed visits, not necessarily 00:02:04.175 --> 00:02:06.476 two people or two purchase opportunities. 00:02:06.596 --> 00:02:12.476 For a practical study, maintain separate fields for eligible visits, identifiable customers and completed 00:02:12.476 --> 00:02:18.355 purchases. That distinction will become essential when we turn observed movement into a contribution 00:02:18.355 --> 00:02:18.775 forecast. 00:02:19.445 --> 00:02:25.339 Suppose your team counts Site A on a payday evening and Site B on a quiet weekday morning. Precise 00:02:25.339 --> 00:02:27.201 counting will not repair that comparison. 00:02:27.321 --> 00:02:34.613 Create a schedule crossing the time periods and day types that matter, then observe both sites within 00:02:34.613 --> 00:02:35.471 comparable strata. 00:02:35.591 --> 00:02:41.427 Expand each stratum using the number of similar operating periods it represents, rather than 00:02:41.427 --> 00:02:44.345 multiplying one convenient hour across the month. 00:02:44.465 --> 00:02:49.519 Record unusual weather, promotions and closures so that analysts can investigate anomalies. Repeated 00:02:49.519 --> 00:02:56.516 observations within one day may share the same shock, so do not treat every minute as independent evidence. 00:02:56.636 --> 00:03:02.813 Before leaving the field, reconcile observer definitions using a short joint counting exercise and 00:03:02.813 --> 00:03:03.695 review disagreements. 00:03:04.365 --> 00:03:11.752 Our illustration assigns each site a score for demand, access and cost attractiveness. Higher always means 00:03:11.752 --> 00:03:18.216 better, including for cost, where a cheaper comparable operating arrangement receives a higher score. 00:03:18.336 --> 00:03:25.266 The numbers are useful only after you define their anchors. For example, specify what zero and one 00:03:25.266 --> 00:03:28.936 hundred mean and why intermediate performance deserves its position. 00:03:29.056 --> 00:03:35.679 Otherwise two assessors may attach different meanings to eighty. Avoid scoring the same benefit twice 00:03:35.679 --> 00:03:41.861 through correlated criteria such as footfall and forecast sales unless the model explicitly separates 00:03:41.861 --> 00:03:42.744 their contributions. 00:03:42.864 --> 00:03:49.088 Keep the raw measurements alongside the converted scores. When someone disputes the ranking, you can 00:03:49.088 --> 00:03:54.067 then distinguish a disagreement about observed performance from a disagreement about priorities. 00:03:54.737 --> 00:04:03.062 We now apply demand, access and cost weights of fifty, thirty and twenty percent. Site A receives forty 00:04:03.062 --> 00:04:08.611 points from demand, eighteen from access and eight from cost, giving sixty-six. 00:04:08.731 --> 00:04:15.883 Site B receives thirty-two point five, twenty-five point five and fifteen, giving seventy-three. These 00:04:15.883 --> 00:04:19.970 are preference scores, not probabilities of commercial success. 00:04:20.090 --> 00:04:26.282 Elicit weights against the defined improvement from the worst to the best performance on each scale. 00:04:26.282 --> 00:04:31.700 Asking which criterion sounds most important without showing that range can create misleading weights. 00:04:31.820 --> 00:04:37.015 Record whose preferences the weights represent and obtain their confirmation before revealing the final 00:04:37.015 --> 00:04:42.582 ranking. This reduces the temptation to adjust the model until it endorses a favourite site. 00:04:43.252 --> 00:04:49.693 Pause the video and challenge the baseline. A decision maker now proposes a seventy percent demand weight, 00:04:49.693 --> 00:04:54.618 while keeping the relationship between access and cost weights at three to two. 00:04:54.738 --> 00:05:01.281 First calculate the new access and cost weights. Then recompute both site scores without changing their 00:05:01.281 --> 00:05:02.099 underlying performance. 00:05:02.219 --> 00:05:08.771 Finally, write one sentence explaining whether a changed ranking would indicate better evidence or 00:05:08.771 --> 00:05:09.707 different preferences. 00:05:09.827 --> 00:05:15.996 That final distinction matters in practice: a model can change its recommendation even when no new 00:05:15.996 --> 00:05:21.008 observation has been collected. Show your intermediate contributions rather than only the totals. 00:05:21.128 --> 00:05:27.345 If your weights do not add to one hundred percent, correct the model before interpreting any resulting 00:05:27.345 --> 00:05:27.711 ranking. 00:05:28.381 --> 00:05:35.785 The remaining thirty percent divides into eighteen percent for access and twelve percent for cost. Site A 00:05:35.785 --> 00:05:41.012 now scores seventy-one point six, while Site B scores sixty-nine point eight. 00:05:41.132 --> 00:05:47.462 The ranking reverses because Site A's demand advantage receives more influence. Nothing in this calculation 00:05:47.462 --> 00:05:53.791 proves that the original observations were wrong. You have changed the decision maker's valuation of 00:05:53.791 --> 00:05:54.635 performance differences. 00:05:54.755 --> 00:06:01.744 In a committee discussion, show both scenarios and ask whether the higher demand priority reflects the actual 00:06:01.744 --> 00:06:06.678 business strategy. Also check whether all participants understood the same performance ranges. 00:06:06.798 --> 00:06:13.759 A useful sensitivity analysis reveals a consequential disagreement; it does not average that disagreement 00:06:13.759 --> 00:06:18.234 away and present an artificial consensus as objective measurement. 00:06:18.904 --> 00:06:25.173 Instead of trying isolated weights, solve for the exact point where the two scores are equal. Let the 00:06:25.173 --> 00:06:26.567 demand weight be w. 00:06:26.687 --> 00:06:34.567 Access receives three fifths of the remainder and cost two fifths. Site A's score becomes fifty-two plus 00:06:34.567 --> 00:06:39.666 twenty-eight times w. Site B's becomes eighty-one minus sixteen times w. 00:06:39.786 --> 00:06:46.027 Equating them gives forty-four times w equal to twenty-nine, so the switching point is approximately 00:06:46.027 --> 00:06:52.267 sixty-five point nine one percent. Below that value, B ranks higher; above it, A does. 00:06:52.387 --> 00:06:58.059 This threshold is useful because decision makers can assess a meaningful proposition: do we really value 00:06:58.059 --> 00:07:03.730 the defined demand improvement enough to allocate roughly two thirds of the total weight to it? 00:07:04.400 --> 00:07:11.305 Weight sensitivity addresses preferences; it does not capture uncertainty in measured demand. Treat those as 00:07:11.305 --> 00:07:12.225 separate layers. 00:07:12.345 --> 00:07:18.625 Revisit footfall estimates, conversion assumptions and operating costs using defensible ranges linked to 00:07:18.625 --> 00:07:23.456 actual evidence quality. Then examine combinations that can occur together. 00:07:23.576 --> 00:07:30.340 A road disruption may reduce visits and raise delivery costs simultaneously, so independent one-variable 00:07:30.340 --> 00:07:36.620 changes could understate risk. Conversely, combining every imaginable worst case without a coherent 00:07:36.620 --> 00:07:38.553 mechanism can exaggerate it. 00:07:38.673 --> 00:07:44.741 For repeated field observations, consider resampling at the day or other relevant cluster level rather than 00:07:44.741 --> 00:07:45.500 individual counts. 00:07:45.620 --> 00:07:51.856 Report how often rankings change under the stated scenarios, but do not call that a probability of 00:07:51.856 --> 00:07:54.791 success unless you have justified a probability model. 00:07:55.461 --> 00:08:01.480 Now move from relative attractiveness to absolute viability. Assume twelve hundred eligible visits per 00:08:01.480 --> 00:08:07.929 day, a four percent purchase conversion and twenty-six operating days. That gives twelve hundred and 00:08:07.929 --> 00:08:09.219 forty-eight monthly purchases. 00:08:09.339 --> 00:08:14.858 At thirty-five Ghana cedis contribution per purchase, monthly contribution is forty-three thousand six 00:08:14.858 --> 00:08:21.649 hundred and eighty. Subtract fixed operating costs of forty-two thousand and the surplus is only sixteen 00:08:21.649 --> 00:08:22.923 hundred and eighty. 00:08:23.043 --> 00:08:30.121 Contribution means selling price less the costs that vary with the purchase; it is not sales revenue. This 00:08:30.121 --> 00:08:35.233 small surplus should immediately focus your evidence collection on conversion and visit quality. 00:08:35.353 --> 00:08:43.485 It also excludes investment recovery, financing and tax, so a positive result alone does not establish an 00:08:43.485 --> 00:08:44.920 acceptable investment return. 00:08:45.590 --> 00:08:51.487 A pilot is most valuable when it resolves a decision-sensitive uncertainty. If the economic case 00:08:51.487 --> 00:08:56.991 depends on four percent conversion, simply collecting more general footfall data may add little. 00:08:57.111 --> 00:09:04.062 Test the actual offer, opening hours and customer journey at a comparable location or through a 00:09:04.062 --> 00:09:09.275 permitted temporary operation. Record the eligible denominator before observing purchases and track 00:09:09.275 --> 00:09:10.579 repeat customers separately. 00:09:10.699 --> 00:09:17.767 Specify the duration and decision rule before seeing results, including what would trigger redesign or 00:09:17.767 --> 00:09:18.709 further evidence. 00:09:18.829 --> 00:09:24.979 A short successful trial can still overstate normal performance if novelty, discounts or exceptional 00:09:24.979 --> 00:09:26.736 staffing drove the outcome. 00:09:26.856 --> 00:09:32.986 Explain how the full operation would differ and adjust the forecast using documented mechanisms, not a 00:09:32.986 --> 00:09:34.135 convenient unexplained haircut. 00:09:34.805 --> 00:09:41.129 Pause and calculate a combined downside. Daily eligible visits fall by ten percent, and conversion 00:09:41.129 --> 00:09:47.453 falls to three point five percent. The operating calendar, contribution per purchase and fixed costs 00:09:47.453 --> 00:09:48.296 stay unchanged. 00:09:48.416 --> 00:09:54.092 Calculate monthly contribution and the resulting surplus or deficit. Then explain why subtracting ten 00:09:54.092 --> 00:09:58.552 percent from the original operating surplus would give the wrong answer. 00:09:58.672 --> 00:10:05.318 The revenue-generating activity changes before fixed costs are deducted, and conversion has changed as 00:10:05.318 --> 00:10:11.489 well. In your written recommendation, distinguish a ranking conclusion from a viability conclusion. 00:10:11.609 --> 00:10:17.662 A site can remain the best of the compared locations while all available options fail the business's 00:10:17.662 --> 00:10:22.290 minimum economic requirement. Deciding to defer is a legitimate output of professional research. 00:10:22.960 --> 00:10:30.607 Ten percent fewer visits gives one thousand and eighty per day. Multiplying by three point five percent 00:10:30.607 --> 00:10:36.004 conversion, twenty-six days and thirty-five cedis contribution yields thirty-four thousand three hundred 00:10:36.004 --> 00:10:36.904 and ninety-eight. 00:10:37.024 --> 00:10:43.920 After fixed costs, the site loses seven thousand six hundred and two cedis monthly in this scenario. Fixed 00:10:43.920 --> 00:10:48.901 commitments explain why the operating result deteriorates much faster than the visit count. 00:10:49.021 --> 00:10:55.062 You should now investigate whether rent terms, staffing patterns or the service offer can change 00:10:55.062 --> 00:10:59.493 without undermining demand. Recalculate the whole mechanism after any proposed adjustment. 00:10:59.613 --> 00:11:05.921 A lower price may improve conversion while reducing contribution per sale; treating that change as a pure 00:11:05.921 --> 00:11:08.889 demand benefit would repeat the same forecasting error. 00:11:09.559 --> 00:11:14.552 Your decision memo should let a reviewer reconstruct the recommendation without rebuilding the entire 00:11:14.552 --> 00:11:20.259 study. Show which sites passed the eligibility gates, how the catchments were defined and when observations 00:11:20.259 --> 00:11:20.972 were collected. 00:11:21.092 --> 00:11:27.424 Present raw measures alongside scores, identify the agreed weights and display the switching threshold. 00:11:27.424 --> 00:11:32.400 Then attach the operating model with baseline and coherent downside cases. 00:11:32.520 --> 00:11:39.051 State a conditional recommendation in operational language: proceed only if a specified uncertainty is 00:11:39.051 --> 00:11:45.581 resolved, an acceptable lease condition is obtained, or a pilot meets the predefined requirement. 00:11:45.701 --> 00:11:52.826 Assign an owner and deadline to every condition. Preserve the data version and calculation file so that 00:11:52.826 --> 00:11:58.693 the next update changes the evidence transparently instead of silently rewriting the original decision. 00:11:59.363 --> 00:12:05.284 Before using this method in your next assignment, audit one denominator and one scoring anchor. 00:12:05.404 --> 00:12:10.205 Ask whether your visits represent genuine purchase opportunities and whether a ten-point score 00:12:10.205 --> 00:12:16.484 improvement means the same thing to every decision maker. Then calculate both a ranking threshold and an 00:12:16.484 --> 00:12:17.223 economic threshold. 00:12:17.343 --> 00:12:24.479 The first tells you which option is preferred under stated values; the second tells you whether an option 00:12:24.479 --> 00:12:28.047 is worth operating at all. Those are complementary decisions. 00:12:28.167 --> 00:12:35.332 Complete the assessment below, then prepare the two applied tasks using either the hypothetical case or a 00:12:35.332 --> 00:12:36.596 suitably anonymised project. 00:12:36.716 --> 00:12:42.177 Evaluate your work by traceability, coherent assumptions and the clarity of the conditions attached 00:12:42.177 --> 00:12:46.468 to your recommendation, not by how confidently you name a winner.