April 9, 2015
Kylie did a stellar job with exceeding sales targets for her store during the recent promotion, but a report from the corporate office claimed her conversion rates should be significantly lower than what she reported. She knew the data didn’t reflect reality, so she received approval from Corporate to do an extensive audit of the numbers, using an outside traffic consultancy.
Over the next several weeks, the audit revealed some telling results. Since the current traffic solution tallied only walk-in counts, and not everyone who walked in was necessarily a shopper, a better metric was needed to measure business performance. The one suggested by the traffic consultants was store “occupancy”, which is calculated by subtracting the cumulative sum of shoppers exiting the store from the cumulative sum of shoppers entering. Because the system did not provide it out of the box, the metric was manually calculated from the walk-in and out counts and compared against physical counts during specific time intervals. For example, during the time t, from 2:00-2:15 PM, the occupancy from the traffic counters was calculated as the occupancy (t-1) + walk-in(t)— walkout(t) = 6 + 13 - 6 = 13. In the next time interval, t+1, 13 would subsequently be used as the input occupancy.
Comparing the counts from the traffic counter to the physical in-store counts performed by the traffic consultants revealed the traffic counters overcounted the walk-ins by 2 and under-counted the walk-outs by 1. Consequently, the errors in walk-ins and walk-outs led to errors in the occupancy, which accumulated through the course of the day.
What’s more, the more accurate physical counts revealed additional sources of inaccuracy. For instance, 15% of the store traffic counted entailed employees entering and leaving the store. And another 25% represented groups of shoppers, each of which should have only represented a single potential customer. However, the most critical issue was that Kylie’s scheduling system would consistently recommend an additional 2 headcount during mid-shift because of the significant increase in walk-ins during lunch hour. Closer inspection revealed most of the shoppers walking in through the front entrance were walking immediately out the back, using her store as a shortcut to the food court. Kylie’s scheduling system relied not only on inaccurate walk-in counts, but was only factoring walk-in counts to estimate staff levels instead of using true customer occupancy.
With this kind of inaccuracy, no wonder there was such a huge discrepancy between corporate feedback and what was actually happening. When Kylie reported her findings, the corporate immediately assigned IT to evaluate the traffic counting solution currently deployed and investigate alternatives. It was clear the solution couldn’t be relied on to evaluate basic business performance indicators, like sales conversion, let alone be counted on to forecast occupancy for something as critical as staff optimization. Inaccurate traffic data can lead to even more inaccurate traffic forecasts, which can have serious implications on workforce optimization and your ability to accurately measure the effectiveness of marketing and merchandising. After all, is it not realistic to expect garbage in can only result in garbage out?
If you would like to know how accurate your current traffic solution is, here is a quick test you can conduct to validate:
Contact us if you would like to learn more about how your organization can address this challenge.