measuring average month over month performance

On a month-to-month yon could take an average of all the stores and see how the individual compares to this. If there are enough stoes then an individual low or high won't affect the  average too much to be able to see a variation.

I assume the customer rep has nothing to do with the operations of the whole store.
Try
#units of store A/# of units in ALL stores.
advantage - industry wide changes would be minimized. However this metric would favor large stores. Average his score over all his 50 stores. That should reduce the size effect (unless stores are assigned by size).
A person who consistently increases his score is good. Keep him.
One can get more sophisticated but might not help the evaluation process much

This looks like basic econometrics.

Your baseline, your universe, is the chart of total units over time. However, this includes the growth of the entire store chain (more store, more units) so it's unfair to reps who are stuck at 50. Using the average of units per store is therefore better.

This chart shows a monthly trend (the gain or loss in volume) and a volatility. If you have enough stores, this is the volatility of the business branch of the stores speciality. It's defined as the standard deviation of monthly trends: each month sees a percent increase or decrease compared to the previous month, taking the standard deviation of that shows how "rugged" the chart is. Low volatility means stability and predictability (for risk analysis, for example). Volatility is like a moving average, since it needs a certain number of months to become meaningful.

Now each rep has about 50 stores, but not all stores have the same size, and the number can change. Again, it makes sense to use the average units per store as the reps basic chart, but it can be argued that the total units better represents performance.

In any case, you have again monthly figures, their variation, and the volatility. You can compare these charts to your baseline.

There is also the "internal volatility", and "internal coherence". These are again the standard deviation of monthly variations, the first store by store over a period (say 12 months), the second taken for a single month. A stable rep will improve performance steadily over all stores, a "busy bee" will jump from store to store, producing excellent performance in one store while neglecting others, the "admiral" will solely concentrate on the mother store, etc.

For all these charts, one good visualisation is to use last Decembers values as starting point (say 100%), and plot all figures relative to that number. It's very readable, and removes the area effect, concentrating on how a rep (or other factors, naturally) evolves compared to others.

Finally, since many interesting figures involve a form of moving average, the cut-off effect should be mentioned. If there was one exemplary increase, it will "count" as long as it stays in the 12 month window. When it drops out, you notice a drop this month, although it's caused by a large increase last year. The solution to that is to use "decaying averages" or "decaying volatility".

Enough rambling, I just meant to write a paragraph or two...

(°v°)