This is a complete-the-passage item, so the credited choice is the condition the argument logically requires, signaled by the connector "only if." The chain observed a correlation: app-scanners spend 22 percent more per visit. From this it plans a discount to get more customers scanning, expecting average revenue per visit to rise. The hidden gap is correlation versus cause. If app-scanners spend more simply because the kind of customer who already buys a lot is also the kind who bothers to scan, then drawing previously non-scanning (and lower-spending) customers in with a discount will not transfer the 22 percent to them, and average revenue per visit need not rise. The plan therefore works only if the extra spending is caused by scanning itself rather than by which customers self-select into scanning.
(E) is correct: it closes exactly that gap. Negate it (the extra spending comes from which customers scan, not from scanning) and the plan collapses, since converting low spenders would not raise the average. That is the mark of the required condition.
(A) anchors on the eye-catching 22 percent and reframes the goal as covering a discount cost; but the stated goal is higher average revenue per visit, and the size of the discount relative to 22 percent does not establish whether revenue per visit rises. (B) simply restates the premise the chain already measured. A condition the argument needs cannot be a fact it has already stated, so (B) adds nothing. (C) over-reaches: it demands that every non-scanner convert, yet the plan needs only more scanners, not all of them, to lift the average, so this is far stronger than the argument requires. (D) sounds forceful but pivots to what rival chains do, which has no bearing on whether this chain's own average revenue per visit will increase. Only (E) supplies the causal bridge the conclusion depends on.