What Georgia Teaches About Crossing the Digital-Market Threshold

The Georgian small firms in this series began with a problem that looked like a skills gap: internet access was widespread, but meaningful online selling was rare. The combined evidence points to a more precise answer to the question of why training alone may not move firms onto digital markets. For many of the firms studied, adoption depended on whether a concrete opportunity made the work worth doing now, whether the firm was ready enough to respond, and whether it could sustain the channel afterward. This final synthesis turns that pattern into a conditional judgment, while keeping the causal findings, mechanism interpretations, and limits of the evidence separate.

The answer is a threshold, not a single deficit

The first useful connection is between the baseline gap and the intervention results. Georgia’s small firms were not simply divided into businesses with internet and businesses without it. The important distance was between connectivity, an online presence, a completed order, and continued participation. That distance can contain several costs at once. A firm may need knowledge to set up a channel, a reason to expect a return, time to coordinate the work, and customers who make the continuing effort worthwhile.

The training experiment tested one part of that chain. It offered interested firms a three-day course intended to make digital-market participation easier. Only about 56% of firms assigned to the offer attended, and assignment to training did not produce a meaningful average improvement in the measured e-commerce index, its component indicators, firm size, or the later outcomes examined in 2025. The evidence is strongest as a statement about this offer under these conditions: broad instruction did not unlock adoption for the tested firms. It is not a statement that skills never matter, that every participant learned nothing, or that every targeted training program would fail.

The demand experiment moved a different term in the decision. It told eligible firms that a buyer would place an order if the firm could make the purchase happen online within two weeks. Assignment to the low or high offer increased the short-run probability that the team could find the firm online and place an order by about 26 percentage points. This contrast shows that the firms were not uniformly incapable of completing the task. At least some could organize a functioning online transaction when a visible buyer and a deadline made the payoff immediate.

The two results should not be reduced to “incentives beat education.” The interventions addressed different conditions. Training tried to lower the cost of capability; the conditional order increased the immediate value and urgency of acting. A firm can respond to a buyer because the opportunity changes its priorities even if the firm has not become more technically skilled. Conversely, a firm with a strong opportunity may still need help if it cannot create a payment route, present its products, or fulfill the order. The observed contrast tells us which change moved behavior in Georgia, not that one type of program dominates in every market.

The offer-size comparison sharpens the diagnosis. The low order was worth approximately $130, while the high order was worth $780, six times as much. The estimated short-run increases were about 22 percentage points for the low offer and 30 points for the high offer, and the difference was not statistically significant. The result makes a simple financial fixed-cost explanation less persuasive: if more money were the principal missing ingredient, a sixfold increase should have produced a clearly larger response.

But the low offer’s success means that money was not irrelevant. A modest prospective order may have covered some initial costs or made the first investment worthwhile. More importantly, it could have supplied a deadline that reduced postponement. Once a firm was close enough to readiness, a small, credible opportunity may have pushed it over the threshold. For firms with deeper constraints, the larger reward did not change the amount of coordination required. The same owner still had to complete the online work, manage payment and delivery, and make time for a new channel.

This is the strongest mechanism-level interpretation available from the pattern. The remaining barriers are likely to include behavioral and organizational frictions, such as procrastination and limited managerial bandwidth, rather than a simple shortage of capital or basic technical knowledge. Yet “likely” matters. The study’s framework organizes these possibilities; it does not directly observe each owner’s beliefs, workload, trust in the buyer, or ability to outsource the task. The result narrows the field of explanations without identifying one latent cause as settled fact.

Readiness explains why the average effect was not universal.