So, the KPIs approved by the customer for July were bolder than our wildest forecasts. It was necessary to increase the number of transactions by 3 times compared to the indicators of the New Year and March holidays - traditional periods of high sales. And as much as 17.5 times compared to July 2019. Revenue, respectively, also had to significantly exceed the holiday.

L'Etoile allocated a budget 3 times more than usual to achieve its goals. At the same time, we agreed that the SRO in July will also be higher, but we have designated the maximum limit.

We identify problems and solutions.

Firstly, our previous experience with L'Etoile (for example, we optimized the online store after the redesign) showed that July is one of the most "sleepy" months of the year. In summer, buyers are the least active.

Secondly, we had no idea whether we would be able to convince users who have never bought online to decide on this. After all, buying perfumes and cosmetics is associated with a sensual experience. Whether you like the smell of perfume, whether the consistency of the cream is pleasant, whether the lipstick matches the skin color - you can't get these feelings online in any way.

In addition, when the need to leave the house disappeared due to the self-isolation regime, it became completely unclear whether people would continue to use perfumes and cosmetics in the same volume as they did before.

From March to June, we observed an increase in the turnover of the online store: it was clearly higher than in similar periods of 2019. To some extent, quarantine helped us to complete the task. But none of us could predict how big the demand would be in July. I had to act at random and make decisions in a situation of complete uncertainty.

We use SRS channels and increase transactions.

For ourselves, we reformulated the client's task as follows: we need to find the point at which the ratio of advertising traffic and budget will be optimal, and stop in time.

The demand for the customer's goods online increased and the volume of advertising had to be increased in order to "catch" all new customers. If there were a lot of advertising, the budget would be spent not on new customers, but on overpaying advertising platforms for too high positions and too expensive a click. If we had shown less advertising or the usual volume, we would have missed some of the orders. We were looking for a balance. In addition, an uncontrolled increase in traffic and budget without reference to the SRO could lead to a critical drop in the profitability of the placement. And we did not want to waste the allocated huge amount of money.

As a result, we decided that we would use the budget for CPC channels, since attracting traffic with pay-per-click is an opportunity to flexibly and quickly increase the number of visitors and buyers. At the same time, from all available tools, those that will bring transactions in the first place were selected, so RTB and display campaigns were not connected.

We had only a month to decide how to scale the current advertising campaigns, whether we would try new techniques, and, if so, which ones. And how, with all this, to remain within the framework of the planned SRO. As a result, we expanded the usual set of tools in contextual advertising and relaxed a number of restrictions on traffic types:

► In Yandex.Yandex.Direct and Google Advertising have weakened the negative adjustment for mobile devices by 40%. The conversion rate from them is usually lower than from a PC, but we thought that with high demand this would attract additional potentially conversion traffic.

► Increased the volume of dynamic search campaigns in Yandex.Direct, smart banners, dynamic remarketing and smart shopping campaign in Google.

► We have connected new tools that have not been used with the customer before: Sales Discovery and Smart Display network campaigns have been launched in Google Advertising, and search and network campaigns with autotargeting have been launched in Yandex.Direct. However, in the case of autotargeting, they overreacted and chose a pay-per-conversion strategy.

The most daring experiment was the launch of autotargeting on Yandex search.Direct.

We allocated 23% of the advertising budget to this tool and received 14% of transactions from the total volume, as well as a decent increase in revenue. According to this particular channel, the SRO turned out to be twice as high as planned, but we were ready for this, and when we looked at the campaign's contribution to the final result, we were generally satisfied with it.

The network campaign with autotargeting worked much more modestly.

With the SRO 28% higher than in the autostr on the search, the tool brought less than 1% of the total transaction volume.

The Smart Display campaign accounted for only 1.7% of transactions from the total volume, and the SRO for this campaign was several times higher than the agreed average. Sales Discovery brought even fewer conversions, although the SRO was 15% lower than in Smart Display.

Tight deadlines have added difficulties. The algorithms of Google's smart trading campaign and the campaign with autostategies in Yandex.Direct need 1-2 weeks to retrain and take into account new introductory - desired indicators for SRO, profitability and expenses.

Because of this, in early July, traffic, budget and conversions were lower than we had planned, and this inertia made us pretty nervous.
The most difficult thing, really, is to watch and wait. In our case, until the strategies are trained and reach a plateau in terms of indicators.

Active work with smart banners also gave a good result. Previously, we used a retargeting audience and a limited number of feed product categories – only with a high average receipt or profitability. But in July, we added ad groups for all categories and set "Maximum reach (Retargeting + Look-alike)" in the audience settings. As a result, the number of transactions per month increased 7 times compared to June, but the SRO also increased by 33%.

Scaling dynamic search campaigns in Yandex.Direct increased the number of transactions by 2.36 times compared to June, and SRO increased by 86%.
The share of transactions from dynamic ROK increased from 4% to 8% relative to the total volume of all Yandex transactions.Direct.

In general, despite significantly higher indicators for transactions and revenue, in 2020, the SRO is comparable to the value of July 2019. That is, by attracting additional traffic, we did not sacrifice conversion, as is usually the case when scaling. Rather, they responded to the increased need of users to purchase goods online.

The SRO in July 2020 only slightly exceeds the SRO of July 2019.
The SRS increased in July 2020 by 2 times compared to the previous month and by 4.6 times compared to last year. However, we kept the SRO indicator by 23% below the permissible maximum value agreed with the customer. The total share of the new audience that we attracted through advertising increased to 37.71% in July. In June 2020, it was ~24%, in May - 16-17%. For comparison, in January 2019 it was 12.23%.

Perhaps the main result is that the effect of the campaign persisted throughout the year, although it lasted only a month. The number of transactions and revenue remained consistently higher than in 2019, and the share of online sales increased from 10% to 20% during the peak period. The July growth proved that large investments in online advertising are justified. Despite the fact that the budget of the following months was lower, it was still several times higher than the budgets of similar periods in 2019, and the task of attracting additional traffic became permanent.

For us, the most difficult thing in working on the task was to make a decision in conditions of uncertainty. In a typical year, the customer has pronounced seasons of high and low sales, which we could focus on and make forecasts. But no one could predict how buyers locked up at home for many weeks would behave.