Ecommerce Case: How to Recognize the Market Decline and Stop Draining the Budget on Contextual Advertising


No contextual advertising guru will save your business if you work in a falling market. We took up such a project: all macro indicators pointed to the fact that the conversion rate decreased twofold and the market sank by 70%. Let us tell you how we tracked the market decline, what growth points were found, and what came of it in this new case.

In Jam Agency we manage paid traffic for eCommerce projects. In 2021, we were approached by an online store for baby strollers. This story is about how we believed in ourselves to be able to do better than the market signals indicated, because we found a number of kinks and growth points in old advertising campaigns and in the end could not justify contextual advertising for the client. Were we deluded and defiant or made logical steps? Decide for yourself. Here is how we tracked the market decline, what work we did and what came out of it — and how you can understand that advertising will not have a net positive effect on your e-commerce project context.



The project is for a chain of regional baby stroller stores in CIS regions. Before the pandemic, there was a store in a large city, but now only a pick-up point remained. Despite lower prices and wider range than that of competitors, since 2016, revenue from online sales and, in particular, contextual advertising has been declining. The client could not understand whether the problem is with the market or the contractors. He wished to remain unnamed, but it’s not necessary for a failed case, right?!

At the start of negotiations, the client’s position “all contractors are bad” was alarming: either the client was difficult to find a common language with, or the causes were beyond the advertising campaigns. From experience, if the business model is working, then Google Ads will bring net zero or positive effect even with sloppy settings, and our task in this case is to find growth points, weaknesses and scale up the effective segments.

However, If the ad campaigns have a net negative effect, the issue is probably with the business model and the market. Therefore, first of all as part of the audit, we began to explore the market.

We start each project with an audit for two reasons:

  • We demonstrate our expertise right away on a specific project: we find weaknesses, show what exactly we plan to do and how it will affect revenue;
  • We dive deeper into the details to understand whether to take on the project and how soon it will pay off. The agency’s economic model is tied to LTV and pays off with long-term cooperation. It is good if in the first 1–3 months the project can operate at break-even. In fact, the agency starts making profit only after 3–4 months, since there is too much work at the start, especially in the e-commerce — we are oriented towards long-term projects.

Audit tasks are limited in time, so we always go from the general to the specific and dive as deep as the allotted time allows.

Market Analysis: Project Indicators are Falling Along with the Market, which is Being Taken Over by Marketplaces

Initially, you need to understand the state of the client’s business, and at this stage it is easy to do this by analyzing traffic in Yandex Metrika (Google Anlytics equivalent in CIS).

First of all, we found out that advertising clicks brought 20% of all traffic, therefore we will consider advertising channels later. Now let’s focus on the key sources.

Traffic sources

About 70% of the traffic comes from search. This means that the business model is working: “cold” traffic from the search is converted into an order or lead. But to make conclusions more accurate, we look at the trend over the last few years — and we find that since 2017 the amount of free search traffic has decreased by around 4 times:

Traffic trend from organic search

A drop in organic traffic indicates either a drop in demand or a decrease in the client’s market share. Another signal that indicates a decline in brand positions and a drop in the volume of orders is direct traffic. The graph is very similar to the one describing search traffic:


Trend of site direct traffic

The next step is to evaluate the trend of brand queries. To do this, we use Google Trends, which stores data for 5 years, and Wordstat (Google Trends equivalent in CIS), which provides information for 2 years:

Trend of brand queries

We see that the brand mentions declined by almost 3 times since 2016 and this correlates with a decrease in direct traffic and organic traffic. We need to check how things are going for direct competitors. To do this, we have collected the most popular sites in the organic search results for our topics over the past 1.5 years.



We singled out direct competitors from the list, looked at the trends of brand demand for them — and came to two conclusions:

  • The demand for the client’s brand decreases proportionally with the decrease in demand for competitors’ brands;
  • The demand for competitors’ brand decreases in direct proportion to the decrease in demand in the market — in general, the influence of direct competitors is minimal.
Trend of brand queries in comparison with competitors

It now remains to explore what is happening in the market. To do this, we study the queries for “stroller”: it includes search queries, which in most cases relate to our topic. We see that the market has fallen by an average of 2 times since 2016.

Trend of search queries for “stroller”

For reference: the value “100” on the chart indicates 100% — demand was at the peak in 2016–2017.

In order to clearly show the correlation between changes in market demand and site traffic, we superimposed graphs on each other. In recent years, demand has been declining not only by brands but for strollers in general.

Superimposing graphs of demand and site traffic

The most obvious reason here is the declining birth rate. And according to the birth rate trend in CIS, there has been a noticeable decrease since 2015:

Birth rate in CIS

Another hypothesis is the people’s inclination to save money and buy used strollers. Therefore, we studied the demand for the query “strollers + ads board name” as a synonym for “used strollers”. However, even here the demand has decreased several times:


Trend of demand for used strollers

The falling demand was compounded by marketplaces taking over the search results. We checked the demand for the stroller segment through Google Trends and Wordstat for “strollers + marketplace name” and got this data for search results:



For clarity, we compiled a table with the composition of organic search results by month, where marketplaces and aggregators are highlighted in green:



And in the last six months, another marketplace has been actively entering the market through contextual advertising:



For the convenience of understanding the overall picture, we combined all the data into a single table:



Also, for clarity, we built a graph with the main trends:



In total, we came to the conclusion that the market is shrinking and at the same time the share of marketplaces and aggregators is growing, which reduces conversion, since the already small demand is split between a larger number of players;

Despite this, the client had presence in dozen regions and had sales, and also there were competitors who had been using contextual advertising for a long time and, most likely, it paid off for them — so it was possible to work, and we began to study advertising campaigns.

Contextual Advertising Analysis: Campaigns Were Managed in Isolation from Business Indicators

Recall that the share of advertising in total traffic was ~20%, of which Google Ads accounted for 40% and with others further down, as shown in the screenshot below.


Since Google Ads was the main advertising channel, we analyzed data from this system. We studied the period from 2018 to 2019: data on conversions were not collected until 2018 and data on orders was available only for 2018–2019. In general, a pattern can be traced across all paid channels: in 2018, the company rapidly expanded coverage, but in 2019–2020 traffic returned to low values again. Perhaps, paid traffic was supposed to compensate for the decrease in orders due to the decline of organic traffic. And, judging by the graph, they were able to maintain visits but could not regain the revenue because advertising was turned off. We needed to figure out what was done, why it didn’t work and whether it was possible to fix the situation.


First of all, let’s exclude the traffic cannibalization: if we see purchases in the reports, we want to be sure that these purchases are “honestly” coming from contextual advertising, and not “stolen” from search traffic.

There is an easy way to check the level of cannibalization — plot a graph of free and paid traffic. In the graph below, we see that when contextual advertising is disabled, there are no sharp changes in organic search traffic, which means there is no crossflow.


Visits by paid and free traffic

Purchase graph for paid and free traffic

We Consider the Economics of the Project: What Should Be the Cost of an Order, Lead and Add-to-cart.

The client had no understanding of the cost of conversions and we had a suspicion that the previous contractors did not calculate the project profitability. Therefore, first of all, we figure out how much a paid order should cost and what are the intermediate steps on the way to the order taking into account current conversion rates in order to compare with actual figures.

However, an adjustment should be made for offline sales, since the client has physical stores in some regions, and among the visitors of the online store there are many who prefer to go to the offline store immediately after viewing the site to evaluate the goods in person and make the purchase. To do this, we use a correction factor, by which we will then multiply orders online to get their real number, taking into account offline sales.

To connect online and offline, we need an event on the site that correlates with the subsequent visit of users to the offline store. In our case, it is a visit to the page with addresses and contacts of stores, which makes sense, since if the user looks at the address of the store then most likely they want to go there.

The client has about 10 stores in different regions. According to the client, several outlets are visited only by users of the online store, since these outlets have low foot traffic and are located away from the city center. We take these outlets and compare their visits with visits to contact pages in these regions, and we see that every 3rd visitor to the contact page visits the store — that is 30%. The comparison is rough, but a tentative model is better than no model.

We will skip the detailed calculations. We’ll just clarify that for every online order there are 1.18 orders in the offline store (total of 2.18 orders).

The online store works throughout CIS, but not every region has an offline store, which gives us 2 data segments: without offline and with offline; to the latter we will apply our factor of 2.18 to get the total number of orders, and not just online:


Table legend: Cart — goods added to cart; CR_Cart — add-to-cart/click; CPL — cost/add-to-cart; Purchases — Purchases through the site; CR_Purchases — Purchases/Clicks; CPO — Cost/Purchase.

The problem is that the rate of conversion to orders of 0.3% is too low, at least 0.5% is needed for a quality study, because the data is “smeared” into a thin layer over a large number of parameters. Therefore, we will rely not on orders but on add-to-carts, and the cost of this conversion.

The break-even point according to the client is $65,5 per paid order. By multiplying the cost per order by the conversion rate from add-to-carts to orders we get the maximum cost of add-to-cart of $4,55.

Now we can have a quality evaluation of effectiveness of the old campaigns.

Effectiveness and Problems with Campaigns in Regions with Stores

The graph below shows costs and CPL (cost per add-to-cart), the red horizontal line at the level of $4,5 per add-to-cart determines the payback threshold.



We see that for the first three quarters advertising had a positive effect, but starting from the 4th quarter the margin went negative due to CPL rise to $10.

First of all, let’s check the seasonality, since this hypothesis is right on the surface. Perhaps the low season began, no one managed the bidding and therefore the CPL increased. According to historical Google Trends data, the high season comes in spring with demand bottoming out in October, which generally corresponds to our case. But the demand in Q4 compared to Q3 was changing by 10–15%, while CPL almost doubled. It turned out that the influence of seasonality, although present, was minimal.


We looked for the causes further, in the data for Q3 and Q4 2018. In our study we highlighted in red the campaigns that significantly affected the growth of CPL in this period:



If we look at the trends of this campaign vs total by quarters, we will see an increase in costs by ≈250%, and CPL by ≈200%:


But the main negative impact on advertising in general was campaign the New_search_stroller_shop_msk. In Q3–4 2018 and Q1 2019, it accounted for more than 70% of all costs and its CPL exceeded the acceptable limit by more than 4 times.



There are other campaigns that have caused CPL increase in the reviewed period, but their impact was insignificant, so we will not dwell on them in more detail.

Detailed analysis of the campaign “Search ads — Strollers — Cities” revealed that there were no large changes in CTR and CR, but CPC was twice the average:



The campaign worked under automated bidding strategy “Maximize clicks”. And although in our experience, this type of automated bidding strategies is not the best option to use, with the correct settings it can work effectively.

After checking the automated bidding settings, we found that they were constantly changing. Over six months, they increased the acceptable CPC limit and weekly budget from month to month. Since the change history is preserved only for the last two years, we couldn’t demonstrate what campaign settings had been earlier, but judging by the increase in cost-per-click, the initial settings were about 4 times lower.

In the end, in February, the settings were already as follows:



Most likely, no one tracked KPIs or they were calculated incorrectly. Therefore, the campaign was managed in isolation from business indicators. To manage bids effectively, they need to be calculated based on the cost of add-to-carts, and track how changes affect other target indicators.

In our experience, it is better to use an optimization strategy based on the target conversion cost or break-even cost.

After advertising was disabled in Q2 2019, it was restarted in the Q3 but with a different provider. And if earlier search campaigns accounted for the largest share of costs, in Q3-Q4 2019 display campaigns share was ~57%:



One can see with a naked eye that there are no orders on the site. This is probably due to the fact that orders were made offline, but even if the actual number of orders is greater than zero, with such an add-to-cart cost, these campaigns were unlikely to pay off. Remember that the maximum acceptable cost of add-to-cart is $4,55, and with an average cost of $7 advertising was loss-making.



Display network campaigns showed a trend for CPC gradual increase — a slight dependence of the increase in CPL and CPC on the decrease in the number of clicks can be noted:



The first suspect was manual increase of cost per click. However, the issue could also be with the platforms, therefore we considered them in more detail. The add-to-cart cost is consistently lower for all campaigns and it was worth removing the inefficient sites:



Effectiveness and Problems of Campaigns in Regions without Stores

This group of campaigns accounted for ~28% of the total advertising budget and generated ~18% of all orders: Here the allowable add-to-cart cost is slightly lower — $3. However, CPL for all campaigns is between $3,75 to $11,25.



Another feature of regions without offline stores was that the conversions to orders was 4 times lower: 0.1% vs. 0.4%. As a rule, in the regions, the conversion is affected by delivery terms, since there would always be a local competitor or a large aggregator / marketplace that would offer more favorable conditions or leverage an offline outlet. And later on, this hypothesis can be tested with an A/B test, where one audience is shown the current delivery terms, and another the more attractive terms.

Dynamic Remarketing is the Only Campaign with Net Positive Effect

In terms of profit margin, dynamic retargeting had a positive effect, because the cost of the order was below the acceptable $62,5:



However, this campaign could have been more effective. After checking the settings and the actual indicators, we found that the actual cost per goal was 3 times higher: $35 on the screenshot with the data against $12 set in the settings:





This happens when there are not enough conversions for automated bidding strategy to learn from. On average, there were less than 10 conversions per month, and you need at least 20–30. And therefore it would be more appropriate to use an intermediate goal with a sufficient number of conversions — for example, add-to-cart.

With the correct settings, the efficiency of dynamic remarketing could have been 15–30% higher, i.e. we would receive not 124 but 160 orders, for example. It is also worth considering that the weekly budget also reduces the effectiveness of the campaign, so you could realistically get 180–200 orders or more.

General Conclusions and Optimization Plans

Together with organic traffic, the number of orders and revenue decreases. There is a general drawdown in the market due to hard competition with marketplaces. And the only way to maintain revenue is to buy traffic. In the first 3 quarters after the launch, advertising was bringing profit, but due to campaigns being managed in isolation from business indicators, the profitability was sliding down. Restarts didn’t help, the same mistakes were made.



Theoretically, if the campaigns were restarted, but with correct settings and proper management, they could prove to be profitable. To understand whether this was expedient, we assessed the market potential for the main high-frequency keywords.

Due to time limitations of a free audit, we made a forecast without working through keywords inside high-frequency keywords. For example, we took the query “buy a stroller” and reduced the number of clicks by 30% to take into account the nested non-target keywords.

The forecast was calculated using the Google “Keyword planner” tool without taking into account the potential of dynamic campaigns and other eCommerce tools. We divided it into three parts so as not to mix the segments “with store” and “without store”. Also we segregated large cities so as not to distort the data. We built the forecast taking into account the current site conversion rate.

  • forecast by regions with stores taking into account the correction factor:

  • forecast for large cities: CPO is on the verge of break-even; for starters, it is worth launching campaigns with the most optimal indicators:

  • forecast for regions without stores with a negative profit margin.

In the end, without considering dynamic campaigns with Google (regions with stores and large cities), it is possible to get ≈100 orders at a cost of ≈$60. The market shows potential and search campaigns and display campaigns with the correct automated bidding strategy settings should be tested. The bottleneck of the current campaigns was definitely the bidding strategy: incorrectly chosen strategy, incorrect settings and management in isolation from KPI resulted in negative profitability.

Taking into account the information obtained and our experience from other eCommerce projects, we drew up a work plan with KPIs for the first three months. It is difficult to predict for a longer period, given how many variables and factors affect the result.

  • During the first three months, we planned to reach 70–125 orders with a budget of ≈$3 750 in the regions with stores and in large cities;
  • Running ads in regions without offline stores is not recommended: there is a high risk that ads will not pay off due to low conversion;
  • In the first iteration, we use dynamic campaigns, as they are more likely to show efficiency above classic campaigns, and with lower expenses at that.


Test Run in February-March 2021

The agency has about 40 active projects, 35 of them being online stores. In most cases, 50% of their revenue is generated by automated feed-driven advertising campaigns.

Launching such campaigns is easier and faster than the classic ones based on search or in social media. Therefore, we decided to start with automated campaigns on this project as well:

  • Dynamic remarketing in Google Ads and its analogue in Yandex Direct;
  • Google Shopping.

It makes no sense to launch advertising for the entire product range at once — we chose the most profitable segments: 30% of products that generate most of the revenue. We used the same products to create category pages for general queries, such as “3 in 1 stroller” or “luxury strollers”, since in addition to feed-driven shopping campaigns, we decided to launch classic search advertising campaigns.

Dynamic Remarketing

This advertising format enables products to be shown to website visitors who have viewed them or added to the cart. They look like product cards:


The campaigns were launched in a large city with a strategy of maximizing clicks at a given price — “maximum clicks for $0.5.” At the start, this is the optimal solution to get the campaigns rolling, get conversions, and only then switch to conversion optimization strategy. In our case, that is “maximum add-to-carts for $5.”

After the launch, we linearly raised the bids every few days in order to get more clicks and look at their payback. We used the brute force method to find the optimal cost per click and meet the add-to-cart cost limit of $5. This way we minimize the risks of draining the budget.

After 5 weeks, we received 44 add-to-carts, 5 leads and 0 sales. We decided to disable dynamic remarketing, since their cost of add-to-cart was 2 times higher than the allowable $5:



We reduced the bids for mobile devices, where the add-to-cart cost was 2 times higher than the average, we cut the bids for the male audience, turned off inefficient sites with low impressions… But we still did not reach the target indicators.

The client suggested we test regions with earlier high conversion rates. We looked at the demand, in total they would have given 5% of the large city volume, so we didn’t even test it. The smart banners in Yandex Direct work on the same principle, we tested them but they didn’t get any results just the same.

Google Shopping Campaigns

It is important to clarify that of the two types of campaigns — standard and smart — we first launch the standard one, where you can manage bids by product groups and cut off search queries. And only if it turns out to be effective, we launch a smart shopping campaign based on it in parallel.

Ads in Google Shopping campaigns are shown in search results and in networks by keywords, which are mostly generated from feed product titles.

Example of ads in the search result page

Campaigns were launched on the same principles as dynamic remarketing: by linearly raising bids and monitoring the add-to-cart cost. Despite the relevant search queries, the situation was similar: zero sales, several leads and about a dozen add-to-carts.


CPA trend for Add-to-Cart

Search Network Campaigns

After testing the previous tools, we decided to launch the narrowest possible campaigns for specific brands and models in order to eliminate unnecessary variables in the form of automated campaigns and see how relevant traffic does not pay off even with the best-selling models.

The client didn’t mind. Firstly, he saw good indicators of user behavior: visit depth, time on site, etc. Secondly, testing costs $250, and it is better to exhaust all possibilities and close the issue than to be left with the uncertainty.

But even this did not get us sales, only a few cart adds. At the same time, we observed that relevant visitors clicked through target queries to product pages with market-level prices, but there were no conversions.


CPA trend for Add-to-Cart

Our Miscalculations and How they could have been Avoided

With an average cost of our services for a client of $875 per month, to some extent we invest the time of our specialists in projects at the beginning of cooperation. In order to carefully assess the prospects at the outset, we conduct an audit and score projects according to our own criteria.

On the one hand, all macro factors indicated that the conversion rate decreased by 2 times and the market sank by 50%. And on the other hand, there was a period of 3 quarters when the campaigns worked stably, which means they paid off for the client. And we thought that if we balance the financial performance, eliminate all the weaknesses and use all the tools, we can reach optimal indicators.



In fact, much more was done than just launching the campaigns listed above: 60 working hours were spent on research alone. Operational time of the team lead accounted for another 5 hours, work with analytics 14 hours and work with advertising campaigns 37 hours.

In this case, we weighed the risks and the money and went for the project. We started from the fact that the client had a chain of stores, was a competent and responsible person, which meant that if we managed to achieve payback on advertising campaigns, we would get a stable client.

Moreover, there were several growth points, and to be honest, we were well immersed in the project and eager to go “to battle”:

  • Priorly, automated tools that have fundamentally different mechanics (smart banners, shopping campaigns, dynamic retargeting) had not been used;
  • There were fundamental mistakes in the current campaigns, which, even in good market conditions, would make advertising break-even or loss-making;
  • The cost of testing was $1250 — $2500 including the agency’s fee, which is commensurate with the client’s current revenue and risks.

Despite having extensively researched the market and competitors and found that external factors were evidently at play, we still took up this project. We believed in our strength, we went against the market — and failed.

This is a good lesson. It didn’t cost us as much as the client’s experiments with various providers. But the conclusions are such that despite the correct campaign settings, careful work with keywords and good traffic quality, the project financial indicators were unbalanced and were unlikely to be balanced without global changes at the business level.

We would like to get your questions and suggestions on how to work with similar situations?

Hope you like it and our experince can be helpful! If you have an online store and problems with Google Ads, then write to us, we will dive into the project and think about how we can help.

Get touch!

Write to us about your project, we will contact you within a couple of hours.

Previous
Case Studies
Online auto parts store
Online auto parts store

Online auto parts store

What was interesting about the project:

  • Made a matrix of coverage by categories of goods and car brands.
  • We picked up the correct attribution model.

Sales Funnel

$ 2300

Advertising budget for a month

8490

Clicks

467

Leads

Customer opinion

Сергей

«I came to 1jam.ru agency with the problem of high cost per lead and incorrect analytics, which we could not use to compare effectiveness of advertising channels with their expenses. The agency's experts developed a custom model to match conversion to channels, implemented electronic commerce, found problems in Google Analytics measurements and fixed them. What was exciting is that we got a matrix of our current and potential market reach with breakdown by car brand and spare part category, from which we drew a conclusion that we need to expand. Now we are systematically expanding the market reach and reducing the cost per lead. Contextual advertising brings a revenue of 400-600 thousand rubles per month»

Sergey Shumaev, Director, city-tuning.ru
opt@city-tuning.ru
Confirmation document

Online-shop of garden equipment
Online store of garden tools

Online-shop of garden equipment

What was interesting about the project

  • Investigated the seasonality of each category of goods and launched them sequentially.
  • Developed a KPI project taking into account the outflow of customers to offline stores.

Funnel

$2553

Advertising budget for a month

7500

Clicks

130

Leads online

30%

Offline revenue growth

Client's opinion

Владимир

«We started working with the agency in April 2018, the project team suggested we launch the categories of goods in sequence over the season in step with demand growth during the summer. One of our problems is that we have several offline stores for which we could not measure return on advertising investments. To solve this problem, we implemented call tracking, set up goals on the site and manually tracked the revenue trends by categories to collect the overall data and calculate the acceptable cost per lead by category. In general, I consider the task accomplished, we continue to work, I can recommend the team.»

Vladimir Frolov, Head of Marketing, technodacha.ru
v.frolov@astari.ru
Confirmation document
Read more

Get your Google Ads audit

After you leave a request: interview ~15 minutes → guest audit access ~15 minutes → audit within 2-4 days → proposal approval → first iteration start. In our experience, it is realistic to get the process rolling in 2-3 days.

Project manager Alex
Alex

Project manager



Write to — info@1jamagency.com or skype — jam.agency