AdWords
1.8K members online now
1.8K members online now
For questions related to Google Shopping and Merchant Center. Learn to optimize your Shopping ads
Guide Me
star_border
Reply

Shopping Campaign Initial Success Dries Up - Repeatedly

Visitor ✭ ✭ ✭
# 1
Visitor ✭ ✭ ✭

I'm working with a company that operates a handful of eCommerce stores for different segments (some product overlap, B2B and B2C, some active at weekends, some more active on weekdays, etc). I've noticed a strange pattern that has affected each of the websites, several times. When I first set up a Product Listing Campaign, and now Shopping Campaigns, in the first two weeks, I get a lot higher conversion rate than I do, following that.

 

I've been working with AdWords for about a decade at this point. I do understand something about Quality Score, and Negative Keywords and the impact of bids. I understand that Titles, and Descriptions are very important for Shopping and Product Listing Adverts, and that there is impact from Product Categories to affect the search queries to which an advert is exposed.

 

I do not believe that this early success is because buyers are exposed to a new website. This company has been active for many years, and has been dominant in organic search for several years, for some of the websites. I won't believe explanations based on consumers exploring a new provider - the latency for the sales is mostly measured in hours or days, not weeks or months. In the time periods that I'm seeing this (about two weeks) most of those who started to search, will have been satisfied; in my estimation, two weeks is about four sequential generations of searchers for most of the websites. One of the websites has repeat buyers, with a less-than-one-year interval, but a more-than-three-months repeat purchase model. So this effect is not "OOOH, they're new, I'll try them". Established, multi-year business, supplier to the London Olympics, Commonwealth Games - multi-million contracts for supply to premium clients and a wide range of consumers, not a dodgy geezer operating from a shed, just set up.

 

The feeds are relatively static. I record the product feed, and I have a database of feed changes over the period. The products that do well, and then decline, are not being substantially changed in the feed. Typo corrections, some HTML attribute changes - nothing major, and sometimes no change at all (no Title, Description, Category, Type or Bid changes, no Landing Page changes other than "related products" and "featured products").

 

I can cite one product, in the last five weeks, in a new Shopping Campaign, where I've had 6 conversions, one for just 0.01. Since the first week, I've added exact match negative keywords, and I've seen the conversions drop. I've seen this same pattern, up to three times per website in different seasons/months, since I started putting together PLA Campaigns around this time last year, for this company - first time I've had one company with such a spread of eCommerce websites, common software platform, common feed system, etc.Previously I've seen this effect, but attributed it to site changes and feed changes. I now have more detailed monitoring and can match websites side by side.

 

Examples?

 

Set up Website 2 with PLA in September 2013, and it peaks by the end of September. Normal sales peak for this website is November (pre-Christmas purchase cycle). It should have gone on to do well. PLA did... modestly; nothing like the early sales peak suggested.

 

Set up Website 5 with PLA in late December 2013, and the site sells for two weeks, and then plummets. It wasn't site changes. What happened?

 

Set up Website 3 with a new weekend-targeted Shopping Campaign, to replace a Weekend PLA Campaign, and it does well for two weekends. And then falls off a cliff for conversion rates.

 

I can cite many more examples from across this small set of websites, just like this.

 

Initial success for the first week or two, followed by a much harder struggle to achieve sales - even though the impression volumes (adjusted for the exact match negative keywords) are the same or increased.

 

What am I adding Negative Keywords for? Exact matched negative keywords. Well, at 116 clicks on one search query, and no conversions, on B2C sales... I think most AdWords folk would be wondering about landing pages and whether the search query was right for the site/LP. OTOH, the same company can sell six conversions for one product in 120 clicks. The conversion funnel works; the right product presented to the right user on the right search can sell. IOW, I'm pretty sure that I'm picking the right negative keywords - we'd lose money if I let these high click volume non-converting search queries through.

 

What it *looks* like, is that Google understands where users are in the purchasing cycle, and is directing users closer to buying, to websites that it prefers. When a new Shopping Campaign/PL Campaign is put up, Google doesn't have the data to select impressions, but later, it does. Initially you get an unfiltered stream, and after a couple of weeks you get lower quality impressions, lower conversion rates. Or perhaps that initialy you get a test stream of impressions, and you then get an unfiltered stream. But there's something special about these early impressions.

 

This suggests that there is some Quality Score metric that not only affects whether the advert shows, but what stream of impressions is shown to a Campaign.

 

1/ What Quality Score Metric is available for Shopping Campaigns, and Products?

 

2/ Apart from randomly rewriting Titles, and Descriptions, what can be done to improve the traffic to get the higher converting streams?

 

3/ Am I alone in noticing this pattern, and if so, why? What are you doing differently that you see sustained conversion rates rather than an early two week peak? And would you mind sharing that? Smiley Happy

 

4/ I'm here because I don't think I reach people who deeply understand search when I talk to Google AdWords Agency support. I think they've been well trained, but this is not a normal issue. When I've tackled this, admittedly tangentially, I'm told sheer rubbish such as "you should only have one Shopping Campaign per feed because of the quality score". I don't trust advisors who don't understand that there is no useful shared quality score between, ohh, shoes and pillow cases, or why there is a Priority field. Neither shoes nor pillow cases are examples of what these sites sells - but the range on any one site is likely to be that diverse - very different needs; and splitting the feed is a perfectly viable tactic, especially if using the Shopping Campaign Priority - that's exactly why there is a Priority, to prioritise one set of products over another set, in the same feed.

 

1 Expert replyverified_user
1 ACCEPTED SOLUTION

Accepted Solutions
Marked as Best Answer.
Solution
Accepted by topic author Jeremy C
September 2015

Re: Shopping Campaign Initial Success Dries Up - Repeatedly

[ Edited ]
Top Contributor
# 7
Top Contributor


first, you're welcome.

typically, physical inventory is considered as well as the site content --
ownership is critical but differing business-models or the feed details

are generally not considered with respect to either duplicate (physical)

listings or the duplicate (similar) website policies; which can impact both

quality and performance over time.

 

the business-model is a separate quality factor.

certain aspects of b2b will cause items to perform poorly on google-shopping --
such items can perform poorly or be dropped over time depending on the details;
as just one example, direct-to-consumer shipping is required for all items.

 

note that these issues are separate and apart from the common-ownership

and similar details of the adwords (multiple-account) double-serving policies.

 

how feeds are registered, maintained, uploaded, and handled

with respect to physical inventory and the rules can all impact

quality over time.

description should simply describe the physical item or
how that physical item might be used -- nothing else.

like title, there are (character) limits related to
both strict requirements and overall quality --
these should both be considered.

also like title, description should be submitted without any html or similar,
promotional text, or non-standard capitalization -- using standard (english)
spelling, punctuation, grammar, and plain (ascii) text.

generally, promotional text is not allowed in a feed; promotional-messages can
be set under adwords or by participating in the merchant-promotions-program --
otherwise, submitting any promotional information in a feed, especially within
titles or descriptions, is grounds for lower quality or removal from the auctions.

any attribute that is submitted but, not following a guideline
or, any attribute that is not being submitted but, should be,
can impact quality; such affects are not always directly visible
or obvious but can impact performance over time.

depending on the number of products and performance statistics
across inventory, per-product bid tactics can sometimes perform
rather poorly over more strategic broad-based models.

custom-labels are usually used for more abstract or strategic bidding models --
such as bidding higher on high-margin items or lower on seasonal products.

note that relevancy is not considered with respect to campaign-priorities --
only the bid and quality factors are considered; priorities (are) promoted
as an alternative to pausing campaigns during the transition from regular
product-listing-ads to shopping-campaigns; certainly priorities can be used
to great effect when creating bid strategies around temporary promotions
tied to specific custom-labels, for example.

dynamic-remarketing is potentially a much more far
reaching strategy for targeting repeat customers --
than anything homegrown.

 

View solution in original post

Re: Shopping Campaign Initial Success Dries Up - Repeatedly

Participant ✭ ✭ ✭
# 2
Participant ✭ ✭ ✭
Hi Jeremy,

I am sure another user with more experience will be able to help you out more succesfully. But in the meantime it sounds like your ads are being given an initial QS based on other advertisers and then when your ad is establishing its own QS then your ads are dropping off as their own QS is much lower than the initial QS.

It is my understanding that the QS is based on your product description versus the users Search query and can be bolstered through product reviews. QS with regards to PLA ads is not available to you.

This all may have changed in recent times, so perhaps wait for a response from someone currently working with extensive Shopping campaigns but these are my original thoughts on the situation.

Hope it was helpful in some way!

Re: Shopping Campaign Initial Success Dries Up - Repeatedly

Visitor ✭ ✭ ✭
# 3
Visitor ✭ ✭ ✭
Hey Jeremy,

We have yet to experience any fluctuations in our clients' conversions/sales due to QS. I will give you my two cents on what may be some of the issues you are facing:

1) I cannot comment on QS because it really has not come up once for any of our clients however.

2) Focusing on the SKU level as opposed to product groups is how our clients are succeeding. There may be three products in a group (or example) and though the group is performing well, only one of those products is actually converting enough to be considered profitable. From there we help re-appropriate funds to high performing products and slowly weed out the poorly performing products. Over time we have been able to account for many issues such as product feed errors, on-page optimization and CPC bidding neglect.

3) I must stress that you'd have to run Shopping Campaigns for at least 90-120 days whilst performing SKU level bidding optimization and trying to measure actual values of profit or loss on the SKU level as well. Determining profitability and loss solely with AdWords/Analytics/Merchant Center and your in-house sales portal is not the easiest thing to do but it is how we have been helping our clients succeed.

4) The only time we speak with Google is when its regarding us as a Google Partner. I cannot comment on their ability or lack there.

5) Clynton already answered the QS question because its heavily reliant on Title + Description since Shopping Campaigns do not receive positive keywords, only negative.

Overall I would need to know more about the Shopping Campaigns you'll be running, how many products for each account, how you are managing your bids & budget and how you are analyzing and reporting on those Campaigns to provide you more insight into any issues that you may be facing. Shopping Campaigns did not do that much to PLA's since PLA's are still the same. Shopping just attempted to clean up AdWords and it added a bunch of (in my opinion) nonsensical features that are (in my opinion) of little use early on.

We've been running Shopping Campaigns for that past, I'd say 2-3 weeks since we are transferring all our client accounts over to the new Shopping format. I cannot comment on specifics but we built our own software to help us manage Shopping Campaigns in-house and we truly won't know the impact of the migration for at least another 5-6 weeks.

Re: Shopping Campaign Initial Success Dries Up - Repeatedly

Visitor ✭ ✭ ✭
# 4
Visitor ✭ ✭ ✭
@Clynton - thanks. If the QS was the issue, Then I'd expect to see a declining impression volume. What I see is that I get approximately the same volume of impressions. They're just worse converting. I even have a few examples where a search query has early conversions, and later... doesn't. Turns up at about the same frequency, but they stop converting. This is why I'm suspicious about determining intent... If we get early stage searchers now, but we used to get late stage searchers, it'd account for the variation. Maybe.

@Bryan - detailed response. Thanks for the time and attention!

Every SKU has its' own AdGroup. I use the API to crawl the feed (there's not enough change and volume to justify an active Merchant Center API feed - plain daily read, matched by my automation to collect new products and add new AdGroup, Product listing advert and Partition identifying the Google Product ID). These items, as presented in the feeds, are not groups. So, no product groups of any sort represented in the Campaign. Every item you see, is a product. It's about as much like the old PLA stuff as you can get. Smiley Happy

Yup, I know that old style PLA and new style Shopping aren't substantially different. API tweaks and UI tweaks. New features for bigger vendors, AFAICS.

Product counts? Each website has from 130 to 1500 products. Each website has its' own AdWords account. Each AdWords account also runs keyword search, mostly for generic terms, and a few specific product terms where we get better responses from the keyword than we do from PLA/Shopping appearances.

Bids - for the initial few weeks, usually kept pretty static. I have sales records for years, and Analytics account matching that. I have a pretty good idea of the normal conversion rates from various sources. These are not fast paced markets. Prices for the products are seasonal more than anything and competitors generally also slow paced SMEs. This is not, oh, flights (which I've also coded AdWords APIs for - 40,000 unique adverts a year, with pricing templates, before AdWords had templates, bid on 15 minute intervals).

They're mostly nice, slow paced, annual fluctuation wesbites, with periodic sales by all competitors, and mostly pretty average landing pages (us and them). We've not got an Amazon monstering the segment. Smiley Happy

I agree with the sentiment on analysis. Mine is aggravatedly complex - one of the sites sells most over the phone, on leads from the web. It does do some online sales, but what is shown as selling in Analytics is not always where the money comes from; have to maintain a healthy input of internal sales with weak tracking from PLA/keyword to revenue - we know it is there, we know what kind of percentage of the business it represents and we can basically only guess what might actually be working - margins aren't enough to justify full on phone tracking (done that, elsewhere/elsewhen) with unique numbers for each presentation.

Budget is always in excess of demand. I have a good idea of the weekly demand profile for each website over the year, and the impression share, and we make sure the budgets are in decent excess (20% or so higher) of any anticipated demand, while being low enough to effectively cap issues in time. Company is well established and knows that revenue comes from spend, if the spend is well targeted - so budgeting is pretty flexible, as long as conversion rates are sustained.

Analysis? Analytics has eCommerce tracking. AdWords and Analytics are linked. I get search-query reporting from all adgroups, daily, with conversions. I know that conversions and search queries lag a few days, so impressions and clicks from the last two days are excluded from analysis. So, I know (mostly - there's always a handful of missing data) which search queries turn into online sales. I know the repeat selling search queries, and the ones that Google gives us that no-one buys with... or not in economically viable quantity. I temper that with knowledge of internal sales (there is no lead generation, other than the websites - no hunting sales folk, just farmers).

What I think I'm getting, is that you see QS playing over a two week early period, based on Title/Description. And after that. we're tagged with an invisible QS that determines the quality of traffic we see. I can kinda see that. I'm not happy with the feed. The copy is taken straight from the website description. IMO, there are two different purposes for the copy and they should be two fields, one for the website description and conversion, and the other for PLA/Shopping and there to get the right person to see and understand the offer... before they click. So... yes, it's not, IMO, the best copy for the Merchant Feed. And without a QS, the only way to tell is to see that the conversion rate is assumed, and as knowledge is built, declines. That... makes sense. Is that what you're suggesting?

Re: Shopping Campaign Initial Success Dries Up - Repeatedly

[ Edited ]
Top Contributor
# 5
Top Contributor


generally, duplicate listings are not allowed; any item overlap whatsoever
violates current policies and is grounds for lowering quality over time and
eventually having the items removed from the listings, item or entire site
disapproval or suspension from the program, at any time; google has a one
listing per (physical) product policy -- regardless of the number of websites.

duplicate or similar website-content is also generally not allowed.

these issues can both have (negative) impacts on quality,
items being listed at all, and eventual status of items,
websites, and accounts in the program -- no changes
to the data (feeds), sites, or accounts are required.

of course, without specific information being specific is rather difficult.

(1) more generally, quality is based on many factors including
all rules, policies, and guidelines that are published by google
for the supply-chain, physical inventory, all data submitted,
and the website; google has indicated that other factors such
as ratings and the historic behavior of all users over time,
clickthrough-rate, etc., are also likely involved.

importantly, each type of item may have different rules and policies.

currently, there is no such directly correlated quality-score that is
displayed in the account for product-listing-ads (shopping-campaigns).

however, measurements such as clickthrough-rates, impressions,
clicks, etc., can be seen and are general, visible, quality indicators.

(2) most all data submitted and the website can impact quality.

certainly, if there is any overlap of physical inventory
or the websites' content, this can eventually negatively
impact quality and depending on the details will be grounds
for disapproval and suspension, at any time -- if so, such
issues should likely be addressed to both improve quality
overall and reduce the risk of eventual policy issues.

of course, title and description are important; however, about
the only attributes not indexed for search and not directly tied
to quality are those related to id; all other attributes can impact
quality both in a positive and negative direction.

the relationship between the data submitted, the rules, policies, and
guidelines, the types of (physical) items, and site (landing-pages)
can also be critical.

(2) again, most all attributes are involved in quality.

generally, consider each rule, policy, and guideline, for the item supply-chain;
then, consider each rule, policy, and guideline for the website and landing-page;
then, consider the rules, policies, and guidelines, for the type of physical term;
then, consider the rules, policies, and guidelines, for each attribute.

in addition to changes or additions, consider removal of
information that is even slightly outside the guidelines.

for example, the title must simply name the physical item
without any promotional text or excessive capitalization --
using standard (english) spelling, punctuation, and grammar
with plain (ascii) text.

there many other rules, policies, and guidelines -- just for title.

for example, there is a limit of 150-characters
that is critical for quality (search relevance);
but, only the first 40-70 characters or so may
be seen in most contexts within an ad display --
so the first 40-70 characters is likely critical
from that perspective.

as just one example, other than title; variants cannot be submitted
unless identical item_group_id values with at least one other unique
variant attribute are submitted per variant group -- otherwise,
submitting variants is a policy violation and grounds for lower
quality or such items eventually being removed from the listings.

id and item_group_id can (negatively) impact quality indirectly --
such as any item changing id or not submitting item_group_id
with proper values for variant products.

attributes such as custom_label_0 can impact how the campaign
might be structured, therefore the bid, therefore performance.

attributes such as google_product_category, product_type,
image_link, the images themselves, link, the link landing-page,
condition and availability and their accuracy with respect to
the physical item and the landing-page, also price, any and
all global-trade-data (gtin, brand, mpn), identifier_exists,
gender, age_group, color, size, shipping, shipping_weight,
multipack, is_bundle, adult, and all other attributes can
all have both a positive and negative effect on quality --
depending on the item details.

google may add or remove attributes or change
which values are allowed or considered inside the
guidelines -- generally, all attributes in all feeds
should be reviewed (hand-inspected) periodically.

google is constantly checking and rechecking
(feed) data and websites for quality; not simply
during feed processing.

all attributes with respect to the website landing-page
can impact quality -- for example, any mismatch between
any data being submitted and the landing-page, inventory
status, or the physical item.

(3) of course, both the bid and quality factors impact ad-rank and performance.

importantly, other merchants can effect performance --
merchants improving their (feed) quality or optimizing
bidding strategies can impact performance over time.

generally, having multiple websites or business-to-business sites should not
be dismissed as a potential negative impact with respect to google-shopping.

(4) data (feed) details and how feeds are structured and handled, can
certainly impact performance and quality, both negatively and positively,
especially over time.

since google can look directly into accounts
there may be certain insights that can be
garnered that forum members cannot deduce
from the information posted here in public.

of course, product-listing-ad campaign strategies can also impact performance.

more recently, campaign structures, bidding strategies, feeds and per-item
details, with respect to any transition from regular-product-listing-ads to
shopping-campaigns for product-listing-ads, can also impact performance.

Re: Shopping Campaign Initial Success Dries Up - Repeatedly

Visitor ✭ ✭ ✭
# 6
Visitor ✭ ✭ ✭
Thanks for the detailed reply, Celebird.

I'll be briefer now (end of my workday, car park emptying and a restaurant calls my name forlornly).

Dups - there are a few products represented on two websites, but they target different audiences, with different shipping arrangements, and different titles and descriptions - e.g. B2B and B2C - where B2B expects 9-5, M-F delivery and B2C... doesn't; also where specific product features are important to one and not the other segment. Pricing, shipping, title description - they're effectively different businesses with a shared office backend.

Hear what you say on Title and Description. I'm not convinced this feed is good for Descriptions. Title, it has nailed. Description - not so good. If that's two weeks to play out... This collapse in performance could be the signal that we really need to address the Descriptions. Category - we can't do much about. These products don't offer a lot of choice about where they sit. Product Type - we've played with; makes a small difference, IME.

Since the Campaign is structured around the Google Product ID/SKU, not groups of products... don't see Custom Labels having any impact... Except that I have a cunning plan to use a Higher Priority Shopping Campaign for repeat sellers, and promotional items, and those products will be identified by signals in the feed... a Custom Label signal. Smiley Happy

Thanks, again, for your insights!
Marked as Best Answer.
Solution
Accepted by topic author Jeremy C
September 2015

Re: Shopping Campaign Initial Success Dries Up - Repeatedly

[ Edited ]
Top Contributor
# 7
Top Contributor


first, you're welcome.

typically, physical inventory is considered as well as the site content --
ownership is critical but differing business-models or the feed details

are generally not considered with respect to either duplicate (physical)

listings or the duplicate (similar) website policies; which can impact both

quality and performance over time.

 

the business-model is a separate quality factor.

certain aspects of b2b will cause items to perform poorly on google-shopping --
such items can perform poorly or be dropped over time depending on the details;
as just one example, direct-to-consumer shipping is required for all items.

 

note that these issues are separate and apart from the common-ownership

and similar details of the adwords (multiple-account) double-serving policies.

 

how feeds are registered, maintained, uploaded, and handled

with respect to physical inventory and the rules can all impact

quality over time.

description should simply describe the physical item or
how that physical item might be used -- nothing else.

like title, there are (character) limits related to
both strict requirements and overall quality --
these should both be considered.

also like title, description should be submitted without any html or similar,
promotional text, or non-standard capitalization -- using standard (english)
spelling, punctuation, grammar, and plain (ascii) text.

generally, promotional text is not allowed in a feed; promotional-messages can
be set under adwords or by participating in the merchant-promotions-program --
otherwise, submitting any promotional information in a feed, especially within
titles or descriptions, is grounds for lower quality or removal from the auctions.

any attribute that is submitted but, not following a guideline
or, any attribute that is not being submitted but, should be,
can impact quality; such affects are not always directly visible
or obvious but can impact performance over time.

depending on the number of products and performance statistics
across inventory, per-product bid tactics can sometimes perform
rather poorly over more strategic broad-based models.

custom-labels are usually used for more abstract or strategic bidding models --
such as bidding higher on high-margin items or lower on seasonal products.

note that relevancy is not considered with respect to campaign-priorities --
only the bid and quality factors are considered; priorities (are) promoted
as an alternative to pausing campaigns during the transition from regular
product-listing-ads to shopping-campaigns; certainly priorities can be used
to great effect when creating bid strategies around temporary promotions
tied to specific custom-labels, for example.

dynamic-remarketing is potentially a much more far
reaching strategy for targeting repeat customers --
than anything homegrown.

 

Re: Shopping Campaign Initial Success Dries Up - Repeatedly

[ Edited ]
Visitor ✭ ✭ ✭
# 8
Visitor ✭ ✭ ✭

Time for a summary of what I've done. What worked, what hasn't.

I've had some copywriters working on the product descriptions in the feed. That has been phenomenally successful. The original website ecommerce system only allowed web page contents through. I had the developers abuse the SEO meta description, and use that in the feed.

About two weeks after one website was altered to allow the SEO description to be used in the feed, and the copy was rewritten to be brief and factual, the performance picked up. It wasn't the quantity of impressions that changed markedly or even the breadth and nature of the search queries, but the quality of the searchers. I think that some kind of QS metric for Shopping/PLAs would probably help advertisers a lot - it simply wasn't clear that this was the problem.

I'm using the API, so I have one product per AdGroup. I manage bids separately for each product - which has proven to be really helpful. I have products that share what would be product groups where I really do want some search queries to be shown to one of the products and not to the others, and where very different bids result in maximising net profit. Separating the products made that much easier to manage.

Thanks for your pointers - all of you (Celebird, Bryan, Clynton).

I think the most crucial thing was being able to convince the company that the copy needed attention. And it is working so well that we're now working on the copy on all products on all the websites, and seeing good results about two weeks after the rewrites.

Not quite sure how to spread the kudos - you each helped! But I think I'll tag Celebird, as the most persistent and detailed respondent.

Shopping Campaign Initial Success Dries Up - Repeatedly

Visitor ✭ ✭ ✭
# 9
Visitor ✭ ✭ ✭

I noticed this too. I have 5 completely different ecommerce stores. The shopping campaigns all do very well with conversions at first, but after 7-10 days the shopping campaing conversions go to zero! Horrible.