Have you ever lost a pile of cash? At Uploadcare we did, to the tune of USD 50K on Google Ads. But rather than wallow in buyers’ remorse (that’s almost the price of a Tesla, right?), we decided to share our pain and insights so that you won’t make the same mistakes.
The irresistible allure of paid traffic for mature products
Uploadcare is an established SaaS product, so we’ve been harvesting the low-hanging fruit from organic and direct traffic for a while now. If you’re in a similar situation, you’ve probably reached the same conclusion: running paid ads is the next logical growth step for customer acquisition. But there can be serious hiccups if you don’t get some key things right, especially if your product is a more complex B2B solution like ours.
What we knew before we launched our Google Ads
Since we’re out $50K, you might be thinking, “Oh, these guys just dove into this without doing enough planning.” That’s the cautionary part of this tale. We had done a fair bit of planning, but we made some key assumptions that came back to haunt us. Here’s where things stood before we launched the Google Ads campaign.
Solid inbound traffic and net negative churn, check
Thanks to developers’ trust in our core infrastructure and their recommendations, we’re fortunate to have a constant and growing inbound flow of leads and net negative churn, meaning the value of usage-driven upgrades outweighs the loss in revenue of subscription cancellations. Seems like a perfect use case for investing in Google Ads to bring more folks to our funnel, right?
A clear funnel with established conversion rates, check
Thanks to Kissmetrics analytics, we have clear visibility into our funnel, including baseline conversion rates for both signups and activations from direct traffic. Our product is a file handling API, so when building our digital funnel, we chose “uploading at least one file” to be our activation metric. Conversions for New Visitor -> Sign-Up and Sign-Up -> Activated were at 3.6% and 30.8%, respectively.
It’s worth noting that we test different hypotheses to improve conversions at every step of the funnel. Google Ads campaigns were explicitly aimed at getting more folks to visit our site. Then, there are experiments with value propositions and main page layout to improve the New Visitor -> Sign-Up conversion. The hypotheses related to enhancing the Sign-Up -> Activated conversion are about improving our onboarding flow. The last two are out of the scope of this article, but let us know in the comments if you’d like to learn more about formulating value propositions, laying out the front page or improving onboarding.
A customer awareness ladder to guide keyword segmentation by purchase stage, check
Another thing that we built was an awareness ladder. We invested time in understanding where certain kinds of customers were on the ladder and what specific pain points they had. Based on that analysis, we built an extensive semantic core that included keywords and combinations for every step of the ladder and every customer type. Here is the snapshot of the ladder for the “File Upload PHP” keyword subset, spreadsheet.
The table above represents the first two steps of the awareness ladder snapshot for “File Upload PHP.” Before getting to the next article section, note the two important things: we tried creating a landing page that would serve as jack-of-all-trades; we looked at the Sign-Up -> Activated conversion as one of our target metrics. Leaping ahead, that was a bad decision: it is a far-off metric to look at since it takes users way more time to activate within a product than click a banner or even sign up.
That was just a single lesson we learned, here are some more.
Lessons learned
Whether we’re talking about startups or PPC campaigns, execution really is everything. No brilliant idea or rock-solid plan is going to implement itself. So while our planning was A-OK, things fell apart due to a few key details. Here’s what we learned (the hard way):
Lesson #1: Test and trim keywords sets before hiring an agency to scale things
We decided to work with an SEM agency to simultaneously launch a whole bunch of Google Ad campaigns. That proved to be costly since agencies typically start with your extended keyword set and optimize the budget over time by narrowing things down and excluding inefficient keywords from the campaigns. Instead, we should have taken a test-and-learn approach internally with small budgets to weed out inefficient keywords before handing things off to an agency.
Don't forget we looked at users signing up and activating before making a final decision on this or that campaign quality. Every hour spent waiting for folks to sign up and activate meant losing money on expensive clicks.
Lesson #2: Focus on page quality and CTR when doing paid tests
Two ideal testing metrics for paid campaigns are landing page quality and clickthrough rate (CTR). Why? Optimizing your Quality Score lowers your cost per click (CPC) and cost per conversion, and CTR is the most important component of Quality Score. Not to mention that measuring CTR and landing page quality is way faster than waiting for conversions to happen within your product.
This approach means you need to create that on-page content first because you can't really test landing page quality any other way. Sure, you could run paid tests on the quality of your competitors’ pages to see what works before you create your own. But you’ll still need to run tests on your own landing pages to optimize for CTR on your domain, because Quality Score factors in your historical Google Ads performance, among other things.
Once you establish baseline page quality scores for your content, you can further optimize your landing pages by testing a few different content hypotheses to see what will boost the Quality Score. Once you’re happy with the quality and CTR metrics, you’re ready to ramp up your spend.
Lesson #3: Stick to keywords that you have landing pages and content for
From the get-go, we should have axed keywords for which we had no landing pages and content since ‘orphan’ keywords that don’t have a home in your page content aren’t going to yield great CTRs.
More to that, you should at least a single landing page for every other step of your awareness ladder. We have already told you about our jack-of-all-trades landing: even though we thought it to have content for everybody, it best suited the third level of our ladder, see the spreadsheet. We got it when we figured that cost per lead was ten times below the baseline for a specific subset of keywords.
Segment your extended keyword set according to the actual content you already have. Some of those existing pages can serve as landing pages, depending on how optimized they are for conversion.
Set aside keyword groups that aren’t well represented on your site. For example, although our solution is a good fit for mobile platforms, there was no dedicated landing page content to demonstrate this to site visitors. Until we have that in place, mobile-oriented keywords are best to wait.
Lesson #4: Don’t assume organic conversion rate will hold true for paid
Remember those conversion rates from earlier? For paid, one would assume they would be equal to, or better than, non-paid traffic because people show intent when they click on an ad, right? Well, that did not turn out to be the case, not even close. Conversation rates from paid campaigns were significantly lower than the baseline.
The reason? Our old nemesis, Quality Score. Specifically, low landing page quality for existing campaigns. Even with a decent clickthrough rate, poor page quality meant a cost-per-click of about $0.97 and a customer acquisition cost way above one-third of customer lifetime value.
In other words, the suboptimal conversion rate was scaling the quality failure: regardless of how much we would spend, we’d only get losses in return. And because the paid conversion rates were so far off our baseline rates, we wasted even more time and money on making decisions about the quality of inbound traffic for the campaign.
Lesson #5: Analytics will save (some of) your bacon
As bad as things were, they could have been much worse. Without analytics in place, we wouldn't have had a clue about what campaign leads did on our site and which elements they interacted with.
Naturally, you’ll want to be sure your analytics are fully optimized with reporting configured to suit your needs before you throw paid traffic into the mix. If you don’t spot something with organic traffic, it’s just a missed opportunity. With paid, the first thing you may notice is your spend spiking, at which point the damage is already done.
Lesson #6: Revisit your awareness ladder often to validate and update it
As we got deeper into our campaign, it became clear that our awareness ladder was structured wrong. Here are the two main issues:
- The keyword groupings we associated with each step were irrelevant because they weren’t in line with how people would perform their searches. We figured we had to shift our keywords sets one step down (the example awareness ladder in this article already comes with the fix). I.e., folks are already aware of the problems and search for a solution in form of a library or service. So, there are two branches of competition: “library vs. DIY” and “library vs. library.”
- Content. We believed that a single landing page could work well for different steps of the ladder. That’s fundamentally wrong, and our tests (see below) showed the landing page better suits the L3 of Awareness Ladder.
In short, any awareness ladder is a dynamic concept that you’ll want to revisit and optimize based on campaign insights, or even restructure completely from time to time.
$50K of lessons learned, summarized
Let’s recap our lessons learned:
- Narrow down your extended keyword set and focus on the keyword groups you have polished content and landing pages for, especially when you have a complex product.
- Use an awareness ladder to inform keyword segmentation by purchase stage but revisit it often to validate and adjust.
- Do not use the same landing pages for different steps of your awareness ladder.
- Do not wait for leads to become paid users to decide on the quality of paid ads campaigns: focus on quick metrics and tailor experiments to one step of your funnel at a time.
- Take a test-and-learn approach with small budgets to quickly fine-tune campaigns, focusing on page quality and clickthrough rates.
- Run tests to optimize page content, which will reduce your cost per click.
- Once you find your winner, you’re ready to go all-in. Now’s the time to pass it off to an agency to scale things up, if you are contracting out campaign management.
In the end, we dialed down our spend and did a series of quick paid tests, which helped us boost ad results to three times our baseline conversion rate. Specifically, we figured our content works best for the L3 of the Awareness Ladder. The landing page was better structured to help folks that were already aware of the problem and existing solutions and would look for a ready-made plugin, i.e., “image upload plugin” or “jQuery image upload.” Take a look at the Search Ads examples below.
After optimizing for Quality Score and CTR/CPC we scaled things back up again and started looking at other steps of our funnel. Next time we feel like burning some cash, we probably should just throw a big party and invite all our Twitter followers.
What have you learned from failed marketing projects? Jot down a comment below.
Anton Elfimov is a Data Analyst at Uploadcare, a file-system-as-a-service that does uploads, storage, and media processing for Web and mobile apps, so you can ship products faster and scale them painlessly. (Your mom says you should try it; she’s worried you’re working too much.)
I’m a Xoogler with 5 years tenure in the SEM team. I’m shocked to read the lessons you’ve learnt after the loss of money. They should’ve been a prerequisite before even thinking to approach Google Ads but I’m not surprised anymore after what I’ve been experiencing here in Los Angeles with so called experts and media agencies, not to mention the lack of technical knowledge on data and analytics.
Hi Gennaro, agree with your points. Do you know an agency that cares about knowledge of data and analytics? We believe that consulting agencies can help us study the product, not SEM. Isn't that SEM agencies work to just scale campaigns and help configure them?
Very interesting article. If I may ask, for how many months were you running those campaigns? I mean 50k is quite a bit of budget so I'm curious how quickly did you burn through it. I strictly not advise my clients to start advertising with throwing big budgets...but starting slow. As always slow means different things for every vertical.
If I may cheer you up, you don't have full control over QS (quality score). Especially the CTR part which is estimated on the average of every AdWords account that is bidding for this particular keyword.
Hey Kamil, great to see you on the thread. We burned through the budget in 2-3 months. We started with way too many keywords and way too few landing pages.
We didn't face any significant competition and were able to get rather low CPCs. Would you recommend us with any further reading regarding the CTR part estimation in relation to other keyword bidders?
Hi Anton,
ok, let's break it down.
Let's assume the average CPC was $0.68
If you burned 50k in 2-3 months...let's assume it was 3 months you were burning per month 16.6k
So you were getting 244118ish clicks...
What can we conclude from this flood of clicks?
Statistical significance. You had enough data to make decisions much sooner than after 16.6k. Over 244 THOUSAND clicks (in a month) is enough data to make decisions.
I mean, you focus on QS, CPC, CTR and other metrics concerning purely AdWords.
As an AdWords guy myself, I can only say the harsh truth..I don't care about them.
Wow, I said it..you may wonder who the hell is this guy.
Give me a sec to explain myself.
I don't care about those metrics until the moment I have a proven flow of conversions.
If and ONLY IF we have a flow of conversions..that's the moment to squeeze more at a lower cost.
I mean, yes you can spend days writing super crazy good copy to increase CTR on the ads...but if with 244k clicks in a month you can't do too many conversions...what's the difference if you get additional 50k clicks?
Push QS up and get more clicks...at lower cost...clicks that...don't convert?
I hope you get my point.
It's a numbers game. Focus on flows, on getting profitable.
When something is proven..that's the time to scale (CTR up, CPC down, QS up).
Why even bother?
It's like making a beautiful foam on a coffee when you forgot to put in coffee and you have hot water.
We call it a premature optimization. Always a bad thing (even in coding!)
Btw great article. It was a pleasure reading it.
Kamil, thanks for the breakdown.
I got your point regarding premature QS optimization (CTR up, CPC down, QS up). I believe it wasn't premature: the SEM agency was up to optimizing things while monitoring stats. The problem was rooted in the great number of campaigns and keywords: the huge number of clicks was split into hundreds of ads. So, we had to wait for Signed Up - Activated conversions to decide on the traffic quality from individual campaigns and keywords.
Sure Anton,
we can't change what happened.
I just can't stop wondering what would happen with a much more limited keyword selection focusing on the lowest hanging fruits based on the persona metrics from organic traffic.
Yes! I'm currently in the early-stage of my startup journey and I'd love to know how you set it up, how frequently you test your assumptions, etc.
Hi there, that's a super broad question :) I'd say you should start with defining a clear goal: which metrics would you want to influence to get traction. There's no frequency of testing assumptions, I'd put it as "rate of testing hypotheses," where you aim for quality and then quantity. Implement a hypotheses prioritization framework such as RICE and track your way through them via something like HADI:
— RICE: https://www.intercom.com/blog/rice-simple-prioritization-for-product-managers/
— HADI stands for Hypothesis-Action(ToDo to implement it)-Data(to check if you were right about its impact, etc.)-Insight(was it a quality hypothesis, maybe there were outcomes you did not expect, etc.)
Cool article. We've had similar issues with Paid Search where I currently work except our issue was that the space was far too competitive and effectively priced us out of making any profit as we didn't earn a high enough return and downstream conversion rate for any leads.
Yeah, it's more or less understandable when your ad space is competitive. In our case, the entry conditions seemed perfect: 46.3% direct/28% organic, previous success in reaching out to the developer community, etc. "Let's run some ads!" Technically, we didn't have enough resources to craft tight landing pages matched with tight keywords. You may find the HN thread of the article useful btw, there are many good points: https://news.ycombinator.com/item?id=18904625#18905182
You forgot the most important lesson: Vet your SEM agency. These are all rookie mistakes, and that's not your fault (that's why you hired an agency, right?) But they should never have let you learn those mistakes with your own money. It's their job to coach you, not the other way around.
It seems that product analysis is more of a consulting nature. The SEM agency helped us create and optimize Google Ads campaigns, but our product is a complex technical platform, and it's not easy to understand what keywords will give better ROAS for us. We write this point in Lesson #1. The insight is that we were wrong focusing on features, not value propositions (like, "File Upload PHP" is a feature, not a value prop)
Why do you think that a good SEM agency wouldn't be able to tell the difference between a feature and a value proposition? This is like saying you can't expect a running coach to know that hydration and warming up are important before a big run.
I understand that your product is technically complex, and forecasting ROAS for keyword themes may not always be possible.
But a good agency would have challenged your models and assumptions right out of the gate, because it's not a given that organic performance translates to paid performance. Different possibilities must be considered in advance, and it sounds like they didn't even mention that.
They would have built a strategy (or "awareness ladder" as you're calling it) that maps actual keywords to actual LPs with a focus on the buyer. The spreadsheet your team created doesn't do this (again, no offense - this is the job the agency, not you!) They would have also kept you in the loop about low QS issues and worked with you to combat it so you were able to pivot before it became a $50K "lesson."
They would have not expected you to hand them the right keyword set; they would have done the work and research to know what keywords to start with, and optimized from there based on the data. At best they should have thanked you for your additional contribution and gone to work, but they should never have "relied" on you to provide this.
They would have insisted on / demanded / facilitated the creation of landing pages that speak to the needs expressed in keywords and search terms. That you had to "learn" this lesson honestly baffles me, since even beginning SEMs know enough to say "dedicated LP" even if they don't know how to tell a good LP experience from a bad one.
None of this is your fault. But if you don't learn from all of this the role an SEM agency can play in your success or failure, I'm worried you might learn similarly avoidable lessons all over again in your next round of advertising.
And quite often you would find something like "do not outsource to an agency since it's not worth it" https://news.ycombinator.com/item?id=18911159
We're trying to find an optimal balance between in-house test/consulting/outsourcing
I really like the idea of SEM agency challenging our models, however, we connected with a good number of those and none would say something aligned with the concept. Maybe there are some "magic words" you say to an agency for them to understand you need a keyword and segment-matching research from them. Would you recommend such an agency btw?
P.S. On the HN thread the comments like "you should've performed the research on your end are not uncommon": https://news.ycombinator.com/item?id=18904911
I think there will always be tension between "DIY so you can solve your own problems" and "find someone great to solve your problems so you don't have to DIY." Doesn't matter what industry or sector, there's just room for both approaches!
Regarding the "magic words" you can ask an agency, here's a checklist I put together that might help get you started: https://paidsearchmagic.com/wp-content/uploads/2018/04/How-to-Choose-a-Paid-Search-Agency-_-PSM.pdf
I'd pay attention to things like "what sort of things tend to go wrong when a campaign doesn't work?" and "what weaknesses, opportunities and threats do you see for us to start running on Google Ads?" The quality of those answers will let you know whether they're worth moving forward with.
Any marketing agency (or sales team at a marketing agency) can paint a pretty picture of possible rewards, without ever mentioning possible risk or what can happen when things go wrong. But things can and do go wrong. Expectations don't always get met. If everyone you're talking to only knows how to speak to the upside, you need to keep interviewing.
Thanks a lot, got your points. We'll look through the checklist and definitely include those in the upcoming interviews.
Ouch. I'd be curious to see what kind of landing page(s) you were using for your product. As a landing page, I'm biased, but it seems to be one of the root causes of your problem.
Matching landing page content and visitor awareness is one of the keys to that process, and I go over that pretty extensively in the Landing Page Cookbook.
I'm also surprised you wouldn't outsource keyword selection to the SEM agency, as that's also a step where you can add a lot of noise involuntarily, while specialists should be able to orient you towards keywords with better ROAS potential.
Hey David,
Yes, I agree with you! One of our mistakes is that we use only one landing page for different levels of visitor awareness. We speak about it in the article. Now we know that first we should do good landing pages with content that relevant to visitor's awareness.
We test different variations of landing page. Here is the last version https://uploadcare.com/lp/file_upload/widget2/. Thanks for your book, I will read it soon.
SEM agency help us to create and optimize Google Ads campaigns, but our product is a complex technical platform and it's not very simple to understand what keywords will give better ROAS for us. We write this point in Lesson #1.
Feel free to ask any other questions. Thanks!
First of all, I am sorry to hear that.
Just from a quick look at the Main Funnel, it doesn't seem to be a problem of Google Ads. Rather, the conversion rate was too low.
Question: why didn't you check the conversion rate after the first 10k spent to make sure it was a good strategy?
P.S. I know, it is easy to appear as experts right now, by any of us.
Hey Batlle,
The Main Funnel represents a baseline for rather mixed traffic. We were getting way better rates from certain keywords/referrals. And even with the baseline, we got A-OK results. So we decided if mixed traffic works, then targeted keywords would be much better.
And they were. The problem is largely with the jack-of-all-trades landing page design. What we do now is distributing the inbound traffic among appropriate content.
You're right we should've set a spend threshold to some reasonable amount. On the other hand, we did want to know how keywords from a large semantic core would do. Also, the SEM agency did a really good scaling job 🙂
It's also worth noting we have a long activation cycle and even longer purchase cycle (the avg. time between singup and purchase is 182 days, median — 23 days). So we had to wait for quite a bit to understand the funnel changes. And that was an expensive wait. That's another thing that led us to the understanding that we shouldn't consider far-off conversions.
Just another question. Where did you spread/distribute the landing page other than though google ads?
No, our goal for this launch was to check Google Ads as scalable acquisition channel.
Have you tried retargeting?
Hi Virgil. Sure, there were retargeting campaigns set up. The point is that if you don't convey value propositions on the first touch, retargeting effeciency suffers badly.
Awesome. We'll try it next month for FirstPromoter, any quick advice that you can share? We'll use PerfectAudience for now(any others we should try?).
Thank you
The advice is mostly about giving value (providing an A-Ha moment) first and then showing retargeting ads. We used AdRoll, but the quick googling identifies PerfectAudience as a good alternative, will take a look at it.
http://gainbits.com/adroll-vs-perfect-audience/
https://woorkup.com/adroll-vs-perfect-audience-retargeting-comparison/
Also,
— if a user visited a site but didn't sign up, it's worth showing banners with product benefits and customer stories
— if users did sign up, it's worth showing banners with the "ideal image" of how their life will be better with your product or after its implementation
That's really great info. Thank you!
This comment was deleted 4 years ago.
Tony, glad to see you around the thread. Technically, the answer to your question is in the title of the article 🙂“Wasted” literally meant “almost zero” for us. That’s why we wanted to share the experience and minetraps to avoid.
This comment was deleted 4 years ago.
Tony, just to summarize (idk the product you want to run Google Ads for, so):
— Ensure you can track user activity (analytics platform is in place)
— Test Google Ads or FB Ads keyword-based to get higher CTRs
— Depending on which ad wins the experiment, either update existing content or craft a new piece/landing
— Slowly expand your budget
This comment was deleted 5 years ago.