What's the Story on Mozscape Updates?

[randfish]

Hey gang,

As you may be aware, we were considerably late with our last index release. You have my sincere apologies for that and the apologies of the entire team. In the interest of transparency, I want to try to explain what's been going on.

Since stepping down as CEO, I've been asked to take on a few roles in the company. One of those is product architect (basically the product owner) of our Big Data team, who produces the Mozscape link index. For several years, that team has been almost exclusively focused on getting us closer to a near real-time indexing system that does not have scalability issues. Mozscape is currently smaller than our major competitors, and we're also often slower. Our metrics (PA, DA, MozRank, MozTrust, Spam Score, Social Data, etc) have been the unique value we provide, but it's not enough. We need to be competitive on size and freshness.

Building a raw link index (without processed metrics like PA/DA et al) is hard, but it's possible. Building a link index with those metrics is really tricky, and requires computer science knowledge and skills far beyond the scope of my understanding. That's what our team's been working on, and they've made some progress, but it's been slow, hampered by unknown unknowns, and materially hurt by a lack of experienced talent we can hire to help (we've had open job posts for years now).

In the meantime, our historic Mozscape index structure keeps encountering challenges - this latest round is still somewhat unexplained (we believe there's hardware issues compounded by how the system is architected to handle large domains, but there may be other issues). The team's struggled to split time between keeping the old Mozscape running and hunkering down to finish the new system. I'm trying to help them balance things as best I can, and we're going to be putting effort toward making sure we get index releases out on time. However, to do that, we'll need to scale down size, and then rebuild back up. We think we can do this while also improving the prioritization of which links we crawl (e.g. deeper on important domains that link out, less so on deep pages that don't link anywhere) so the index overall improves.

However, I don't want to minimize the risks - we may have some slow updates, some smaller indices, and some less-than-ideal data in the next one or two indices while we work to remedy this issue. I HOPE we don't, and that things actually get better immediately, but we can't promise that until the work gets finished.

TL;DR - Mozscape V2 is in development and will let us as big and faster as any link index. In the meantime, current Mozscape's having issues & we're making smaller indices in an attempt to diagnose and repair.

As always, thanks for your understanding, continued support, and if you have any questions, feel free to leave them below. I realize that this level of service/product quality is NOT OK, and I'm doing everything in my power to fix it.

[YairSpolter]

Thanks for the update, Rand, and good luck on getting Mozscape V2 up and running ASAP. I'm sure I'm not alone in wanting to see Moz emerge as the leader in this industry, bringing the tools up to par with the amazing community assets that you guys provide.

Looking forward to seeing the mustache disappear!

-Yair

[DonnaDuncan]

Hi Ken,

It's normal to see fluctuations in link counts from update to update, especially if a large number of your links are coming from relatively few domains. Take a look at this Q&A - http://moz.com/community/q/why-did-my-da-and-links-go-down. See if it helps shed light on why this happens.

[LesleyPaone]

Yeah basically on the split index. I think it would give people a better view of how they are actually doing without exposing their data across the internet. It also seems like it could be a way of more accurately determining the DA, PA, and trust as well. And could also add more value to the product. Right now the link index only picks up a portion of the total links and then tried to determine a ranking method off of that. But in GWT there is a totally different set of links usually.

I think if you merged the GWT data, GA data, and the Moz data together things could be very powerful.

I realize a lot of people want to see just what they can do to improve their ranking and nothing more. But I tend to look at SEO as a whole picture and just want more traffic. I could really careless if it comes from Google, Bing, or some guys blog. So I think a link section would also help too. This is what I imagine in my head. Another section in our account that scrapes the referral pages out of our GA accounts and shows us how our backlinks are performing over time. That would be nice, because part of what a lot of us do is buy links in some way or another. It would let us know at a glance if a blog we want to target again is still pulling the traffic it was before. Things of that nature. But I think a real value is putting everything under one roof. GWT and GA are totally separate, having the information from both of those platforms merged into one would be awesome for me.

[DonnaDuncan]

"our social channels and Q+A reaches the Moz audience, which is mostly marketers and very, very few software engineers (even fewer with the right big data skills)"

True. But we might know some. I know some. We're all well connected.

[randfish]

Unfortunately, our social channels and Q+A reaches the Moz audience, which is mostly marketers and very, very few software engineers (even fewer with the right big data skills). We've got a recruiting team that's going hard at this problem, though, and yes, we've just changed to accept remote candidates for technical roles, so hopefully that will help.

[randfish]

Well, there's probably only a few thousand people who've worked on big data issues of this scale and complexity, and hiring ex-search-engine employees who've already made their millions in stock is pretty hard

Thanks for the positive wishes - we'll keep working hard to try and get this to where it should be.

[randfish]

I'll ask Martin from the Big Data team to answer the first question - I believe we use a number of different languages across the various services that power Mozscape.

On the split-index, I don't think I quite understand? Do you mean there'd be a unique version of our index that crawls the links we see from GA referrers (via the connected profiles in Moz Analytics)? That would be possible, but a ton of infrastructure work, and we'd likely need to build a unique version of the index for each customer to keep that data private.

As for GWMT - we don't have access to those, as folks don't connect/oAuth their GWMT accounts to Moz.

[keboydjr]

So what are the accurate numbers, last time or this time? My total links went from 15,000 down to 3,000. I really depend on the consistency and whatever changes were made, problems encountered, have really screwed up my reporting to the executive team.

[MattAntonino]

A giant game of Whack-a-Moz ... yeah, that's fun until it's not. I can feel the ... what is it? Distress? Worry? I know you want these things to come out on time & exactly how you want but that's rare even in the land of Google & Facebook.

You guys have great company benefits. It's amazing you find it so hard to hire the right people and as you said, have had openings for 2+ years.

Glad to hear the issues are at least unrelated. That gives us hope for the site crawls & other bugs to get fixed whatever happens to the link index.

Thanks again for the update & good luck getting through it all!

[DonnaDuncan]

A suggestion.

Put out a call in Q&A and social channels. Be specific about the skill set needed. Link to job descriptions. Ask for referrals and shares. Use social channels to the do the same thing. Surely we can get a few viable candidates to step forward...

Edited for one final thought. Are you open to remote candidates?

[LesleyPaone]

I deal with a few enterprise grade clients on the server configuration end, just out of curiosity what language platform do you run on?

Also, while you are here I would like to make a feature request or inquire if it is in the pipeline.

I would like to see the index split into two subsets that are used to generate the rank / trust. I would like to see it use the GWT links and also the GA referrers. It could show a better scope and at the same time keep that data private.

[randfish]

Moz has a hybrid cloud we built ourselves on a datacenter with hardware we customized. We were previously using AWS, but moved to give ourselves greater control and huge cost savings.

[randfish]

My guess and hope right now is that we don't have another delayed index. However, based on the inconsistencies and not-totally-clear diagnosis of issues, I'd say there's some risk associated with the next index.

Re: crawl updates - those have been behind, but should be catching up. I think the Moz Analytics team said 2-3 more days for completion of all the lagging reports.

[randfish]

The good news is we do know what we need to do and we have the people at the architecture level to make smart decisions (IMO). However, finding folks to help us execute on this plan has been insanely hard. We've had open positions on this team for literally 2 years, so if you know anyone, send 'em our way! I'm certainly working my networks to try and help, too.

[DonnaDuncan]

So Rand, does the engineering collective know what needs to be done or do you need a technical architect to offer an assessment of the situation and recommendations? I don't have that skill set, but perhaps there's someone in your vast audience that does and would be willing to help out.

[LesleyPaone]

Just out of curiosity are you running on your own network or off of a cloud like AWs or Google?

[CFCU]

Hi Rand,

Appreciate the update.  Are we probably seeing another delay this round? Ive usually had my crawl updates by now.

Cheers for the good service all round.

[randfish]

Unfortunately (or maybe fortunately), those issues are unrelated. The campaigns getting behind on updating was due to some server configuration challenges as I understand it - somehow our monitoring falsely told us things were fine, and thus, even after we fixed the issue, we've been trying to play catch up on thousands of weekly/monthly reports. I'd seen an email from the team lead last night that we should be nearly clear - maybe a day or two away.

Re: memberlist - that's a separate team, and they're aware of it, so should get a fix out soon. I think one of Moz's challenges is that we've bitten off a bit more than we can chew (that's my opinion, not necessarily representative of the whole company) and with 6+ engineering teams handling 100s of operations across a giant software suite, it feels like we're always playing whack-a-mole. We fix one thing, and the next week, 3 more things we never imagined could go wrong do. It's a frustrating problem, to be sure, and I honestly think the only way we'll break out of the cycle is to cut back to fewer projects and staff up teams, which will take time, discipline, dollars, and some good luck.

[MattAntonino]

Great to read the update Rand! You know I'm passionate about Moz keeping up in index size so I'm happy to wait to get a better v2 as long as that wait is reasonable.

Is this causing other issues on the site?  The memberlist isn't updating and it feels like bits of the site here & there are either very slow at updating at the moment or just not updating at all. There have been quite a few Q&A questions about campaigns not running/updating/performing alongside this.  Are these related or should we be reporting them as they happen?

Thanks again for the update & good luck! It's a huge project but you have a great team to help accomplish it!

[randfish]

Yes - that should be possible and I will endeavor to make sure we're keeping the updates honest and correct for the future. Thanks Donna.

[DonnaDuncan]

Very much appreciate this update. Thank you!

As you might expect, some of us rely heavily on the Moz API update. Consistency and quality is more important than frequency and volume. So while I now can better appreciate some of the difficulties and obstacles the Moz team is facing, I would love it if you could give us more than a few hours notice when an update is going to be delayed. Is that possible?