DIY Analytics

Thinking about building your own analytics? Don’t.

Original article: https://amplitude.com/blog/2016/04/21/thinking-building-analytics-dont

 

One big-ass decision up in tha game of any company, especially tech g-units whoz ass have straight-up pimped out engineerin crews, is whether they should loot a analytics solution or build it theyselves. In some cases tha fervor ta build all dat shiznit yo ass seems almost religious — but there be straight-up nuff upsides ta rockin a third-party analytics provider fo’ yo’ infrastructure.

Although it may not seem like a funky-ass big-ass deal, especially early on, tha “build or buy” decision has a enormous impact on yo’ company’s productivitizzle n’ speed, n’ ultimately on yo’ mobilitizzle ta drive growth.

I can KNOW tha appeal of buildin yo’ own analytics solution — it is ghon be perfectly customized ta fit yo’ needs, n’ you’ll have complete control over how tha fuck yo’ data is handled. Y’all KNOW dat shit, muthafucka! Some tech giants like Airbnb, Zynga, n’ Facebizzle have built impressive data infrastructures, so why shouldn’t yo slick ass?

Unfortunately, nuff g-units overlook tha legit costz of buildin it theyselves (and I’m not just poppin’ off dollars). If you’re thankin bout or is up in tha process of buildin up yo’ own analytics, make shizzle you be thinkin bout tha ways dis could potentially hurt yo’ company.

Buildin analytics diverts resources from yo’ core product

If you’re bustin yo’ time thankin bout how tha fuck ta set up yo’ event data pipeline, dat means you’re not thankin bout yo’ user behavior data n’ how tha fuck it can inform yo’ thang roadmap.

Companies often underestimate tha time n’ resources dat go tha fuck into buildin analytics. They be thinkin they can throw all of they data up in Hadoop, or maybe Redshift, n’ be done wit dat shit. But fuck dat shiznit yo, tha word on tha street is dat if you wanna ensure dat mah playas at yo’ company can access tha data they need (more up in tha next section on why dis is so critical), it will take much mo’ time, or like scrilla buyin third-party visualization tools, ta make dat infrastructure usable.

On tha other hand, payin fo’ a analytics platform means dat yo’ data will automatically be self-service ta mah playas whoz ass needz it, allowin mo’ time n’ resources ta be all bout yo’ actual product.

Fareed Mosavat, whoz ass leadz tha growth crew at Instacart, faced dis same build vs. loot decision when he joined tha company fo’ realz. At tha time, Instacart was rockin a hodgepodge of self-made tools fo’ trackin n’ analysis yo, but dat shiznit was hard as fuck fo’ data end-users (like tha growth n’ marketin crews) ta access dis data theyselves. In addition, they was reachin tha point where they system wasn’t goin ta work fo’ theygrowin data volume, n’ would require hella mo’ investment on they part ta scale.

Fareed decided ta go wit Amplitude fo’ they analytics platform, saying, “I’m much mo’ interested solvin tha core thang problems than buildin technical infrastructure fo’ analytics.”

Without accessible data, yo’ company won’t gotz a thugged-out data-informed culture

In addizzle ta takin away focus n’ resources from pimpin-out yo’ product, buildin analytics make it incredibly hard as fuck ta make data accessible ta tha rest of yo’ company.

Accessibilitizzle means dat mah playas at yo’ company (meanin fo’sho, even one of mah thugs whoz ass don’t know any SQL), can explore yo’ data n’ answer they own thangs, doggystyle fo’ realz. A big-ass part of dis is data visualization yo, but static dashboardz aint enough — ideally, tha end user should be able ta slice dis data up in different ways n’ discover insights on tha fly. Lil’ Bow Wow Porterfield, co-smoker of Looker, puts it well: “Da right analytics infrastructure is one dat make it just as easy as fuck ta share insightful data visualizations (graphs, charts) as it is ta dig down tha fuck into da most thugged-out granular details.”

Yo ass might be wonderin why data accessibilitizzle is so blingin. If yo’ company has a analyst or data science crew whoz ass can write SQL fo’ tha thang n’ marketin crews when they need lyrics, isn’t dat enough?

build vs loot infographic cta

Fareed from Instacart puts it perfectly: “If you say you’re data-driven but every last muthafuckin thang has ta go all up in a analyst, you’re not straight-up data-driven.”

If data is siloed ta tha analytics or data crew, then all dem fools dat needz a answer (even suttin’ simple, like yo’ everyday actizzle playas segmented by hood), is goin ta gotta go all up in a analyst. This creates bottlenecks n’ longer turnaround times, meanin dat (1) yo’ data scientists is backlogged wit requests n’ thus less productive, n’ (2) yo’ entire company loses momentum. Which brangs our asses ta our next point…

Yo crazy-ass company loses speed

Many high-growth g-units operate on tha principle dat speed is tha definin characteristicseparatin successful g-units from tha rest of tha pack. If you wanna operate as quickly as possible, self-service data access be a MUST.

Self-service data eliminates any bottleneck between tha data n’ tha end users, shortenin tha time ta insight. This determines how tha fuck quickly you can move all up in tha cycle of thang iterations n’ improvements, n’ ultimately how tha fuck quickly yo’ company can grow.

Now, that’s not ta say dat raw data access n’ freestylin SQL don’t have they place. It’s bout enablin thang n’ marketin ta answer 90% of tha thangs they have theyselves up in a analytics platform. Da straight-up bangin, complex shiznit — tha other 10% — is still incredibly blingin, n’ it’s where yo’ data scientists can shine.

Yo crazy-ass data scientists waste time on “janitor work”

Choosin ta build analytics means dat yo’ data scientists n’ engineers do mo’ grunt work, n’ less core data science work fo’ realz. As Porterfield says, “Data crews too often create bottlenecks fo’ tha rest of tha company. IT shouldn’t be bustin tha work of librarians, retrievin n’ interpretin data fo’ dem requestin dat shit.”

build vs loot ctt

In addition, engineers will need ta spend time buildin n’ maintainin tha data pipeline infrastructure over time fo’ realz. All of dat data “janitor work” be a funky-ass big-ass time suck; up in fact, data mungin takes up 50-80% of data scientists’ time.

Instead, imagine a scenario where tha data pipeline n’ infrastructure is straight-up taken care of by a third party. No one has ta spend time or scrilla ta constantly maintain it, n’ on top of that, mah playas can answer they own thangs. Think bout how tha fuck much time yo’ data scientists now gotta focus on complex insights n’ problems.

Still thankin bout it?

Whether you’re evaluatin a in-house or third-party platform, here is all dem thangs ta ask yo ass:

  1. Dum diddy-dum, here I come biaaatch! Who tha fuck up in yo’ company needz access ta dis data?
  2. For dem people, do dis system make it easy as fuck fo’ dem ta explore tha data ad-hoc, visualize tha data, n’ pimp they own insights?
  3. (This one is specifically fo’ evaluatin third-party systems). Yo ass betta access tha raw data, n’ up in what tha fuck format, biatch? Do it require hustlin custom scripts n’ data cleanup ta extract dis data yo ass n’ git it up in a query-ready format, or is there a option fo’ data pipeline n’ warehousing?

An alternatizzle ta buildin in-house

If you’re thankin bout a third-party analytics solution, you probably already know dat there be a ton ta chizzle from. Many of y’all may be familiar wit self-service analytics tools like Gizoogle Analytics n’ Mixpanel. While these hook up tha basic needz of tha non-technical end user, you can’t dig straight-up deep, which means additionizzle custom analysis is ghon be required. Y’all KNOW dat shit, muthafucka! But fuck dat shiznit yo, tha word on tha street is dat straight-up gettin tha raw data up ta do dat analysis be a big-ass pain; you’ll either need ta devote resources ta extractin n’ transformin dis data, or collect tha same data separately tha fuck into a Redshift clusta — neither of which be a ideal setup. Yo ass deserve mo’ betta n’ shit.

Unlike other analytics solutions, Amplitude combines a gangbangin’ flexible, intuitizzle analytics platform wit raw SQL access via Redshift. This gives you tha flexibilitizzle ta work wit tha data however you want. Yo crazy-ass marketin n’ thang crews can use tha Amplitude intercourse ta answer tha vast majoritizzle of they thangs n’ track they core metrics fo’ realz. At tha same time, you git instant access ta yo’ data up in a thugged-out dedicated Redshift clusta dat we maintain – all you gotta do is log in. I aint talkin’ bout chicken n’ gravy biatch.  Yo ass can connect any SQL editor or visualization tool on top of yo’ cluster, like Wagon, Looker, or Mode.

Smoke a cold-ass lil culture of data

I don’t be thinkin I need ta convince mah playas on tha importizzle of makin sound bidnizz decisions based off data insights, n’ you can put dat on yo’ toast. But is yo’ company currently set up ta do that, biatch? Do mah playas have access ta data, biatch? Can they KNOW what tha fuck tha data means fo’ they own goals as well as tha company’s core metrics, biatch?

Yo crazy-ass chizzle of analytics platform, whether you’re buildin in-house or buyin third-party, has big-ass ripple effects on how tha fuck yo’ company operates, n’ ultimately on yo’ company’s success.

Leave a Comment

Your email address will not be published. Required fields are marked *