Find­ers Keep­ers. Fix Site Search, Fix Sales.

Some things nev­er change

In our new nor­mal world of changes, uncer­tain­ty and weird­ness, it’s reas­sur­ing to know that some things will always stay the same.

• Wile E. Coy­ote nev­er catch­es the Road Runner.
• The Simp­sons will nev­er not pre­dict the future. 
• And site search will always be neglected. 

It’s not an irony, just the sug­ar rush feel­ing of find­ing low hang­ing fruit. This is a com­pet­i­tive advan­tage that’s there for the tak­ing. Evi­dence indi­cates your com­peti­tors are prob­a­bly doing it wrong. Let’s see the numbers:

• A look at the lead­ing 250 ecom­merce brands shows an aston­ish­ing 72% fail site search cus­tomer expec­ta­tions (Bay­mard). Num­bers go much high­er for the rest of us.
• 21% exit the web­site after two failed attempts (Econ­sul­tan­cy)
• 68% of cus­tomers con­sid­er the frus­tra­tion of poor search suf­fi­cient to aban­don (For­rester)

And if that’s not enough to see how big search is, check this out:

• 64% go direct­ly to search when they have an ​“I want to buy” moment (Google)
• Con­sumers who use search are 2.4 times more like­ly to buy (Demand­ware)
• Con­ver­sions near­ly dou­ble when vis­i­tors actu­al­ly find what they are look­ing for (Moz)
• 69% of online shop­pers go straight to the search bar when vis­it­ing ecom­merce sites, but 80% leave due to a poor expe­ri­ence (Nos­to)

The evi­dence is over­whelm­ing. If you want to geek out on this, check this over­whelm­ing and suc­cinct list of site search stats from Algolia.

And yet it remains one of the most neglect­ed areas of investment.

It’s not about size, either

Bay­mard ran a thor­ough bench­mark on the top 250 lead­ing US and Euro­pean eCom­merce sites, mea­sur­ing effec­tive­ness against their 40 rec­om­mend­ed UX guidelines.

The find­ings per category:

• “Exact” search queries (42% of sites have issues)
• “Prod­uct Type” search queries (71% of sites have issues)
• “Symp­tom” search queries (52% of sites have issues)
• “Non-Prod­uct” search queries (39% of sites have issues)
• “Fea­ture” search queries (22% of sites have issues)
• “The­mat­ic” search queries (36% of sites have issues)
• “Com­pat­i­bil­i­ty” search queries (30% of sites have issues)
• “Slang, Abbre­vi­a­tion, and Sym­bol” search queries (49% of sites have issues)

In their words:

This must be hard to fix though. And expen­sive. Right?

Wrong. It’s eas­i­er than it looks.

A guy walks out of a bar and finds his neigh­bor crouch­ing under the street­lamp, fran­ti­cal­ly search­ing for something.

– What are you look­ing for, Ned?

– My glasses 

– How on earth did you lose them here?

– I didn’t, but here’s where the light is!

Good expe­ri­ence is about shin­ing a light in all the right places, mak­ing it easy for cus­tomers to do what they came to do.

Luck­i­ly, this is get­ting eas­i­er by the minute. Tech­nol­o­gy is push­ing cer­tain areas exponentially. 

If you fol­low our agency, you know we’re skep­ti­cal about the Hype AI-Train. How­ev­er, we’re hap­py to admit that site search is an area where nat­ur­al lan­guage pro­cess­ing and arti­fi­cial intel­li­gence actu­al­ly comes in handy and can move the nee­dle. We’re all inclined to trig­ger behav­ior based on our avail­abil­i­ty bias. Tai­lored, fast and accu­rate results strike our love for auto­mat­ic deci­sion mak­ing. It’s use­ful and feels better. 

There are big advances in tools that, despite hav­ing been in the mar­ket for a long time, are now super­charged, avail­able to any­one, with reduced com­plex­i­ty and a price point with­in reach. Let’s see a few.

Bril­liant search should allow our cus­tomers to be free to look for what they actu­al­ly want, instead of just a prod­uct name, SKU or spe­cif­ic string of con­tent. With AI step­ping up to the plate, we’re see­ing a shift from mere key­word search­es to under­stand­ing user intent. This evo­lu­tion isn’t just about find­ing things quick­er —which is not bad at all— but about mak­ing things more intu­itive so you can search like you think and like you talk.

In a nut­shell, you can move beyond search­ing for what you want, and into what you want it for. From ​“Nike red shoes” to ​“I want com­fy red shoes to run in wet cold weath­er”. We’re mov­ing from sim­ple key­word match­ing to con­text under­stand­ing. And it will keep evolv­ing as the con­text includes our past pur­chas­es, activ­i­ty, sup­port requests or com­plaints, mar­ket­ing inter­ac­tions and more first-par­ty data. 

I know. None of us has Walmart’s bud­get, size or resources. For­tu­nate­ly, we don’t need to. Check out the much more down to earth Rodo​.com. Their bril­liant AI-pow­ered car rec­om­men­da­tions are not only enjoy­able for cus­tomers, but also a good media move that got them press exposure.

Made for me

Imag­ine search is fixed. Your cus­tomers now have a blaz­ing fast auto­com­plete, search by pic­ture, intent aware search­es and sug­ges­tions. What’s next?

A web­site made just for you. 

Advanced Com­merce, Algo­lia, Bloom­reach, Search­spring or Cov­eo, just to name some of the lead­ing brands, are com­pa­nies that make this pos­si­ble with lit­tle dev or main­te­nance effort. By mix­ing real-time cus­tomer data, behav­ior and prod­uct infor­ma­tion, these tools take care of tra­di­tion­al­ly annoy­ing prob­lems like pri­or­i­ti­za­tion in search results, cat­e­gories, list­ing pages of any kind or ​“fea­tured prod­ucts” sections. 

AI-pow­ered mer­chan­dis­ing will rearrange your prod­ucts for each vis­i­tor based on pat­terns spot­ted dur­ing nav­i­ga­tion. Every cus­tomer inter­ac­tion helps these appli­ca­tions become a lit­tle bit smarter and accu­rate. On top of that, you have the abil­i­ty to fine-tune the process depend­ing on your busi­ness spe­cif­ic needs or pri­or­i­ties. Push over­stock prod­ucts, pro­mote top sell­ers for easy wins, or give more rel­e­vance to under­per­form­ing prod­ucts you’d like to sell more. You have con­trol over huge sets of para­me­ters that, when mixed with your cus­tomer pref­er­ences and habits, cre­ate a per­fect­ly bal­anced lay­out that favors prod­uct dis­cov­ery and engagement. 

Find­abil­i­ty at its best. Tai­lored, smooth ux that makes users go ​‘Wow, this plat­form real­ly gets me,’ in a good, not at all Black Mirror‑y kind of way.

Con­ver­sa­tion­al shop­ping is com­ing, too. We’ll see high­ly enjoy­able, on-brand bots ask­ing you four or five ques­tions and return­ing a tai­lored, unique ver­sion of the web­site along with per­fect rec­om­men­da­tions from an expert vir­tu­al assistant.

The hard (but doable) part is feed­ing the machine

Arti­fi­cial Intel­li­gence is entire­ly arti­fi­cial and not very intelligent.

You are what you eat, and so is the AI. Feed­ing it requires con­sis­tent and qual­i­ty data. The more you feed it, the more accu­rate and use­ful it will be.

So the more infor­ma­tion you give them about your prod­uct, the bet­ter the AI will be at respond­ing, find­ing pat­terns, and get­ting results.

Depend­ing on what plat­forms you choose and how well equipped your team is, this path will be more or less chal­leng­ing. Ide­al­ly, a deep inte­gra­tion with your ERP can feed enough con­tent into the prod­uct indexing. 

Dream Thing

The ide­al set­up is to mix and match MACH soft­ware. API-first, com­pos­able ecom­merce plat­forms like Big­Com­merce, Com­merce­tools or Vtex have a flex­i­ble archi­tec­ture that can con­nect you to any­thing, in every pos­si­ble way. Add a PIM to the mix and you have a per­fect feed­ing loop. 

PIMs (Prod­uct Inven­to­ry Man­age­ment plat­forms) are extra­or­di­nary. They detach the prod­uct data – specs, copy, dig­i­tal assets, sto­ry­telling pieces, every­thing – from the ERP or ecom­merce plat­form. This has many advantages:

• Prod­uct data is cen­tral­ized, includ­ing trans­la­tions, bring­ing con­sis­ten­cy to every­thing you pub­lish and allow­ing you to edit once and update everywhere.
• You can eas­i­ly swap your ecom­merce in the future with­out los­ing a sin­gle bit of con­tent. You final­ly own the qual­i­ty con­tent that’s so hard to build over the years.
• PIMs give your mar­ket­ing and prod­uct teams inde­pen­dence with­out involv­ing them in bloat­ed ERPs with end­less fea­tures and so many things to break. 
• PIMs play nice­ly with automa­tions and mul­ti­ple chan­nels. So, from a sin­gle source of prod­uct truth, you can push your con­tent tai­lored to your web­site, mar­ket­places, dis­trib­u­tors or any­where you want. 
• And of course, you can tai­lor chan­nels to feed the machine.

Mak­ing a men­tal note to write more about PIMs in the com­ing weeks. Bot­tom­line: with flex­i­ble, AI-pow­ered search, smart auto­mat­ic mer­chan­dis­ing and sol­id prod­uct data you can com­plete­ly trans­form your customer’s buy­ing journey.

Yes we can

I’d like to debunk the idea that solv­ing this is too expen­sive or com­plex. I’m pret­ty sure search sucks for oth­er reasons:

• We’re too busy fix­ing oth­er tech and UX prob­lems because every­thing breaks all the time. The urgent stuff gets in the way of what’s important.
• With­out prop­er tools, search fea­tures are hard­er to exper­i­ment and attribute fail­ure or success. 
• Efforts go to sales, mar­ket­ing, or man­ag­ing products.
• Or per­haps we sim­ply haven’t paid any atten­tion to search, yet.

Also, very rarely do we get cus­tomers call­ing sup­port to say search sucks. They just leave in silence. If we’re reac­tive and pri­or­i­tize fix­ing only the loud­er prob­lems, this is one that might fall through the cracks.

I know. We just have oth­er pri­or­i­ties. There’s a lot to do. Yet, I’m pos­i­tive that deliv­er­ing world class search is doable with a small bud­get, some effort, and sol­id teamwork. 

In a nut-search:

• Choose a great search app to integrate.
• Make sure you feed the robot enough prod­uct data.
• Fine-tune results.
• And work out a fan­tas­tic, per­son­al­ized and smart results page.

It will be worth every pen­ny, and you will final­ly spare your users from the U2 dilem­ma.