Absolutdata and the art of the next move

A hotel can know a surprising amount about a guest who will never come back. It knows when the reservation was made, when the guest arrived, and that there was, at least once, a reason to choose this hotel. The difficulty is turning that memory into another booking. A database is very good at remembering. Persuasion is a different occupation.

• The job: turn customer and business data into a recommended next action.
• The machinery: NAVIK software plus analytics and data science services.
• The useful lesson: test the recommendation against ordinary practice before scaling it.

The guest who stayed once

In a 2017 account of Absolutdata’s hotel experiment, a global brand prepared a dozen offers for guests who had stayed once. One group received offers assigned randomly. In another, NAVIK MarketingAI matched particular offers to particular customers. The experiment changed the allocation of invitations, a small intervention in a business where rooms and guests are expensive to acquire.

Absolutdata reported 31% more bookings and 51% more revenue from opened emails in an eight-week controlled test. Those are distinct measures. Revenue from opened emails is a narrower denominator than all campaign revenue. Neither figure tells us the profit after discounts. Still, the experiment gives the reader something more useful than a general declaration that AI improves marketing: a decision, an alternative and a period of observation.

This is the territory Absolutdata chose: the gap between knowing what a customer has done and deciding what a business should do about it. Its software addresses sales, marketing and research; its services supply the data science and business interpretation around them. The buyer is an enterprise team with decisions to repeat, data to connect and an appetite for measuring the difference.

A company named over drinks, built on questions

The beginning was considerably less solemn than most enterprise software biographies. A 2013 account places Anil Kaul, Sudeshna Datta and Suhale Kapoor at Delhi’s Golden Dragon restaurant, discussing their proposed market research outsourcing company over Absolut vodka. Datta supplied the name. Corporate naming committees have achieved less with larger budgets.

Founded in 2001, the company brought a research background to what would become an AI proposition. Kaul had worked at McKinsey and held graduate degrees in marketing from Cornell. In a 2010 interview, he described building the business with Datta, his wife, and Kapoor. His professional subject was how buyers behave, a useful starting point for a company that would eventually recommend what to offer them.

By August 2012, Fidelity Growth Partners India had invested $20 million. The stated uses were concrete: expand global delivery and hire people for advanced and big data analytics. The company then employed 275 professionals. The funding story was about supplying expertise at greater scale, years before NAVIK became a family of AI applications.

Monday morning, specified

A sales representative faces a version of the hotel’s problem every week. There are more possible calls than useful hours. A CRM can hold contacts, purchases and opportunities. Deciding which person to approach, with which product and which argument, remains work.

NAVIK SalesAI, launched in 2016, packaged that work into a weekly plan. It prioritized contacts, suggested likely purchases and recommended communication channels. The dashboard also showed reasons for its suggestions. That detail matters: a salesperson can inspect an argument, challenge it and decide whether it suits the account.

The launch announcement described a pilot involving 50 salespeople across six territories at a Fortune 500 industrial products company. It reported a 4% sales uptick within seven weeks. The later platform announcement separately reported 2% more sales than a control group. The two percentages should remain separate; a change in sales and a comparison with peers answer different questions.

“Infogain now offers its first SaaS platform, NAVIK AI”Ayan Mukerji, December 2020 acquisition announcement

For a buyer, the appeal is the fit with an existing workflow. The person still makes the call. The software supplies a sequence and its reasoning. That makes adoption part of the product’s job: a recommendation that never reaches a conversation has little opportunity to earn its keep.

Research escapes the filing cabinet

NAVIK’s product history follows neighboring decisions. Converter addressed free-to-paid subscription conversion in 2015. Concept Test addressed product ideas. The 2017 platform gathered their successors, MarketingAI and ConceptAI, alongside SalesAI. Absolutdata’s expertise joined predictive modeling, data integration and research methods to business questions specific enough to act on.

ResearchAI arrived in December 2018. Its modules covered concept testing, pricing, segmentation and brand tracking. Research Guru used natural language processing to interrogate existing reports. The proposition was appealingly unglamorous: make accumulated research easier to use, instead of leaving it as an expensive archive that someone vaguely remembers commissioning.

TradeAI followed in 2019, bringing the same decision emphasis to consumer goods promotion calendars at the retailer level. The company also broadened its inputs: SafeGraph location data in 2019, and a Nielsen Connect partnership in 2020. ASK NAVIK, a business assistant for searching dashboards, databases and reports, was part of that Nielsen integration.

The price is more than the software

Absolutdata occupied a particular place between analytics consulting and enterprise software. Public descriptions offered pre-built SaaS applications, platform licensing and customization, backed by service teams. Customers could buy reusable machinery while still drawing on people who understood their data and business requirements. The organization around the model was part of the offering.

The practical alternatives depend on what a buyer needs. An analytics services firm such as Mu Sigma or Tiger Analytics is one route; an internal team working with a platform such as Dataiku is another. Those choices distribute the work differently. Absolutdata’s attraction was the combination of a decision platform and specialist delivery. A comparison should therefore ask who handles integration, model maintenance and the business workflow, alongside the software.

For a prospective customer, a sensible cost calculation includes implementation, data preparation, staff time and the economics of the recommended action. A discounted hotel booking may increase revenue while consuming margin. This is an editorial buying test, not a statement about Absolutdata’s contract terms: insist that the value calculation follows the money all the way to the business result.

Why an engineering company wanted the model

Infogain announced the acquisition on December 24, 2020. The fit was understandable. Absolutdata brought NAVIK and analytics expertise; Infogain brought a broader software, cloud and experience engineering organization. A recommendation needs systems that can deliver it. Joining the two businesses gave their customers a wider set of capabilities under one parent.

Funds advised by Apax announced an agreement to acquire Infogain in June 2021. By 2026, the parent was operating as Tenarai, presenting itself as an enterprise AI acceleration company. That is the current parent-company context. NAVIK’s earlier launches and pilots belong to their own dates, and are best read as the evidence behind Absolutdata’s original proposition.

Borrow the experiment

The transferable lesson is pleasantly manageable. Choose one repeated decision: which guest receives which offer, or which account gets the next call. Define the outcome before the trial. Keep a comparison group where practical. Record whether people followed the recommendations. Separate additional revenue from additional profit.

The logic depends on conditions. The data must connect to the decision; the business must have offers or products worth recommending; someone must act; and outcomes must be observable. Where those conditions are absent, an elegant model can produce an elegant suggestion with nowhere to go. The useful question is therefore quite specific: which decision will become better on Monday, and how will we know?